The surfer was us. The wave was AI. My message: this wave is coming for our customers, our business, our society. Existential threat. Existential opportunity. You either drown or you ride it. There is no standing on the beach watching.

The room was energized. The timing felt right. AI labs were rewriting the rules every few weeks. AGI was around the corner. White-collar jobs were going to be eliminated in one to two years. Trillions of dollars flowing from the smartest investors in the world. Every conference, every newsletter, every board conversation carried the same message: move fast or fall behind.

We leaned in.

February and March: enabling our teams. Tools, training, AI-assisted workflows across every function. Goal: internal productivity — get everyone using this before it uses us.

February through May, simultaneously: rearchitecting our products to embed AI capabilities that would create real customer value. A parallel track. H2 rollout target. Full company mobilization.

April: the electricity hit. Teams were visibly more productive. Engineers shipping faster than we’d seen in years. The surfer was upright. It felt like the wave was exactly what I said it would be.

May: the first budget meeting that felt wrong.

API bills. Token costs. Inference at scale. None of it behaves like SaaS seats — it runs while you’re not watching, compounds while you sleep, surprises you on the 15th. We weren’t alone. FinOps Foundation found 73% of enterprises reported AI costs exceeding original projections in 2026. Uber burned through their entire AI coding budget in four months. Expensive company to be in.

June brought more budget surprises. Model limitations I hadn’t fully anticipated. And the quieter disappointment — the gap between what these models can actually do and what the keynotes suggested.

Six months of emotional roller coaster. Not as terrifying as the predictions. Not as exhilarating as the demos. But here’s what that middle place gives you: clarity.

So what actually happened?

1. Quick productivity gains are easy. Sustainable ROI is hard.

Only 18% of professional services firms actually track AI ROI. Firms getting real returns moved to value-based pricing before deploying AI — the business model determines the return, not the tool. Same technology. Completely different economics.

2. The AI budget is a new animal. Most CFOs haven’t met it yet.

Token costs, agent costs, inference at scale — none of it behaves like traditional software spend. One enterprise spent half a billion dollars in a single month after failing to set usage limits. Our May was a smaller version of that lesson. Treat AI spend as a separate P&L line or it will find you.

3. Your model providers are not your partners.

Alex Karp said it publicly last week: enterprise CEOs are privately livid about paying for tokens while AI labs absorb their intellectual property. We’ve lived a quieter version of that frustration. Models we’d built workflows around — proven, affordable, working — started getting deprecated. Migrating to newer, more powerful models means higher costs, revalidating everything we built, and no guarantee the next generation won’t be deprecated in six months. There is no end in sight. For professional service firms, your client data is the alpha. “Controlling your weights is controlling your fate.”

4. The entry-level pipeline is quietly narrowing.

Lawyers and accountants aren’t being replaced. But workers age 22 to 25 in the most AI-exposed roles have seen a 13% employment decline since 2022 — not because they’re fired, but because firms stopped hiring them. Future partners and senior advisors are a shrinking cohort. Nobody’s saying this at conferences yet.

5. Adoption is table stakes. Quality of output is the differentiator.

Thomson Reuters’ 2026 report: 79% of legal professionals use AI, 69% of accounting professionals. Results across the board: “marginal productivity improvements.” Everyone has the tools. Whether what comes out is good enough to put in front of a client — that’s the actual race now.

6. Governance failure is a client trust event.

KPMG quietly withdrew a major client report after discovering AI-generated hallucinations. Gartner warned in May that uniform AI agent governance leads to enterprise failure. Most firms invested in tools. Almost none wrote a policy for what gets reviewed before it leaves the building. Not caution — liability.

7. AI-native competitors are forming around your commodity work.

Not a large firm with a bigger AI budget. A small shop with the same foundation models, lower overhead, and no legacy process to defend — targeting structured, document-heavy, junior-dependent work. They’re already bidding against you.

8. Client relationships and the data from them — that’s the moat.

Every professional services firm now has access to the same Claude, GPT, Gemini. Firms that pull ahead will build proprietary datasets — client feedback, engagement history, relationship signals — that compound over time and can’t be copied by a competitor who just signed up for the same API.

9. The pricing model is breaking.

EY data: AI-augmented audits now complete 35% faster. Clients know this. Firms still billing by the hour for AI-assisted work are heading toward a conversation they don’t want to have. Every PS firm has a billable-hour problem hiding inside a productivity story. It just hasn’t arrived as an invoice dispute yet.

10. AI is raising what clients expect — without raising what they’ll pay.

Clients know you have the tools. Faster delivery, deeper insight, more strategic counsel — that’s the baseline now, not the premium. Firms still delivering at 2022 speed with 2026 pricing find out at renewal. Not in a confrontation. In a silence.

January was: ride the wave at all costs. H2 is: read it first.

Three adjustments. We consolidated our internal AI tools from two to one — simpler to govern, easier to track. We set an upper limit on token spend and have a designated leader watching usage like a hawk. And we shifted the filter for every new AI investment: does this create measurably better outcomes for customers? If not, it doesn’t justify the cost. Every good investment in H1 had a specific problem, a clear owner, and a human review before anything reached a customer.

Not as scary as January suggested. Not as simple as the demos made it look.

The wave is real. I just know more about how to read it now.

Two weeks before that, the Commerce Department pulled Anthropic’s most capable models for entire user categories. No explanation. No timeline. Just pulled.

Everyone in enterprise software is treating this as a plot twist. I’ve been waiting for it since January.

What my customers told me first

Early this year, I sat down with a group of customers to understand what they needed from us as we rearchitected our platform for AI agents. Expected to hear about features. Response times. Integrations.

What I heard: trust. Security. A privacy track record that could hold up when their most sensitive client data is inside the model.

Not talking points. Hard requirements. Specific questions about where data flows, what gets retained, who has access to what. And then one question I didn’t expect: which company is actually running the AI.

One customer said it without hesitation. Don’t pick that model provider. Their CEO keeps talking about human extinction. We don’t trust him.

I didn’t argue. I wrote it down.

No benchmark comparison. Just a direct statement about a CEO’s public behavior and what it signals about the company behind the model.

They’d already made the call. Just waiting to see if I had too.

Why clients got there first

Regulators ask: what’s the risk to the system?

Clients ask: what’s the risk to me?

And in professional services, where a managing partner’s name is on every client deliverable, that second question gets answered faster, with more urgency, than any government review board.

Thomson Reuters asked thousands of PS professionals the same question this year. 85% of clients want their firms to disclose which AI they’re using. Fewer than one in three actually know. That gap is exactly where trust gets destroyed quietly, before anyone has to call it a crisis.

The clients asking hard questions about model providers in January weren’t being paranoid. They were pricing in risk before the market did.

Government just caught up.

The treadmill

Here’s the part nobody in enterprise software wants to say out loud.

Worst model is always the current one.

Everything ahead will be more capable. More regulated. More expensive. And the useful shelf life of each generation keeps shrinking. One major provider launched a new flagship model this month at double the previous pricing — ten days’ notice before existing plans stopped covering it. Another provider’s most powerful models got pulled by a federal agency for users in certain geographies. No warning.

Model providers are in a capability race. Capability attracts regulation. Regulation restricts access. Restricted access creates pricing power. Pricing power funds the next leap. Repeat.

You’re not picking a vendor. You’re picking a treadmill. Speed gets set by people who aren’t thinking about your renewal cycle or what you committed to a client last quarter.

Open source is a real alternative. I looked hard at it. But the economics of running it at enterprise scale aren’t obvious enough yet for me to stake the platform on. Watching closely. Not ready.

What we actually decided

When we made our model decision, I kept returning to the same questions. Which provider meets the security and privacy requirements our customers named explicitly? Which one extends the infrastructure we’ve already committed the business to? What happens to our clients if that provider changes pricing overnight, or if a government agency decides their models can’t reach certain markets?

If you haven’t asked yourself all three, you haven’t made a decision. You’ve let your engineers make one for you.

One answer covered all of them.

