AI is a margin reset, not a technology shift.
Ninety-five percent of enterprise AI projects deliver no measurable P&L impact (MIT NANDA, 2025). I have watched the same script play out for 27 years, four waves in a row: web, mobile, data, cloud, and now AI. Across five Comex mandates in PE-backed mid-caps, mostly in France and continental Europe, no technology has changed the laws of value creation. Each new wave enforces them.
I have not been spared by the lesson. I have backed projects that should have been killed earlier. I have followed CEOs whose conviction was louder than their strategy. I have championed initiatives that never anchored to a customer outcome you could measure. The convictions that follow are sharpened by these calls.
Technology never changes the rules. It enforces them.
The web rewarded customer focus. Mobile rewarded customer focus. Data rewarded customer focus. AI will reward customer focus. The pattern is not a coincidence. Technology removes friction, and friction was the only thing protecting weak strategies from their own consequences. I have seen organizations spend through two consecutive technology waves without ever anchoring to a measurable customer outcome. The issue was not execution speed. It was strategic absence. The wave changed. The lesson did not.
AI is a utility, not a strategy.
Strategy answers where you want to go. AI answers how you can get there faster, cheaper, or with fewer people in the loop. Confusing the two produces the AI-first decks that fill consulting bills and never reach a customer. Bpifrance 2025 documents that 73 percent of AI projects in French mid-caps are launched by the CEO directly. That is also why most of them never reach P&L impact: nobody else owns them.
The illusion of cheap building is destroying what is actually rare.
AI makes coding feel free. AI makes content feel free. AI makes design feel free. The trap is that 'free' tempts boards to throw away the disciplines that took decades to build: product culture, brand craft, customer empathy, creative judgment. These are not commodities AI replaced. They are the only things AI cannot replace. Code became a utility. Product culture did not. Distinguishing the two is the new core competency of any value creation plan.
The org chart is the next thing to flatten.
Reporting in cascade, validation in cascade, coordination in cascade existed because information moved through humans, slowly, and needed echelons to filter and route it. Algorithms now do this work in seconds, at no marginal cost. The businesses that flatten in the next eighteen months will defend their multiples. The ones that wait will discover at exit that their EBITDA was inflated by activity, not value.
Repurpose. Do not eliminate.
The talent that handled the repetitive work is the talent that knows the customer, the product, and the operations. Redirect them toward what creates value. Layers go away. People stay. That is the difference between a business where institutional knowledge accumulates over the hold and one where it walks out the door every time the org chart is redrawn. The first compounds. The second pays consultants to recreate what it just lost.
The KPIs that compound. Everything else is theater.
Satisfaction drives repeat. Repeat drives organic growth. Organic growth drives the multiple. Every other metric is a proxy or a distraction. The portcos that win their hold periods will be the ones whose customer NPS, repeat rate, and share of wallet move quarter after quarter, with everything else aligned around those three. I have seen the difference firsthand: when a portco removes layers and redirects middle management toward the customer, repeat revenue moves within the hold. When it does not, the multiple at exit reflects the gap.
Engaged teams are an asset on the balance sheet that nobody dares to value.
Gallup 2025 measures the cost of disengagement: 9.6 trillion dollars of unrealized productivity globally. France Invest 2025 measures the upside of value-sharing: a fifteen percent performance gap. These are not HR numbers. They are equity story numbers. The fund that learns to invest in this asset will generate a structural advantage over the funds that keep treating it as a cost line.
Inaction is the most expensive decision of the hold.
The cost of waiting is invisible until it is too late, which is exactly why it is the most expensive decision a board can make. Wrong action shows up in three quarters and gets corrected. Inaction shows up at exit and cannot be. The Partners who confuse prudence with paralysis will discover that prudence has a price tag, paid in the multiple they will not defend. There is no neutral ground anymore. There is acting with discipline, and there is paying for not having acted.
The new investment thesis for any hold period currently in flight is not optional. It is operational.
- Customer obsession at scale.
- Flat organizations.
- Engaged teams.
- AI used where it actually creates value.
If you have a business where AI spend is rising faster than EBITDA, it is worth a conversation.