29% of employees admit they're sabotaging their company's AI strategy.
Not ignoring it. Not dragging their feet. Actively working against it.
Feeding garbage data into tools to make them look broken. Using unauthorized apps on purpose. Generating bad output so they can say "see, it doesn't work."
“”””
A year ago, I was deeply criticized for two posts I wrote on the structural challenges facing Accenture and the urgent need to address them. The argument was simple.
The market was underestimating the structural threat Al posed to the traditional consulting and systems integration delivery model. A year on, the numbers speak for themselves. Accenture is down roughly 38% year to date, close to 49% on a one-year total shareholder return basis, with nearly half the market cap gone.
The issue is that one of the most successful professional services operating models of the last 30 years is broken.
Accenture built an extraordinary business by industrializing enterprise transformation. Massive delivery programs, huge offshore workforces, over 300,000 developers in India alone, thousands of engineers deployed across long-duration
implementations. At its peak, it was arguably one of the greatest labor-arbitrage businesses ever created.
Generative Al attacks the economic foundation of that model directly. It sottware development becomes increasingly automated, what happens to firms built around monetizing vast pools of human delivery capacity?
The challenge is brutal because Al creates a paradox.
The technology can massively improve delivery efficiency, but improving delivery efficiency compresses revenue in businesses still priced around hours, utilization and headcount. The internal response tells you everything. Accenture has cut roughly 11,000 roles, with headcount down close to 14,000 over the past year and stealth layoffs running for 10 of the last 11 months in consulting. Over $2 billion spent on severance.
Promotion eligibility for senior managers now tied to measured Al tool usage. Julie Sweet has been explicit: people who cannot reskill to Al-native delivery are being exited on a "compressed timeline." Quietly, contractually, painfully, the firm is conceding the point.
Meanwhile, Al-native competitors are emerging without the same constraints. Smaller teams, lower overheads, faster deployment cycles, productized delivery.
Engineers augmented by Al from day one rather than retrofitted into legacy structures. Many of them will never need tens of thousands of developers because the work no longer requires it. The consulting industry spent two decades scaling through labor intensity.
The next decade will reward orchestration, IP, workflow ownership and Al-enabled leverage instead. This does not mean firms like Accenture disappear. They still have extraordinary client relationships, distribution and global reach. But the model has to evolve faster than any incumbent of this size has ever attempted.
Funny thing I’ve noticed, as the ACN share price discovers new floors, so do leadership’s expectations of us. High causal relationship.
Used to be “this is consulting, work hard or get out.”
Now? Silence. The pep talks have stopped. The hustle preaches have gone quiet.
Because they know. The consulting throne built on information asymmetry, prestige, and the promise of “Partner or bust” is wobbling.
AI is doing the analyst work. The pyramid is a liability.
And the talents who actually understand that? Either won’t join, or will immediately threaten the people above them.
So the new strategy is: say nothing. Expect less. Hope nobody notices the org chart is slowly becoming a museum exhibit.
The stock knows. The silence confirms it.
My average team member today uses M365 Copilot, ChatGPT and Claude, plus Accenture’s own AI interfaces like Amethyst.
While no one knows exactly which tools will be available in the future, they’re already throwing their workflows at AI.
Furthermore, no one can accurately calculate the cost of agentifying their workflows. More agents mean more parallelised agents. These are internal workflows, and each employee’s labour costs have increased while outputs are virtually the same. And clients have access to virtually the same models.
And no one can prove they’re generating more revenue thanks to the agentic workflows and tokens they’re burning.
When token costs become actual bills, I don’t see how Accenture can afford them. Tying token costs to token output is difficult, but all this “everyone must use AI for promotion” approach won’t be sustainable.