We chose the provider already powering our infrastructure. Not the highest benchmark score. The one our customers would accept — and the one we could defend out loud when asked.

We use a different provider internally for employee productivity. That’s a different trust context. My team’s workflows aren’t client deliverables. Different bar.

The question behind the question

When a managing partner at a 200-person AEC firm asks which AI powers your platform, they’re not asking a technical question. They’re asking whether your judgment about risk is something they can rely on — the same way they rely on your judgment about everything else you do for them.

Most SaaS CEOs are treating model selection as an engineering call. Their clients stopped treating it that way sometime last year.

Government caught up this week.

Your clients were the first regulators. They made the call before any of this was a headline. Before Washington.

Only question is whether you were listening then — or starting now.

My honest answer: I hadn’t heard their name from a single customer or prospect in our pipeline. Not once.

When I looked them up, I understood why. They’re a horizontal platform — their model is to integrate data from every system at an enterprise, layer AI on top, and sell the output. No deep vertical. No embedded workflow. No domain expertise built over years in a specific industry. Just data aggregation and an intelligence layer.

We’re something different. Deeply verticalized. Intelligence is an add-on to the core value we deliver to a specific customer in a specific workflow. That distinction matters. And not just for us.

Every market narrative right now is pushing SaaS CEOs toward a binary that isn’t real. AI-native platforms command premium multiples. AI-powered is the consolation tier. Get with the program or get left behind.

Notion didn’t get that memo. Neither did Atlassian, Monday.com, Canva, or Adobe — which added Firefly on top of Creative Cloud and didn’t apologize for it. These are not companies in retreat. They’re some of the most valuable software businesses on the planet, and AI makes them better without defining them.

There is a real category of AI-native software. Cursor is an AI-native code editor — remove the AI and the product doesn’t exist. Harvey built a legal reasoning engine where the intelligence is the product. Glean is enterprise search built entirely on models trained on your company’s data. These companies exist because AI created a category that didn’t exist before.

Real, valuable, and smaller than the narrative implies.

Trying to become it when your business is something else — burning 18 months rearchitecting a platform that works, chasing a multiple before validating whether your data can actually support what you’re building — is not strategy. It’s panic with a roadmap attached.

ChatGPT can write a blog post in 30 seconds. People are still writing their own. Our world doesn’t operate in binaries. It operates in grey.

So the real question isn’t AI-native or AI-powered. It’s: which one are you, and do you know why?

Most CEOs don’t. Not because they haven’t thought about it — they have, constantly — but because they’ve been thinking under pressure without a clean framework for the decision. So they do what people under pressure do: they copy the most expensive-looking option.

Here’s the audit I ran on our platform. Four questions, in order.

1. What data do you actually own?

Not what you have access to. What is proprietary, rights-secured, and unique to your platform? Public data doesn’t compound. Semi-proprietary data doesn’t compound. Only signals that exist because customers use your platform — and only yours — build a flywheel. Map every data asset. Classify each one honestly. Most CEOs find less in the proprietary column than they expected.

2. Where do your customers seek value you can monetize?

Not where they use your product. Where they would pay more if the product were better. Find two or three workflows where your proprietary data makes AI materially better than any generic alternative. Name them specifically — if you can’t, you’re not ready to build.

3. Can you deliver that value better than the alternatives?

Your customers have options. They can wait for a horizontal platform to solve the same problem. They can build something themselves. Your moat question is whether your data, domain knowledge, or workflow depth gives you a durable edge. Speed erodes. Data compounds. Relationships compound. Vertical depth compounds.

4. What does it actually take to build it?

Before committing, prototype. One workflow. One agent. Real customers. Measure whether the outcome actually improves. Companies I’ve watched burn capital on AI transformation mostly skipped this step — they modeled the destination before validating the path.

Run those four questions honestly and the call usually makes itself.

If your data is so central to the product that removing the AI removes the product — and every customer interaction generates a training signal that makes the system smarter — you may be building toward AI-native. Real destination. Real architectural commitment. Real price tag.

If your platform delivers genuine value independent of AI, and AI makes that value better, faster, or more defensible — you’re AI-powered. Own it. Your moat is the embedded workflow, the vertical expertise, the customer trust built over years. AI sharpens it. The platform is still the platform.

We ran the audit. We came out AI-powered. Our platform has real value to customers whether or not AI is in the picture. That’s not a concession. It’s a foundation most AI-native startups are still trying to build.

Before you make the call

• Have you mapped your data assets and classified them — public, semi-proprietary, or uniquely yours?

• Can you name two workflows where your proprietary data makes AI better than any generic alternative?

• Have you stress-tested your moat against horizontal platforms building toward the same problem?

• Have you prototyped one AI feature with real customers — not planned it, built it?

• Does your platform deliver value when AI is off?

• Are you pricing AI tied to measurable customer outcomes, or bundling it in and hoping?

• Is your advantage vertical depth and domain expertise — or is it the AI itself? One of those is a moat. The other is a feature.

Most CEOs can work through those questions. Fewer do it before the fear arrives.

The fear is real. FOMO hits SaaS businesses the same way it hits the stock market — it makes disciplined operators do things they’d never do in a calm room. Chase a multiple before validating a moat. Build for the narrative instead of the customer.

Two questions cut through the noise: does what you’re building create surplus value your customers couldn’t get elsewhere? And is that value durable — compounding, deepening, harder to replace over time?

That board member’s AI-native competitor hadn’t shown up once in our pipeline. Not because they weren’t real. Because they weren’t solving a problem our customers had.

Our world doesn’t operate in binaries. Your strategy shouldn’t either.

SpaceX going public is the kind of milestone that makes you stop and actually read the document. I’d been following the company for years, but the narrative shifted significantly when xAI became part of the story. So I spent thirty minutes with the S-1.

I read it as a business builder, not an investor. Missions, TAMs, values — those are the foundations I care about. They’re what I think about when evaluating any company, including my own.

Thirty minutes in, my brain was spinning. Here’s my take on what I saw.

SpaceX’s stated mission, as written in the filing:

“to build the systems and technologies necessary to make life multiplanetary, to understand the true nature of the universe, and to extend the light of consciousness to the stars.”

Read that carefully. Not a mission to improve life on Earth. Not a mission to connect the unconnected or make space more accessible. Humanity’s home planet is framed as a single point of failure — a problem to be solved by leaving. That framing does something specific in the document. Creates a container large enough to hold rockets, satellite internet, a social media platform, an AI chatbot, and orbital data centers that don’t exist yet, and makes them all feel like chapters in one inevitable story.

When your mission is to extend the light of consciousness to the stars, a $26.5 trillion TAM sounds almost modest.

My question throughout: does each business actually earn that mission?

Rockets

SpaceX was founded in 2002 with one animating idea: make humanity multiplanetary by dramatically reducing the cost of access to space. Rockets are the direct, unambiguous instrument of that mission. Not rhetorical. It drove product decisions, twenty years of financial sacrifice, and breakthroughs most of aerospace considered impossible.

Falcon 9 reduced launch costs from $18,500 per kilogram to $2,700 — an 85% reduction. SpaceX now launches more than 80% of all mass to orbit globally. Revenue was $4.1 billion in 2025.

TAM for rockets is real. Also modest. Roughly $20 to $30 billion by the early 2030s. SpaceX dominates a market it largely created from scratch. You’re not betting on whether reusable rockets work. You’re buying hindsight.

Mission verdict: authentic. Mission and business are the same thing.

Starlink

Here’s where it gets interesting.

Starlink is the most compelling business in the S-1 — $11.4 billion in revenue in 2025, 50% growth year over year, $7.2 billion in EBITDA. Zero enterprise customers above $750K have voluntarily left. During the Maui wildfires in 2023, Starlink deployed 650 terminals when all other infrastructure failed. During Hurricanes Helene and Milton in 2024, it was the only communication system still standing.

What I noticed: the cosmic mission actually undersells what Starlink is doing. Connecting 3.3 billion underserved people to the internet is a genuinely important earthly mission — and it doesn’t need to borrow from rockets.

