Analysis
Almost no large enterprise has been "really, truly reinvented" by AI, Accenture's global AI chief technology officer Marc Warner said at Fortune's Brainstorm AI event on October 15, putting the share at "basically close to zero," according to Fortune. The comment came on a panel moderated by Fortune AI editor Jeremy Kahn, alongside executives from AXA, Mastercard and Mistral AI.
The Panelists
- Accenture -- global AI consultancy; Warner says the pace of AI change won't slow until roughly 2030, with Western energy constraints starting to bite in the early 2030s.
- AXA -- the French insurer started with about 400 scattered AI use cases, then narrowed its focus to a 2027-2029 plan targeting €500 million to €700 million in recurring pretax AI value per year by 2029, drawing on roughly 4 billion unstructured documents.
- Mastercard -- VP Greg Ulrich said the biggest gains come from redesigning entire workflows rather than speeding up one step, which just creates a backlog at the next step in an 80-step process.
- Mistral AI -- executive Marjorie Janiewicz, who spent a decade in cybersecurity before joining the French AI lab, joined the panel.
“- Mistral AI -- executive Marjorie Janiewicz, who spent a decade in cybersecurity before joining the French AI lab, joined the panel.”
Warner's framing is that AI transformations have become "politics projects," not technology projects -- a reversal from how he saw them until recently. AXA's Matthieu Caillat said the central question now is how to bring people along as the technology improves. Ulrich echoed that from Mastercard's side: the company leans on CEO sponsorship, including weekly updates with the CEO's direct reports, but "no one has solved the politics of large organizations."
That's a sharper admission than most AI vendors make in public. Most enterprise AI spending to date has gone toward peripheral tasks -- marketing copy, meeting notes, transcription -- rather than the core workflows that would show up as reinvented processes, Warner said. AXA's own trajectory illustrates the gap: narrowing 400 pilots down to a handful with a real financial target is itself evidence that most of the original 400 weren't moving the needle.
However, a panel of AI vendors and buyers describing their own transformation as slow is also a convenient story to tell: it lowers the bar for what counts as progress and buys more time before anyone asks why the ROI isn't showing up yet. For VCs underwriting enterprise AI startups, Warner's 2030 timeline is the more useful data point than the "close to zero" line -- it suggests buyers still have years of budget cycles ahead before the market matures, which is good for revenue growth but bad for exit timing.
