Why I Left Consulting to Build Enterprise AI From the Inside
After 10+ years at Accenture, Capgemini, and KPMG, I went in-house. The transition from advisor to builder changed everything.
The Consulting Trap
I spent over a decade in management consulting — Accenture Song, Capgemini Invent, KPMG Advisory. I wrote beautiful decks about AI transformation. I presented them to C-suite audiences. I delivered "strategic roadmaps" with three horizons and a maturity model.
And then I left before implementation.
The hard truth: most AI strategies I wrote as a consultant never got fully executed. Not because they were wrong — the analysis was sound, the recommendations were reasonable. They failed because the deck doesn't capture the organizational immune system that fights change.
The IT team that's been burned by the last "transformational" initiative. The operations manager who's terrified of losing control. The legal team that can't approve something they don't understand. The vendor who sells the dream but can't deliver the reality.
You don't see any of this from the outside.
What Changes When You Go In-House
You Own the Outcome
In consulting, you own the recommendation. "We suggest implementing a GenAI-powered customer service solution across three channels." Beautiful sentence. Zero accountability.
In-house, there's no "we suggest." There's "I'll make this work." The digital assistant I'm responsible for handles tens of thousands of conversations every day. If it fails, real customers have a bad experience. Real revenue is impacted. My name is on it.
This changes how you think about risk. In consulting, you optimize for being right. In-house, you optimize for not being catastrophically wrong. Different game.
Stakeholder Management IS the Project
In consulting, stakeholder management is a workstream. In-house, it's the entire job.
The technology is maybe 20% of the work. Getting alignment between IT, operations, legal, compliance, customer experience, and the business unit is the other 80%. And unlike consulting, where you parachute in with authority borrowed from the CEO's mandate, in-house you have to build your own credibility. Every day. With the same people.
I spend more time in meetings explaining what GenAI can and can't do than I spend actually building GenAI systems. That's not a complaint — it's the reality of enterprise transformation.
You Learn to Love Constraints
In consulting, "best practice" is aspirational. You recommend the ideal architecture, the cutting-edge model, the optimal team structure. Whether the client can actually implement it is, frankly, someone else's problem.
In-house, you work with what you have. The legacy CRM that can't be replaced. The vendor contract that locks you into a specific platform for two more years. The budget that got cut by 40% in Q2. The team that has three developers, not the twelve your roadmap assumed.
Constraints aren't obstacles — they're the design space. The best solutions I've built in-house are ones I'd never have recommended in consulting, because they're tailored to specific realities rather than generic best practices.
The Moment It Clicked
I was sitting in a call center, headphones on, listening to actual customer conversations. Our digital assistant was handling a pension inquiry. The customer had a strong regional accent. The ASR transcribed "pensione" as "pensare." The bot responded to a philosophical question about thinking instead of a practical question about pensions.
No deck in the world prepares you for that moment. You need field observation — the AI equivalent of gemba walks. You need to hear the system fail on real humans with real problems.
That's when I understood why so many consulting recommendations stay on slides. The people writing them have never sat in the call center.
What I Kept From Consulting
Not everything was wasted. Ten years of consulting gave me:
- Structured thinking: MECE problem decomposition, hypothesis-driven analysis. These work everywhere.
- Executive communication: The ability to explain complex technical decisions to a board in three slides. Invaluable.
- Stakeholder mapping: Knowing who to align with before launching anything. This saved me months.
- Pattern recognition: Having seen 20+ organizations attempt transformation helps you spot common failure modes early.
What I Had to Unlearn
- The obsession with frameworks over action. A framework is a thinking tool, not a deliverable. Ship something.
- The "best practice" mindset. Best practice for whom? In what context? With what constraints? Stop copying and start designing.
- The belief that a good strategy survives contact with reality. It doesn't. Plan to iterate, not to execute.
The Bottom Line
Consulting taught me how to think about transformation. Building taught me how to survive it.
If you're a consultant considering the jump in-house: do it. You'll learn more in six months of building than in three years of advising. But check your ego at the door — the people who've been running the operation for years know things you don't.
And if you're an enterprise leader hiring consultants for AI strategy: make sure at least one person on the team has actually deployed a production AI system. Strategy without execution experience is just expensive fiction.
Thinking about your own consulting-to-builder transition? Or need someone who's done both? Let's talk.