And this TAM is Musk-independent. Teledesic, backed by Bill Gates, attempted LEO satellite internet in 1990 before going bankrupt. OneWeb raised billions trying to bridge the digital divide. Amazon has committed $10 billion to its constellation. Demand was proven before Musk arrived. He executed better.

Great business. The cosmic framing just gets in the way.

AI

This is where I slowed down.

AI — Grok, X, orbital compute — claims a TAM of $26.5 trillion. Let me describe what believing that requires.

Total worldwide IT spending — software, hardware, data centers, services — is expected to reach $6.15 trillion in 2026. SpaceX’s AI TAM is nearly one entire additional US GDP. And the segment generated $3.2 billion in revenue in 2025 while losing $6.4 billion.

Orbital AI compute — data centers in space powered by the Sun — doesn’t exist yet. No precedent. No customers. Just the ambition.

Grok gets framed as a “truth-seeking AI built to help humanity understand the universe.” That description elevates what is currently a money-losing chatbot competing against ChatGPT, Gemini, and Claude into something operating on a different plane. That $26.5 trillion isn’t a market forecast. It’s a bet on civilizational convergence — one that makes normal scrutiny feel small-minded.

Mission verdict: cosmic cover for an unproven business.

The Musk Premium

Here’s what tied all three together for me.

For rockets: the market required Musk’s obsession to create, but the creation is done. You’re buying a proven position.

For Starlink: the TAM is Musk-independent. That problem was real before he arrived. Its business speaks for itself.

For AI: the TAM is the Musk premium in its purest form. That number requires his specific relationships — Tesla for chip manufacturing, X as Grok’s data engine, his political capital. Remove him and the AI TAM doesn’t get smaller. It becomes incoherent.

Before You Decide

SpaceX’s total S-1 TAM is dominated by that $26.5 trillion AI figure. Rockets: $20 to $30 billion. Real, proven, largely maxed out. Starlink: growing but bounded by actual connectivity markets. AI: $26.5 trillion, almost entirely unproven.

When north of 80% of a company’s claimed TAM is unproven, you owe yourself three questions.

First: do you believe in the mission? Not the cosmic framing — the actual business underneath it. Rockets earned that belief over twenty years. AI hasn’t started that walk.

Second: is the TAM real? Not just big — real. What part is proven by prior attempts, existing customers, or market data that pre-dates this company? For Starlink: most of it. For AI: almost none.

Third: are you buying at the right price? Nobody likes asking this in a FOMO environment. But when the valuation rests on a TAM that’s 4x all global IT spending, the margin for error isn’t a rounding issue.

For Musk fans, none of these questions land. They’re not buying a business — they’re backing a person with a worldview. That’s a coherent position. Just own it clearly.

For the rest of us — especially those watching AI hype warp normal capital discipline — know what kind of risk you’re taking. Not whether Musk is brilliant. Whether the market has already priced in the part of the story that hasn’t happened yet.

I run a company. Not SpaceX scale, but mission and TAM are the first two questions I ask about any business, including my own. Great mission and large believable TAM is the precondition to a great business. Not a guarantee — the necessary starting point. Get that right, then hire the right people, deploy the right capital, and execute. That’s where moats get built.

My company has a real mission and a real TAM — just not a large one. Mission plus TAM tells you the max enterprise value you can create. That’s the ceiling. Everything after determines how much of it you reach.

What the SpaceX S-1 shows is what happens when one company contains all three versions of that test. Rockets passed it twenty years ago. Starlink is passing it now. AI is still asking the mission to carry weight the financials haven’t earned yet.

Know which walk you’re paying for.

That’s not a confession. It’s where most SaaS CEOs are right now, whether they know it or not.

Our cloud costs are up 30 to 50 percent year over year. Some of that is growth. Most of it is inference — the cost of running AI features for customers. When my PE investors asked about our AI roadmap last month, I came with three buckets: internal productivity, engineering output, and product value. They nodded. Not because the numbers were large, but because the framework was honest.

What I didn’t have was a clean answer to a simpler question: what is our AI gross margin?

I knew our blended margin. I didn’t know the split.

Here’s why that matters.

Traditional SaaS is built on one of the best business models ever invented. Software costs almost nothing to replicate. You write the code once, sell it a thousand times, and gross margin approaches 80 percent because the marginal cost of the next customer is nearly zero. Every metric we use — gross margin, NRR, Rule of 40, LTV:CAC — was built for that world.

AI broke the assumption underneath all of them.

Inference costs money every time a customer uses a feature. Not a fixed amount — a variable one that scales with usage. ICONIQ’s 2026 data puts inference at roughly 23 percent of revenue at scaling-stage AI companies. For traditional SaaS adding AI features, the margin hit is 12 to 17 points immediately, without raising prices. AI-native products run at about 52 percent gross margin in 2026, against the 70 to 80 percent traditional SaaS has always targeted.

Add an AI layer to a SaaS business without separating the P&L and you’re adding manufacturing economics to a software business. Then reporting only the software number.

Salesforce found out what this looks like at scale. Agentforce processed nearly 20 trillion tokens against $800 million in ARR. Customers ran 80 percent fewer human seats — and got billed 83 percent more. Per-seat metrics stopped working. NRR became a function of consumption patterns, not renewals. Forecasting got harder.

We’re not at that scale. But the dynamic is the same — just smaller and earlier.

The pattern shows up in aggregate too. Median Rule of 40 across public SaaS is 28 percent today. Only 20 percent clear the threshold. The floor didn’t drop because growth slowed — it dropped because AI spend is hitting margins faster than AI revenue is arriving.

Eighty percent of M&A buyers cite AI commoditization as the top risk to SaaS valuations. Twenty-five percent of CEOs do. That 55-point gap doesn’t close with better strategy. It closes when the P&L forces the conversation — at a board meeting, in an LOI, or when a quarter comes in that nobody can explain cleanly.

My forcing function is next quarter’s feature rollout. Others find out later.

What I’m doing right now is simpler than it sounds. I started tracking AI inference cost as its own line — separate from everything else. That one number, once you have it, starts to reveal what the blended dashboard hides. What it’s doing to gross margin. Whether NRR is healthy because customers value the product or because they haven’t hit their usage ceiling yet. Whether the AI layer is getting more efficient as you scale or just getting bigger.

Three numbers I didn’t have six months ago that I want every month now: AI product revenue divided by inference cost — how much revenue for every dollar of inference spend. AI COGS as a percentage of AI revenue — whether margins are improving or just growing. And gross margin by revenue stream — legacy SaaS and AI layer reported separately so you can see which business is actually healthy.

One question before the next AI feature goes in: can you draw a straight line from this spend to GRR, NRR, or gross margin? If you can’t draw it, it’s a bet. Name it as such.

None of this is complicated accounting. It’s a decision to look.

Separating the AI P&L from legacy SaaS isn’t a finance exercise. It’s the difference between managing the business and being surprised by it.

The split doesn’t create the problem. It just tells you how long it’s been there.

I had the wrong worry.

Safety warnings got through. They just went somewhere I didn’t expect. They became the pitch.

Every major shift in Sam Altman’s public narrative over the last five years arrived exactly when OpenAI needed the world to believe something specific. Not when the technology changed. Not when new evidence emerged. When the capital event required it.

Most observers call this maturation. Altman is “evolving,” “navigating complexity,” “balancing competing pressures.” There’s a more precise word for it: calibrated.

Here’s the sequence.

2021–2022: The Prophet. OpenAI as humanity’s guardian. AGI is coming — transformative or catastrophic. We’re the right people to build it carefully. Safety framing makes OpenAI the responsible adult in a room full of irresponsible actors. It also justifies the capped-profit structure that allows fundraising while claiming non-profit mission. One narrative. Two functions.

May 2023: The Regulator. GPT-4 launched. ChatGPT at 100 million users. Microsoft deal deepening. Altman testifies before the Senate Judiciary Committee and says something remarkable: please regulate us. He calls for a global licensing agency with authority to revoke AI licenses. Manipulation of elections is his “area of greatest concern.”

