Most important take away

Traditional SaaS is dying and only “AI-first” companies (those generating the majority of new revenue from AI products and automating internally) will win the next cycle. Founders should disrupt themselves before competitors do, upskill every role around building with AI, and treat the AI talent/chips/energy stack as a geopolitical priority — particularly in Europe, where the answer is to stop whining about regulation and capital and just execute.

Summary

Actionable insights from Daniel Khachab, Co-Founder/CEO of Choco, on pivoting a SaaS unicorn to AI-first, surviving COVID, building in Europe, and fundraising:

On pivoting to AI-first (the central thesis):

• Choco killed its SaaS revenue model so 100% of revenue now comes from AI. Khachab’s logic: either someone replicates your 5–20 years of code in days using LLMs, or you disrupt yourself first. “Always play offense.”
• Define “AI-first” as: a majority of new revenue comes from an AI product, plus AI used internally for productivity. Soon this won’t be a label — it will be the default, like having a social media account.
• He believes great companies will get smaller in headcount, not bigger. The 10-person unicorn is realistic today with the right team.
• Treat AI products as selling an “employee,” not a feature set: it can be hired in fractions, doesn’t take breaks, and pricing/marketing/QA all have to be rethought.

On the death of SaaS UIs:

• UIs exist because users had to learn software. With prompt-based interfaces, there is nothing to learn → adoption curves collapse, especially in traditional/laggard industries (the opposite of conventional wisdom).
• Two interfaces will survive: (1) the prompt where you ask AI to do something, (2) the training interface where users teach the AI on proprietary data. Don’t oversell day-one accuracy — sell rate of learning.
• Predicts Salesforce-style apps become the application layer themselves, with the legacy SaaS layer collapsing into a database underneath.

Tech patterns mentioned:

• Use of OpenAI (primary), Anthropic, and Mistral. Costs dropped ~80% in six months despite higher volume — pure commoditization at the foundation layer.
• Small language models hosted on-prem solve enterprise data-security objections — not everything needs to be a hosted LLM.
• Anthropic’s “computer use” is the next-gen RPA and shows foundation-model companies pushing into the application layer.
• Biggest enterprise blocker today is implementation: data readiness, data cleanliness, and the post-purchase “now what?” gap.

Career advice:

• “If you left Choco tomorrow and didn’t have AI-building skills, your skills would be obsolete.” It is the employer’s responsibility to upskill every role (design, marketing, sales, engineering, HR) on building with AI.
• Product designers now design the character of an agent, not just UI. Engineers must learn how to QA non-deterministic outputs.
• Choco automated away ~100 customer-support/account-management roles. Khachab argues founders have a duty to existing employees not to fall behind — but the conversations are uniquely hard because the people being let go are high performers.
• On work-life balance: he rejects it. Be 100% present and 100% intense at work; same at home. Expect Sunday calls.
• On founder selection: “every cell in your body” must want to do it. If one cell hesitates, don’t. Founders should be wet before they jump in the pool. Commit 15–20 years; treat short-termism as the European disease.

Geopolitics — the AI cold war:

• Governments (US, UK, UAE, Saudi) are personally lobbying AI founders at dinners to relocate. Some offers include the government paying senior AI salaries for 3–5 years plus golden visas.
• True AI sovereignty requires three things: foundation models, chips, energy. Europe has none of the three. Until it does, it’s not sovereign.
• His prescription for Europe: fund the right chipmaker (TSMC/Nvidia, not Intel); solve energy (he’s torn on nuclear); the foundation-model gap is a talent problem and only needs one founder willing to commit a lifetime (Mistral is proof it’s possible).
• “Regulatory” and “lack of capital” are excuses. Capital is now global — there are 5 daily flights from Berlin to London and 2 to JFK. Zuckerberg and Musk wouldn’t have let a notary stop them.

On managing through COVID:

• Lost 98% of GMV in a week. Read Shackleton’s biography for crisis management. Two lessons: never lose your humor, never change your values.
• Mistake: gave the whole team a week off in week one of COVID. Purpose vanished, made things worse. People need purpose, not rest, in a crisis.
• “Always play offense”: rotated sales teams to whichever US state/country wasn’t locked down (Spanish team → Miami became their best launch). Didn’t lay off, burned cash, came out ahead of hibernating competitors.

