Accenture's CTO puts large enterprises truly rebuilt around AI at 'close to zero'

Both figures can be true. Warner said the number of large enterprises or government processes "that have been really, truly reinvented with AI at the absolute core is basically close to zero" ^[1]. The survey's category is fully modernized, AI-led operations, and it rose from 9% in 2023 to 16% in 2024 ^[19], a gain of 7 points ^[21]. We think that category is the wider one and that few firms in it have rebuilt a core process. Two other readings are open: Warner was rounding for effect, or the survey sample is not the large enterprises he meant. The reporting does not define either term.

Ulrich's backlog example has a number inside it. One step in an 80-step process is 1.25% of the sequence ^[16], and if the steps take equal time the other 98.75% still runs at its old pace until someone redraws it. Ulrich put the constraint in his own words: "The AI agent to automate a piece of this isn't... the long pole in the tent for a lot of this," he said. "It's the rethinking. It's the remapping." ^[8]

AXA's target is €500 million to €700 million a year of recurring pre-tax value by 2029 ^[10], or rather €1.25 million to €1.75 million for each of the 400 use cases it started with ^[17], if every one had survived. They did not: Caillat said the insurer started with about 400 and then narrowed its focus ^[9], so each remaining case has to carry more. Any interim figure AXA gives would be the first test of whether the value is building toward 2029.

Accenture's own book is growing in the same direction. Its revenue from generative and agentic AI tripled to $2.7 billion ^[13], from roughly $0.9 billion before ^[18], and its AI and data workforce nearly doubled over two years to about 77,000 ^[14]. The window for the growth gap in its research, 2019 to 2023, ends before any return on 2024 spending could show in it ^[20]. Crypto Briefing, which reported the research, noted that the correlation does not prove AI alone caused the outperformance, since well-run companies may simply be better at adopting new tools ^[3].

We think the bottleneck is organizational. The evidence for that is executives' accounts from Mastercard and AXA, and the counter-thesis sits in Warner's own remarks: he said he still sees the AI builds themselves as technology projects ^[6]. If core processes are risky because the models are not yet reliable enough, sponsorship and remapping will not fix it, and a company that redraws a core workflow would still book edge-sized returns. That result would prove the view wrong. Redesigns will also be built on tooling that keeps changing, since Warner said that, by Accenture's calculations, the pace shows no sign of slowing until something approaching 2030 ^[12].