Press coverage called it humility. It was positioning. Write the regulatory framework and you become the standard against which every competitor is measured. Safety becomes a moat.

November 2023: The Survivor. Four days. Altman is fired for being “not consistently candid.” Eight hundred employees threaten to walk. He returns. Out go the safety researchers. In comes an establishment board. Safety advocates lost. Commercial momentum won.

2024: The Empire Builder. Safety language stays. Emphasis shifts entirely. “Intelligence too cheap to meter.” AGI by 2025. On October 2, 2024, OpenAI closes a $6.6 billion raise — the largest venture round in history — while planning to lose $5 billion that same year. I’d never heard of that kind of capital behind a pre-IPO company. No one had. His logic: win first, safety follows. Scale becomes the safety argument. Every additional dollar of investment means humanity is more protected.

2025–2026: The Reluctant Capitalist. OpenAI converts to a Public Benefit Corporation. Non-profit mission becomes a $130 billion foundation stake in the for-profit arm. On going public: “Am I excited to be a public company CEO? Zero percent.” On critics: “I would love to see them get burned.” Senate humility is gone. Mission still gets invoked — it just sounds different attached to a trillion-dollar valuation.

Map any of these shifts against what OpenAI needed commercially at that moment.

Non-profit framing justified the capped-profit structure that made fundraising possible. Senate humility arrived when regulatory capture was the strategic priority. Scale-as-safety arrived when the largest capital raise in venture history needed a story. For-profit conversion arrived when public markets became the only path to the next funding cycle.

This isn’t a criticism of intelligence. It’s the opposite. Altman is one of the most effective narrative operators in the history of technology. Every pivot lands well-timed, well-worded, and believed by enough people to work. The question isn’t whether he believes it. The question is why he believes it at exactly the right moment.

An IPO will be the first time the narrative meets a number the public can read.

Quarterly earnings replace quarterly vision. Revenue, churn rate, margin on inference costs — these replace “we are building AGI for the benefit of humanity.” Public markets don’t negotiate with mission statements. They price the delta between what was promised and what was delivered.

OpenAI has promised AGI, abundance, safety, and humanity’s salvation. Quarterly reports will cover API revenue and ChatGPT subscriptions.

Playbooks for this moment are familiar. Visionary founder takes company public on the strength of a transformational narrative. Numbers prove harder than the story. Founder transitions to executive chairman. New CEO hired to manage the business. Founder becomes a billionaire investor with a portfolio of new bets. Promises made in the roadshow become someone else’s problem to explain.

Not predicting failure. Predicting the standard exit.

Two years ago I was worried the safety warnings weren’t getting through.

They got through. They just got monetized first.

An IPO doesn’t end the story. It just ends the part where he controls the narrative.

There’s a pattern I’ve noticed watching Sam Altman, Dario Amodei, and Elon Musk over the last three years. Their message about AI has shifted. Not gradually. Specifically. And the shifts line up with something that has nothing to do with the tech.

I’m an Anthropic customer. Building on their platform, not watching from the sideline. Dario’s public message has been striking to track. In 2022, he was warning about civilizational risk. By May 2026, onstage alongside Jamie Dimon at Anthropic’s financial services briefing in Manhattan, he was invoking the Jevons Paradox to explain why AI job displacement is “net positive.” Real economics. But the timing of when he found it useful is the tell.

Altman told a Senate committee in May 2023 he was “afraid” of what he was building and called for government regulation. Two years later, in a 2025 essay titled “The Gentle Singularity,” he described AGI arriving at “a remarkable moment” leading to “a future of abundance.” Musk called AI “one of the existential risks we face” at the UK AI Safety Summit in November 2023, then automated 100,000 federal jobs through DOGE. Their models didn’t change their minds. Something else did.

The Reasonable Defense

They’re updating as they learn. That’s the standard argument. Not unreasonable — models improved faster than anyone predicted, and the risk calculus changes as capabilities evolve.

Partly true.

What doesn’t hold is the direction. Genuine updating produces more nuance as situations get more complex. Their messaging got simpler, and more optimistic, right as each needed outside capital or regulatory goodwill. OpenAI was raising at a $157B valuation. Anthropic crossed $20B+ in ARR. Musk needed a story for Tesla’s AI division. Fear went away faster than risk did.

Not updating. Audience selection.

Three Years. Three Shifts.

Watch the arc.

2022–2023: all three are warning about existential risk. Altman testifies before Congress. Dario publishes essays on catastrophic outcomes. Musk co-signs open letters calling for pauses. Consistent message — this is dangerous, move carefully.

2024: risk still gets mentioned, usually in a subordinate clause. Headlines shift to enterprise partnerships and productivity gains. I was building on Anthropic’s platform that year — hallucinations were a real problem, and we made a deliberate choice not to deploy AI in any customer-facing way that could break trust. The company briefing enterprise customers was navigating different pressures than the company that published the safety papers.

2026: Dario is onstage at a JPMorgan briefing explaining why job displacement is fine, actually. Altman is publishing “The Gentle Singularity.” Musk is running an AI-powered government efficiency operation. Existential risk framing is still available when regulators are in the room. For investors, it’s about upside.

Same people. Same technology. Different rooms.

The Board Meeting

Last week I had a board meeting. One key metric is underperforming in the first half of the year.

My job in that room was to say where we are, what we plan to do, and what we don’t yet know. I flagged that our plans have uncertainties and are macro dependent. Laid out two or three scenarios that could close the gap. Not a comfortable message to deliver. You’re sitting across from people with capital at stake, telling them the honest version — including the part where you’re not certain it works.

Here’s what I know: the alternative is worse. Soften the message, lean into the optimistic scenario, and the board feels better walking out. Six months later, you’re explaining why the surprise happened. Boards can forgive bad news. They don’t forgive hidden bad news. Trust breaks in a way that doesn’t come back.

So you tell the truth. Flag the uncertainty. Give them the scenarios, including the one where things don’t recover fast enough. That’s the job.

The Scale Problem

I shape the message for the room too. Emphasize what’s relevant to employees, what matters to investors, what’s actionable for customers. At company scale, that’s just communication — leading with what’s relevant, not hiding the rest.

The room is everyone. Every worker wondering about their job, every government deciding on policy, every parent thinking about what their kids are walking into. They all took the 2022 warnings seriously. Career pivots. Retraining programs. Policy positions built around the fear. When fear became inconvenient, those people didn’t get a memo. They just noticed that the story got better right as the fundraise closed.

Not audience segmentation. An audience you forgot you had.

The ignored audience calls you a fraud. Not unfairly.

The Standard

Every founder and CEO is meant to be an eternal optimist. That’s not a flaw — it’s the job. Frame hard truths in the best possible light. Lead with the mission and the art of what’s possible. Sugar-coat it if you need to. What you cannot do is change the underlying facts to fit the room.

There’s a difference between “we’re behind on this metric, here are three scenarios to close it, and I believe we get there” and “the metric doesn’t matter the way we said it did last quarter.” One is optimism. The other is a different story. These three crossed that line — not gradually, but specifically, right as the investor changed.

A leader can choose silence. But if you open your mouth, the truth doesn’t get to change based on who’s in the room — especially when the room is everyone.

AI Advantage, the new podcast hosted by Baker Nanduru, CEO of ClearlyRated, explores how leaders across AI, AEC, accounting, construction, and technology are approaching implementation in practical, measurable ways.

Each episode dives into operational challenges, workflow optimization, client experience, data strategy, and the real-world decisions firms are making as AI becomes part of everyday business operations. From automation and project intelligence to efficiency and growth, the conversations focus on what’s actually working and where firms are seeing meaningful impact.

#AIAdvantage #AI #ArtificialIntelligence #Automation #ProfessionalServices #DigitalTransformation

His youngest team member — the most junior person on the floor — had at least $3M in stock options. That was late last year. It’s probably more now.