Fundraising lessons (raised $330M):

• The fundraising “process” is just: bring great people together → set mission → execute → put graphs on ugly slides → hand them out. If the graph goes up and to the right, you raise. If not, no deck saves you.
• VCs are loss-averse to missing deals, not making bad ones. Meet on eye level — don’t beg.
• Take the money if you can. More cash = higher probability of mission outcome, unless you let unicorn status soften the culture. He regrets throwing a unicorn party — culture shifted toward “we made it.” Ordered the flags removed two weeks later.
• Articulate how you become a $10B company. European founders are conditioned not to — that’s a mistake.
• Biggest spend regret: a “special projects” skunkworks team. If your core can’t innovate, that’s the real problem; don’t outsource innovation.
• Biggest underspend regret: not tripling burn earlier once product-market fit was clear, to grab brand share before competitors arrived.

Mental models worth stealing:

• “Lions don’t lose sleep over sheep” — intrinsic mission motivation beats competition-driven motivation.
• Jungle metaphor: the cooler stories from running a company are the river full of piranhas at midnight, not the sunny mountaintop. Be grateful for the shit times — they keep you calibrated.
• “Third-generation companies”: every unit of economic success is mechanically tied to a unit of planetary/societal success (vs. Gen 1 “ignore externalities” or Gen 2 “go to zero impact”).
• On focus: most public mindshare goes to EVs, but eliminating food waste has 5x the carbon impact of removing every car. Be methodical about leverage points.
• Bias for action vs. impulsiveness is judged retrospectively by outcome — so just act.

Chapter Summaries

1. Why Choco killed its SaaS business (open) — A sleepless night staring at the ceiling: LLMs will replicate years of SaaS code in days. Either disrupt yourself or be disrupted. Going AI-first is also how you upskill every role in the company.

2. The death of SaaS UIs — User interfaces exist because humans had to learn software. Prompts remove that friction → faster adoption in traditional industries, not slower. Salesforce-style apps collapse into the application layer; underlying CRMs become a database.

3. Every job changes — Design becomes character/agent design. Engineers must figure out QA for non-deterministic outputs. Marketing/sales now sell a fractional “employee.” HR risks obsolescence.

4. Where value accrues in the AI stack — Foundation-layer prices dropped ~80% in 6 months. Anthropic’s computer-use blurs the line between foundation and application. Choco uses OpenAI primarily, with Anthropic and Mistral. Utility > price.

5. AI and the labor market — Klarna laying off 700 in support is a preview. AI will reallocate scarce labor to under-staffed sectors (kindergartens, healthcare, hospitality). The “shit jobs” go first.

6. Automating Choco’s own org — ~100 customer-service/account-management roles removed. Hard conversations because they were high performers. Founders’ first duty is to the mission and the surviving team.

7. The AI cold war: talent — US, UK, UAE, Saudi governments personally recruiting founders at dinners. Most extreme offers: governments paying top AI salaries for 3–5 years. Berlin chosen for affordability and network, not AI excellence.

8. The AI cold war: sovereignty — A sovereign AI nation needs foundation models, chips, and energy. Europe has none. Europeans get OpenAI Advanced Voice and Apple Intelligence last.

9. Why Europe is f***** (and how to fix it)** — Stop whining. Capital is global. “Regulatory” is mostly an excuse. Fund the right chipmaker (not Intel). Solve energy (possibly nuclear). Backing Mistral-style foundation bets is a talent problem, not a money problem. Need founders who commit 15–20 years.

10. Surviving COVID — Lost 98% of GMV in a week. Shackleton as a manual. Never lose humor, never change values. Mistake: giving the team a week off killed purpose. Won by rotating teams to non-locked-down geographies.

11. Fundraising lessons — Raised $330M. Take the cash. The downside of unicorn status is cultural softening — he regrets throwing the unicorn party. Don’t beg from VCs; the dollar is a commodity. Articulate the $10B path.

12. Where the money went well and badly — Worst spend: special-projects skunkworks team (if your core can’t innovate, fix the core). Best missed opportunity: didn’t triple burn fast enough once PMF was clear.

13. Competition and motivation — Be grateful for competitors. Intrinsic mission > extrinsic enemy. “Lions don’t lose sleep over sheep.”

14. Quickfire — Only AI-first companies win. Headcount shrinks, not grows. Bias for action ≈ impulsiveness, judged by outcome. Pixar for inspiration, Deutsche Bahn as governance horror story. The next generational European company will be a privatized Deutsche Bahn replacement.

15. Founder responsibility and “third-generation” companies — Founder motivation must be the mission, not money. Caltrain still sucks despite trillion-dollar companies along it; tech hasn’t lifted average quality of life enough. Third-generation companies tie every revenue unit mechanically to societal/environmental gain. Focus on the highest-leverage problems (food waste >> EVs on carbon).