emlyon business school just published its first Consulting & AI Barometer, an anonymous survey of more than one hundred professional consultants run between October 2025 and January 2026, led by Jean-Baptiste Vaujour, Director of the Master in Strategy and Consulting. Given that 55% of emlyon graduates go into audit, consulting, finance, banking and insurance, this isn't an academic curiosity for the school, it's a direct read on where their graduates' careers are heading.
The topline numbers confirm what most people in the industry already suspect: AI is fully embedded in daily consulting work now. But the more interesting findings are about who benefits from that, and what it's quietly costing the profession.
Nobody is debating whether to use AI in consulting anymore. Research, synthesis, and first-draft deliverables are just part of the workflow now, and that's where the 72% daily-use and 65% productivity-gain numbers come from.
What's unresolved is where that value actually goes. 62% of respondents say the main beneficiaries of AI gains are consultants themselves, not clients and not the firms employing them. That's a strange place for an industry to be a few years into adoption: everyone's faster, but nobody has cleanly worked out who gets paid for the speed.
There's a governance gap sitting underneath this too. 83% of firms have formal AI policies, but only 18% have consistent, team-level standards for actually reviewing AI-generated deliverables. That's a wide margin between "we allow this" and "we know if it's right." The report frames this well: a year ago the priority was getting people to adopt AI. Now it's making sure someone is checking its work.
The barometer's most counterintuitive finding is that AI doesn't help every part of teamwork equally. It's genuinely useful for reasoning through options and stress-testing hypotheses. It's much weaker at the thing that actually separates good consulting teams from mediocre ones: only 11% of respondents think AI helps meaningfully with prioritizing information or catching weak signals.
That distinction matters more than it sounds. Producing more analysis faster isn't the same as knowing which three insights in that analysis actually matter to the client. The barometer also flags a subtler risk here: when every team is drafting from the same handful of models, the output starts converging. Faster, but more uniform. The firms that figure out how to keep genuine disagreement and independent judgment alive inside AI-assisted workflows are going to out-think the ones that just let the model set the frame.
This is the part of the report I think matters most for anyone early in their career. 43% of respondents expect consulting firms to reshape into what the report calls an "obelisk" structure: narrower at the junior base, with more concentration in expert, manager, and solution-architect roles. AI is absorbing a chunk of the research and first-draft work that used to be the junior tier's job.
The risk isn't that junior roles vanish. It's that the tasks junior consultants used to cut their teeth on, the ones that actually built analytical judgment over a few years, quietly disappear along with them. The report calls this an "apprenticeship gap": consultants who look more productive on paper early on, but who've had less exposure to the grinding, occasionally tedious work that used to build real expertise. And notably, only 46% of managers currently think their firm's AI training programs are actually effective, so the fix for this isn't obviously in place yet either.
The report does push back on the more alarmist read, though: 70% of a junior consultant's workload is estimated to remain genuinely non-substitutable, tied up in things like supporting transformation work, building collective intelligence, and managing the interpersonal side of client relationships.
Jean-Baptiste Vaujour puts it well in the report: "A consulting firm's relevance will increasingly be measured not by the quantity of its deliverables, but by the quality of the judgement its teams apply to what AI produces." Paradoxically, AI adoption is making human judgment more central to the job, not less.
For anyone in or entering consulting, the practical read is this: the differentiating skill isn't prompting a model well, that's table stakes now. It's what the report calls epistemic calibration, knowing how far to trust what the AI hands you, when to push back on it, and when your own read on a client or situation should override the model's. That's harder to build if the tasks that used to teach it are the first ones automated away, which is exactly the tension this barometer is pointing at.
Certifications are helpful, but in 2026 they are more of an indication than a promise. The real value is in hands-on experience, and the data backs that up.
Burning tokens isn't a productivity metric, it's a cost center. Here's how the frugal AI shift translates into concrete architecture decisions on Salesforce projects.
Hooks let you enforce behavior in Claude Code instead of hoping the model does the right thing. The mental model, three Salesforce-specific hooks, and why this matters more in Salesforce than in most stacks.