My take-home as a CEO barely covers my mortgage.

Those two facts exist in the same economy, in the same city, at the same moment in time.

Saturday mornings I run with a pace group. Engineers, founders, operators — people who built careers the right way. Last week nobody talked about training. Everyone talked about AI: what it’s doing to their jobs, their industries, their sense of where they stand. Fear. Envy. A particular kind of exhaustion that comes from running hard and falling behind anyway.

Everyone’s first read on this is wealth inequality. A story about unfairness, about who got lucky, about which lottery ticket you happened to hold when the numbers were called.

That’s the symptom. The disease is something different.

For most of Silicon Valley’s history, there were multiple paths.

You could join HP or Cisco in the late nineties and build a career there. Twenty years of good work, compounding loyalty, a retirement that felt earned. It had a logic: be smart, work hard, pick a stable company, and the math rewards you over time. Not always rich. But stable, respected, fine.

That bargain is structurally gone.

Google, Meta, Microsoft — companies that were once the definition of a good career bet — offer no stable jobs now. Everything else in tech is worse. SaaSocalypse isn’t a buzzword; it’s a business model reckoning that’s been running for three years. And sitting above all of it, disconnected from the normal rules of effort and time, is a cluster of AI-native companies printing wealth for their employees at a scale and speed the industry has never seen.

What broke isn’t the economy. What broke is the time variable.

A 2-year window at the right AI company now outperforms 20 years of disciplined career execution at the wrong one. When that math inverts, people don’t just feel behind. They start questioning whether the game is worth playing at all. That psychological rupture — the sense that rules changed mid-game without anyone announcing it — is harder to recover from than any market correction.

Most people’s answer to this: work smarter, pick better, adapt faster. If you’re falling behind, the solution is in your hands.

That’s wrong. Worth saying clearly.

This gap isn’t about effort or talent. My running group isn’t less capable than the 25-year-old sitting on $3M in unvested options. They made reasonable bets on reasonable companies using rules that were reasonable when they made them. What they didn’t have — what most people don’t have — is access to leverage.

Capital flows to leverage now. Not purpose. Not effort. Not mission. Not a 20-year track record of delivering results for customers who actually needed the product.

Markets aren’t moral. The gap between those inside the leverage machine and everyone outside it is no longer a matter of intelligence or discipline. It’s a matter of which side of a line you were standing on when the capital decided to flow.

So what does a CEO in my position actually do with that?

Here’s what it looks like at my company right now. We’ve cut travel. Delayed bonuses. No pay raises — not for anyone, including me. These aren’t choices I made lightly. They’re what the math requires in this macro.

And yet my team is still here.

Not because they have to be. Because they believe in what we’re building and they’re grateful to be building it together. I watch that every day and I don’t take it for granted for a single second.

That’s not a consolation prize. That’s a different optimization function — and anyone who’s been on both sides of it will tell you it’s the better one. Genuine belief in the mission produces more lasting satisfaction than money does. The person hitting $20M at 30 with no particular reason to get out of bed is going to spend the next decade figuring that out the expensive way.

Believing in the mission harder than ever isn’t the loser’s framing. It’s the right bet if you actually want a life worth living at the finish line — for yourself and for the people running with you.

Sometimes on those Saturday runs, between miles and conversations about job security and AI and who just got rich, I think about those people. Smart, capable, running hard. Trying to figure out which game they’re even playing anymore.

The $3M kid at the chip company isn’t the villain of this story. He’s just inside the machine. Most of the people I run with aren’t, and neither am I.

But purpose is the one variable that doesn’t get repriced by a venture capital cycle. It doesn’t evaporate when the leverage machine slows down. And it’s what makes the work mean something, regardless of which game the market decided to reward.

Capital follows leverage. Purpose follows meaning. You don’t get to choose which game the market plays.

You do get to choose which one you win.

“Our cloud costs are 20% above plan over the last three months,” he said. “You haven’t budgeted for what your engineers are spending.”

I started thinking. What’s the actual budget for this? How do I rank the ROI of AI against retention, expansion, the rest of the list? How do I price something when the cost is moving while I’m forecasting it?

My head started spinning. I owed myself an honest look at how I got here.

What changed in twelve months

A year ago, I could use AI tools without thinking about the bill. Tokens were effectively unlimited, the $20-a-month subscription covered it, and API costs were a footnote.

Now I get notifications four times a day that we’ve exceeded the Claude API budget. One of our developers burned more than $1,000 in a single week last month without realizing he was doing it. He thought he was being efficient.

The cost-per-engineer of AI is no longer a rounding error. It’s a budget line, a growing one, and it just blew a 20% hole in my quarterly cloud forecast.

The two bills I didn’t plan for

Every SaaS CEO right now is buying AI twice. We’re paying to make our teams faster (Cursor for engineers, Claude for ops, ChatGPT for everything), and we’re paying to put AI inside our own product through contextual copilots, auto-generated reports, AI-driven insights.

Both bills are big and growing. My budget isn’t.

So I’m forced to ask the question I’d ask of any other capital allocation decision: is this an investment, or is it a gamble?

An investment has a known cost and a clear expected outcome, sized to the bet. A gamble has neither, and gets sized by peer pressure. Most SaaS CEOs, me included in places, are gambling and calling it strategy.

Bill One — Internal AI

This is the bet I have the most data on, because the costs and benefits land inside my company.

The aggregate productivity data is worse than the headlines suggest. A February 2026 NBER paper found 90% of firms report no measurable AI productivity impact, against executive projections of +1.4%. My own data sits inside that aggregate. Some engineers ship faster with Cursor. Others don’t. Same uneven picture for CSMs working with Claude on QBR prep. The gains are real but rarely match the slide-deck claim.

What’s new this year is that I can finally see the cost, and it isn’t flat. One $1,000-a-week engineer made me rethink every assumption about how the team’s AI exposure was sized.

What turns this into a gamble is committing headcount on the basis of AI’s potential before measuring its performance. HBR ran a piece in January with a headline I haven’t been able to shake: “Companies are laying off workers because of AI’s potential, not its performance.” I won’t do that while my own data is this uneven and the cost is still moving.

Bill Two — Product AI

This is the harder bill, because the customer side of the equation isn’t on my side anymore.

Every one of my top customers already has Office Co-Pilot at work, ChatGPT on their phones, and AI baked into most of the enterprise tools they already pay for. When I show them an AI feature in my product, the question I get is:

“Does your product do better than what we already have?”

What that means: my customers have anchored their AI expectation to whatever they get inside the enterprise software they already pay for. That software’s AI features are themselves subsidized by the same trillion-dollar capital flow propping up the model providers. My customers think they’re already paying for AI. They aren’t going to pay me extra. They want my version to be better at the same price.

When every SaaS vendor in my category is shipping the same AI features in the same quarter, and customers are pricing those features at zero against a subsidized enterprise baseline, that’s a defensive investment, not a growth investment. We’re not getting paid more for AI features. We’re paying not to be the one without them. Table stakes spending, dressed as innovation.

Defensive spending is real and necessary. But it should be sized as defensive: capped budget, no ROI thesis required, no growth story attached.

The cost basis is collapsing

When I look at the financial statements of the companies whose models power everything, the math gets worse the longer I stare at it.

OpenAI is forecast to lose $14B on $13B of revenue this year. It has committed $1.15 trillion in future spend to seven vendors, all of which are simultaneously its investors and suppliers: Broadcom, Oracle, Microsoft, Nvidia, AMD, AWS, and CoreWeave. Microsoft’s April 2026 10-Q discloses, quietly, that OpenAI is now a material accounts-receivable exposure on its balance sheet.

In April, Anthropic killed flat-rate billing on its Pro and Max plans. They told customers directly that open-bar pricing was no longer affordable. That was the first real-world signal I’ve seen that the subsidy underpinning today’s AI economics is ending.

If OpenAI follows (and I think they will within two quarters), every AI feature my industry has shipped this year reprices. Costs go up for my customers and for me, while the wrapper layer that priced itself to a subsidized input gets squeezed in the middle.

I have not seen a single competitive AI roadmap deck account for this.

The custodian’s dilemma

I’m a custodian of this business. My job is to deploy capital where the return is predictable, the cost is knowable, and the bet is sized to the company. That’s the entire fiduciary obligation.

The world I’m operating in is the opposite of that.

My annual board plan is already under pressure from a tough macro and geopolitical instability. Every dollar of spend has to justify itself harder this year than last. And my customers are operating in the same macro. Inflation, tariffs, and AI uncertainty are squeezing their profitability, lengthening their sales cycles to me, and pushing for price concessions. They want more AI inside my product, just not for more money.

So my cost is rising while my customers’ willingness to pay is falling. The macro that already made every dollar work harder is now a permanent condition rather than a passing one. On top of that, I have to absorb a new spending pressure (AI) whose cost basis is being engineered by capital I don’t control, with no commitment that it stays this way.

All of it compounds, and nothing relieves it.

I can’t stand on the sidelines. Every CEO in my seat knows this. The cost of being a year late to a category-defining feature is higher than the cost of overbuilding by a year.

But I also can’t pretend the budget math works when I know it doesn’t. Pretending is how a CEO destroys a balance sheet quietly over two years.

What I can control

I can’t fix the cost basis, manufacture customer willingness to pay for AI features they already get for free, or shorten anyone’s sales cycle. I can’t slow the competitive set.

What I can control is the way I size my own bets.

I price every AI contract as if costs triple, whether it’s an internal tool or third-party model spend in our product. Some still pencil. Some I’ve renegotiated. One I cut.

I won’t make headcount calls on the basis of projected productivity. I measure it in the metrics that affect the P&L — shipping velocity and renewal rate. If the lift is real, it shows up there.

And I make sure my board can see which of our AI investments are growth bets and which are defensive. Growth bets need an ROI thesis; defensive bets need a budget cap. Mixing them in the same line item is how budgets get blown.

I don’t see a happy ending

I want to end this with a clean answer. I don’t have one.

If I’m honest about what I see, this story doesn’t feel like it ends well. Enterprise adoption is being purchased with a subsidy that is ending. The trillion dollars committed to AI infrastructure is being recycled among the same handful of companies in a way that inflates everyone’s numbers without creating new end demand. Customer expectation has been anchored to a price no one is actually charging. Every CEO in my seat is being asked to spend faster than they can measure, in a macro that already punishes spend that doesn’t measure, while their customers are squeezed by the same forces.

Something has to give. When it does, there will be a lot of write-offs, not just at the AI companies that overpromised but at every enterprise that bought what they were selling without aligning pricing and ROI. The capability survives the unwinding; the buyers who priced themselves to a subsidized cost basis don’t.

In the meantime, my customers will keep asking whether I do AI better than what they already get for free at work. My CFO will keep flagging cloud overages. Sales cycles will keep stretching. None of that gets easier when the subsidy ends; it probably gets worse.

So the question I keep asking myself is the one I started with: are my AI decisions today defensible whether the subsidy continues or ends?

If they only work in one of those futures, I’m not running a business. I’m running a gamble.

If it disappeared tomorrow, would anyone mourn it?

Not would the customer find another vendor. Not would the investors lose money. Mourn. As in, would someone sit in their car after they got the news and wonder who they would be now if they had never walked through the doors.

Hold the test in your head. I am going to ask you to apply it.

Sam Altman is running a betting pool. Inside his group chat of tech CEOs, the wager is the year the first one-person billion-dollar company arrives. Dario Amodei put 70-80% odds on this year. The first contender is already here. Medvi, a GLP-1 telehealth startup launched out of a Los Angeles home in September 2024 with twenty thousand dollars and zero employees, posted $401 million in first-year sales. It is tracking to $1.8 billion this year.

Apply the test. If Medvi disappeared tomorrow, who would mourn it?

Customers would find another telehealth provider before lunch. There are no employees to lose their place. No managers who shaped anyone. No offices that became neighborhoods. No first jobs that started anyone’s career.

A unicorn that needs no humans helps no humans. That is what is wrong with the dream.

Here is what changed. AI agents collapsed the cost of execution. Work that used to require people you hired, trained, and paid — the design, the code, the support ticket, the bookkeeping — can now be rented by the API call. Suddenly the floor under “you need a team” disappeared.

But the floor was holding up more than execution. It was holding up purpose. I know because I lost a company once before I built one.

I was an leader at a cybersecurity market leader. Their mission was protecting people, businesses, and societies from cyber threats. Steve was the CEO. He was my mentor and my idol. One Wednesday evening, I was told he was leaving.

I couldn’t work the next day. I wandered the office. I got coffee with the team. I walked hallways I’d walked a thousand times and didn’t see them. That night I grabbed a drink with people I loved working with and came home late to tell my wife. The next morning the gossip had multiplied. What the board would do. Why. What came next.

It became clear to me. I believed in Steve. I believed in the vision we had built around him. Neither was real anymore. The company would keep going. The salary would keep arriving. The title would keep working on a resume. But the future I had bet my career on had quietly left the building.

I quit. No fallback. No other job lined up. It wasn’t a rational decision. It was the right decision.

What I built next was small. Lonely, at first. I left a comfortable paycheck and walked my network for months looking for a problem worth a decade. We landed on a personalized CRM for small businesses — software to help them build the kind of client relationships that survive bad quarters. I raised a little money. Found a first customer. None of that made the company real.

Camila did.

Camila was my first unpaid intern. She showed up because she believed in what we were trying to build, not because we paid her. We couldn’t. Then the first full-time employee said yes for less than market. Then ten more people did the same. They took less pay because they believed in the team and the mission. For two and a half years, what held us together wasn’t revenue or fundraising. It was the people in the room.

I remember the highs. I remember the lows more clearly. What I remember most is who was there.

If you are a founder reading this, you are being told two things at the same time. The first is that AI agents can replace most of the people you would have hired. That is largely true. The second is that you should be glad about it. That is the part that should worry you.

Let me give the solo-unicorn crowd their best argument. They will say: every revolution displaced workers, and every revolution created new sectors that absorbed them. Cotton gins, assembly lines, the internet. Each one was supposed to end work. Each one ended specific kinds of work and started new ones. The solo unicorn is the same story. Productivity rises. New industries form. Workers move on.

That argument has one problem. Past revolutions created new jobs. This one creates one job per company. The founder.

Then comes the second claim. Middle management is bloat. Layers slow things down. Flat is faster. Jack Dorsey told Block it had too many managers. Elon Musk gutted Twitter’s middle layer overnight. Mark Zuckerberg called it the year of efficiency and cut tens of thousands. The solo unicorn is just where that logic ends. Kill the managers. Then kill the team. Keep the cash flow.

What they are missing:

Middle management isn’t supervision. It is connecting tissue.

A manager’s actual job is to take an individual and grow them into a bigger version of themselves. To translate the company’s purpose into the daily decisions of people who can’t see the whole map. To notice when someone is about to break, and when someone is about to break out. Strip that tissue out and you don’t get a flatter, more agile company. You get a founder with a megaphone, agents executing tasks, and a base with no path forward.

Ask anyone who impacted them most in their professional life. They will tell you a name. They will tell you about a manager, a mentor, a CEO who pulled the future a little closer. Nobody answers that question with the name of a product. Nobody answers it with a tool.

Build a company nobody can answer with. That is the solo unicorn.

Here is what is being sold to you, founder.

You can build a billion-dollar company alone. You can keep the equity. You can skip the politics, the hiring mistakes, the culture work, the 2 AM Slack messages, the firings that haunt you.

You can also skip the only part of the work you will remember.

Ten years from now, the founders of the first solo unicorns will be billionaires. They will also be alone. Their company will exist on exactly one LinkedIn page. Their own. Their press will be the press of someone who built a P&L. Not the press of someone who built a place.

Apply the test one more time.

If your company disappeared tomorrow, who would mourn it?

If the answer is just you, you didn’t build a unicorn. You built a hobby with an exit.

I think about the office I walked that Thursday after Steve left. Same walls. Same desks. But the place wasn’t a company anymore.

Unicorns we ever celebrated were not the ones with the largest numbers. They were the ones whose absence the world would have felt. Companies that gave someone their first paycheck. Their first mentor. A promotion that taught them who they were.

So who is the unicorn for?

Choose the answer carefully. You only build one career.

He’s on my exec team. The kind of person you build a quarter around. The kind of person who makes you confident enough to say yes to harder problems because you know the bench can handle it.

He was telling me he’s moving on.

First thing through my head: not now. Not this quarter.

It takes a year to build that kind of bench. You don’t replace it in a quarter.

By Friday afternoon, I’d called one person I trusted to be clear-eyed about it. By Friday evening, I’d read the board in — calmly, with what I knew, what could be impacted, what I was already doing about it.

It took me two days to get over it. I was glad it landed on a Friday.

Most of that work happened inside my own head. Running disaster movies — the team noticing he’s gone, the next person leaving, customers feeling the gap. Recalculating the quarter. Replaying the last three months for the signal I missed. The conversation I should’ve had. The check-in I postponed.

By Wednesday next week — three business days later — I was genuinely excited for him.

You want your loyal people to dream big. You want the ambitious ones to take the bigger seat when it shows up. You can’t keep them forever and you shouldn’t want to. But when the moment lands — when the words show up in your ear or in your inbox — none of that helps. And the world feels like it’s crumbling.

A few weeks earlier, a different kind of email showed up — from one of our top customers. A driver of real growth, with real opportunity ahead.

Here’s the line that landed hardest:

“I want to be very clear — I am extremely frustrated with our experience as customers... another very serious concern.”

I read it twice. Then a third time. Then I started running the math. Five percent of revenue was at risk if I didn’t handle this cleanly. A relationship we’d spent years building, in jeopardy in three paragraphs. Discontent had been brewing under the surface for months — small things, missed signals, basics not done consistently. My point person was on vacation. And the hot potato landed in my lap.

This one took weeks, not days.

I had to acknowledge it without flinching. Look at my own gaps in oversight before I asked anyone else to look at theirs. Hold my team accountable for higher quality without turning it into blame. Do the basics consistently — calls returned on time, commitments tracked, small things stacked. Basics we’d taken for granted because the relationship was strong. Sky was falling. Until it wasn’t.

Today the relationship is mended. Far from perfect. But it’s back.

Both weeks had the same shape.

The gut runs disaster movies. The damage almost never matches the trailer. Not because the threat isn’t real — sometimes it is. But because the brain calculates worst-case in five seconds, and the actual outcome takes weeks of real work to reveal. By the time it does, you’ve usually already moved the dial.

Here’s what nobody tells you when you take the seat: this job puts you in five to ten of these situations every quarter. They are not the exception. They are the work.

Each one is a rep. You only get the rep if you stop running from the feeling.

So now, when the words land, I have a sequence I trust.

Tell one person who’ll be clear-eyed about it. Not to vent. To stay sharp.

Tell the board the same way every time: here’s what happened, here’s what could be impacted, here’s what I’m already doing about it, here’s what I need from you, here’s when we’ll talk next. Five lines. No drama.

Map the blast radius — this quarter and this fiscal year. Then find the smallest workable next step. Take it.

Make sure the person on the other side of it feels respected and supported. Relationships outlast the crisis if I let them.

And give myself a weekend to feel it. If I can.

I used to think the CEO job was not feeling the swings. Hold your face. Show up calm. Be the unflappable one in the room.

Now I think it’s the opposite. The job is feeling them faster. Gut punch. Recalibrate. Act. Shrink the time between the three.

From months, to days, to sometimes hours.

You don’t get to skip the “oh no.”

You only get to practice making it shorter.

He wasn’t complaining. He was reporting. His team is working harder than ever — heads down, putting in the hours. And they’re terrified. Worried AI will take over their jobs. Hearing that SaaS is dead. Getting no good news from anywhere. His exact words: “It feels like we’re working hard to get ourselves out of a job.”

I’ve been a CEO long enough to recognize that moment. I’ve also been honest enough, lately, to admit how I’ve handled it most of my career.

I sent an email.

For most of my career, I’ve treated fear like a weather event. Acknowledge it, give people the gear, wait for it to pass. It works — until the moment it doesn’t.

What I watched TH, my previous manager, do during COVID made me question whether that was ever really enough.

The world had collapsed in 2020. Every employee was carrying fear — professional and personal. Family. Health. Whether their kids would go back to school. Whether their parents would survive.

TH treated it like a campaign. Fireside chats. Virtual townhalls. All-hands meetings. Fun sessions designed just to let people breathe. Unlimited PTO — not just offered, actually used. No layoffs while the world was burning. Every occasion to show up: Zoom calls, flowers, cards, voice notes, showing up personally when someone was in a bad place.

Leadership was one of them. Not above the fear. In it.

Nobody called it a morale program. What it actually built was a safe environment where doing right by the people around you wasn’t awkward or soft. It was just how everyone operated.

The result: we grew at 20% that year. The market grew at 3-4%. Our nearest competitor topped out at 8%.

That delta isn’t a product story. It’s a leadership story.

Now I’m facing something different — and harder.

COVID was brutal. One enemy. Two years with no endpoint in sight. But there was still one enemy. You could name it, point at it, rally against it.

Today, think about what your people carry when they walk in on Monday. They’re running through five questions — whether they realize it or not:

Will I be ok?

Will I have a job?

Will my company be ok?

Will my neighborhood be ok?

Will our society and nation be ok?

During COVID, those questions were frightening — but most still had a fighting chance at a hopeful answer. The fear was acute. But so was the solidarity.

Today, too many of those questions land in the MAYBE or NO column. AI is displacing jobs in ways nobody fully understands. SaaS is getting commoditized. Macro pressure is grinding. Geopolitical instability is constant. The political ground keeps shifting.

No single enemy to name.

No visible endpoint.

When fear outnumbers hope across all five questions, it doesn’t stay home. It walks in with your people every morning. It shapes how much they invest, how much they risk, how much they trust.

What I keep coming back to is trust. Not programs. Not messaging. Built when people feel you’re carrying the same weight they are.

TH built that. What I’m trying to figure out is how to build it in a harder environment — without a single enemy to rally against and no finish line in sight.

I don’t know yet what this campaign looks like at full scale. One thing I know: it can’t be an email.

This time it looks like a multi-year effort — more challenging than what TH and leaders like him faced during COVID. And I’m at the beginning of it.

All I know right now is where it starts.

You listen first.

Listening is good. Action is better. Delivering a safe, productive employee environment is a never-finished task.

That sentence shouldn’t make sense. It does anyway. And that’s the problem.

I got my first computer in the late nineties. I remember the moment I sent my first email — the physical sensation of it, the disbelief that a message had traveled somewhere in seconds. Then a first webpage. Applying to colleges online. Buying a book from Amazon before I fully believed Amazon would survive. Creating a profile on a social network and finding people I hadn’t spoken to in years.

Nobody was alone in that feeling.

John Chambers, the CEO of Cisco, told anyone who would listen: “The Internet will change the way we work, live, learn, and play.” Not a warning. Not a hedge. A promise. Bill Gates called it “a tidal wave” — and framed it as opportunity, not catastrophe. Steve Jobs was building the device that would put the internet in your pocket, telling a generation they were here “to put a dent in the universe.” Marc Andreessen built Netscape — the browser that gave most people their first internet experience — and told the world the optimists are always right. Scott McNealy built his entire company on four words: “The Network is the Computer.”

These were not passive observers predicting the future. Best salespeople a technology has ever had. Their belief was contagious. It worked. Internet changed everything — partly because it was transformative, and partly because the people who built it spent twenty years convincing the world it was going to be great.

Positive. Consistent. Delivered by people the world had decided to trust.

Crypto’s story started differently. Satoshi Nakamoto published the Bitcoin whitepaper in 2008 under a pseudonym and disappeared. No interviews. No Senate testimony. No Twitter. Its original architect chose to have no public identity at all.

Into that vacuum walked Do Kwon, who promised a new financial system and called critics “poor.” And Sam Bankman-Fried, who donated to charity between defrauding customers. Satoshi’s anonymity was principled. What filled the gap was fraud.

Real ideas lived inside that technology. Some still do.

But three years after the peak? Crypto is an afterthought. Limited real-world adoption. Regulatory battles. Occasional headlines when a price moves. What was supposed to replace the financial system is mostly used for speculation.

It didn’t fail because the technology was bad. It failed because there was no Chambers, no Gates, no Jobs. And the messengers who showed up instead burned it down.

No technology wave in history has produced more prominent builders than AI. Dario Amodei at Anthropic leads with safety messaging — even the most careful builder in the room frames his work primarily around risk. Sam Altman signed a statement declaring AI extinction risk should be treated alongside pandemics and nuclear war, then published a blog called “The Intelligence Age” that didn’t mention extinction once. Geoffrey Hinton left Google “to speak freely about the existential threat.” Elon Musk says there’s a 20% chance AI leads to human annihilation — while raising billions to build it faster.

Poor communicators stumbling over hard ideas? No. People who have made a choice — conscious or not — to lead with fear. And unlike the journalists and academics who worried about the internet from the outside, these are the builders. They own the companies. They set the research agendas. They have every reason to want this technology to succeed.

And still. Extinction. Annihilation. Civilizational risk.

Results are exactly what you’d expect. 52% of American workers are worried about AI’s impact. 71% fear it will destroy jobs permanently. In 2024, sentiment flipped: more workers now believe AI does more harm than good. Mental health data is worse. That 56% of Americans who believed the internet would transform commerce for the better — that number has no AI equivalent. Because nobody with Chambers’ platform is making Chambers’ argument.

Yesterday, in my leadership meeting, one of my leaders said it plainly: “My team is shit scared. They’re working extremely hard. They’re worried AI will take over their jobs. They’re hearing SaaS is dead. There’s no good news. It feels like: work hard to get yourself out of a job.”

That sentence stayed with me. Because I feel it too — not just my team.

And here’s what matters: we didn’t create that sentiment inside our company. It came from outside. It came from the messengers. Every Altman Senate testimony, every Musk tweet about human annihilation, every “SaaS is dead” headline — it accumulated. And it landed in my leadership meeting, in my team, in my own head on a Tuesday morning.

More than bad press. Fear that migrates from public platforms into private rooms — and stays there.

AI is genuinely different from the internet. It competes with human cognition, not just human behavior. That’s real. Concern isn’t manufactured from nothing.

But the amplification is. Musk co-founded OpenAI, left, built a competitor, and filed suit against them. Altman calls for regulation that would benefit incumbents while racing to AGI. Competitors using the language of responsibility to fight market battles. That’s what’s happening.

The messenger is the message. Always has been.

Which means your company’s relationship with AI cannot depend on these people getting the narrative right. They won’t. Not because they’re incompetent — because they’re conflicted.

Your job — as CEO of a company trying to navigate this — is to bring sanity back into the room. Not cheerleading. Not dismissal. Perspective. The bad is not as bad as the headlines suggest. The good is not as good as the pitch decks claim. Somewhere in the middle is where reality lives — and where your company can actually operate.

Every technology we have ever adopted found its place eventually. The internet didn’t replace human connection — it changed it. Mobile didn’t replace conversation — it changed it. We’ll find a place for AI too. We always do.

But the wrong messengers have made the jobs and lives of billions of people genuinely miserable in the meantime. That’s not collateral damage. That’s a failure of leadership — not yours, theirs. And it’s a failure that falls to you to correct inside your own walls.

Your team doesn’t need another article about AI risk. They need a leader who has a point of view on what this technology means for them — specifically, in their role, this quarter. Chambers did it. Jobs did it. Nobody’s doing it for AI. Which means it falls to you.

The internet had Chambers, Gates, and Jobs. Crypto had Satoshi, SBF, and Do Kwon. AI has Dario, Altman, Pichai, and Musk — and Musk is suing Altman.

You decide which pattern that is.

And then go build the story your team is waiting for.

Employee NPS was 88. Nothing was broken. Nobody had asked for change. And there I was — unfamiliar face, untested voice, trying to say something meaningful to a team that didn’t ask for me. What I took from that room wasn’t a plan. It was a principle I’ve run on since: don’t touch what’s working. Find the one or two things that unlock the next level of value, change those, and let compounding do the rest. Transformation is incremental. The outcomes aren’t.

Humbling doesn’t cover it.

Now multiply that room by a thousand. That’s what Tim Cook walked into on August 24, 2011. Not a Zoom call with fifty employees — the entire Apple workforce, a press corps that had covered Jobs like a prophet, Wall Street, and millions of customers who didn’t just use Apple products. They believed in them. The founder Cook replaced wasn’t just successful. Mythological, actually. And the question everyone was asking, louder and less charitably than my team ever would: is this guy really the one?

What Most Successors Get Wrong

Successor CEOs face a choice nobody mentions on day one.

They can try to fill the founder’s ghost. Adopt the founder’s language, the founder’s habits, the founder’s way of walking into a room. Some do it consciously. Some don’t notice they’re doing it at all. Either way the result is the same: the ghost gets bigger, not smaller. Because you’re pointing at it every day.

Or they can do what Cook did.

Show up as exactly who they are. Play the game they’re actually good at. Trust that the mission is bigger than any one person’s style — including the founder’s.

Cook chose the second path. Without announcement. Without a manifesto. He just did it.

What Cook Built

Operations was Cook’s superpower. Not stage presence. Not product mysticism. The machine — supply chain, manufacturing, logistics at a scale nobody had attempted in consumer electronics. So he became the best supply chain operator in the history of manufacturing. He didn’t touch what Jobs had built. He built on top of it.

Then the services bet. App Store, Apple Music, iCloud, Apple Pay, Apple TV+. Services revenue hit $109 billion annually — a Fortune 40 company he built inside a company, from near-zero, during his tenure. Nobody was asking for this in 2012.

The Apple Silicon call came in 2020. Biggest architecture shift in 15 years. Pulled the entire Mac lineup off Intel and onto chips his team designed. Every MacBook sold today runs on it. Better performance, longer battery, and Apple doesn’t pay Intel a cent.

Wearables came next — Apple Watch, AirPods — categories that didn’t exist when Jobs died. They’re now worth more annually than most Fortune 500 companies.

He won an Oscar. CODA, 2022. Because why not.

None of it was imitation. Every move was Cook playing to his actual strengths — precision, financial discipline, thinking in decades when the press was thinking in quarters — while trusting the people around him to do what they were great at. He didn’t fill the ghost. He let it rest.

The Scoreboard

Apple’s market cap the day Cook took over: $350 billion.

Apple’s market cap today: $3.7 trillion.

Ten times. In fifteen years. At the largest consumer company on earth.

That’s not stewardship. That’s construction.

The North Star

We are all judged by the results we deliver — customers, investors, employees. All three scorecards. By every one of them, Cook delivered. A dream outcome by any measure.

But the part that doesn’t show up in the market cap is what I keep returning to: he did it with humility and poise, and with a complete absence of the ego that usually comes with that kind of achievement. And the world — somehow — spent fifteen years debating whether he was good enough.

That might be the biggest achievement of all.

For any CEO stepping into a post-founder era — first time in the chair, team that didn’t choose you, customers who have a relationship with someone who isn’t you — Cook is the model. Not because of the $3.3 trillion. The move isn’t to perform transformation. It’s to find the one or two things that unlock the next level — and change those. Build on what’s working. Let compounding do the rest. Cook knew this. Fifteen years of proof says he was right.

I know that room. Fifty people on a Zoom call, wondering why I’m there.

He knew it too. His was just bigger.

Thank you, Tim, for paving the path for many of us.