The debate about artificial intelligence in consulting swings between two caricatures. The first announces the disappearance of the consultant, replaced by models that produce reports in seconds. The second dismisses AI as a gadget generating hollow text with no understanding of organizations.
Our daily practice, equipped with a proprietary suite in production since 2024, has led us to a more nuanced and more demanding conclusion: AI radically changes the economics of consulting, but it does not change its nature. A firm’s value never resided in its capacity to produce pages. It resides in judgment: knowing which question to ask, recognizing what matters in a mass of information, standing behind a recommendation in front of a board.
What AI actually changes
Three concrete shifts, measured in our own mandates:
Absorption speed. A thousand-page corpus of institutional documentation that used to demand weeks of reading can now be mapped in hours. The consultant does not read less; they read better, guided toward the critical zones.
Depth of exploration. Where time constraints used to allow two or three scenarios, it becomes possible to develop ten, with their assumptions, costings, and risks. Blind spots shrink.
Consistency of quality. The tasks where fatigue produces errors (consistency checks, compliance controls, bilingual harmonization) are precisely where the machine excels. The quality floor rises.
What AI does not change
No model will carry the responsibility for a recommendation. No model will sense that an exhausted chief executive needs a three-step roadmap rather than a twelve-step one. No model will decide that an uncomfortable truth must be told to a client, and how to tell it.
That is why we speak of augmented advisory rather than automated advisory. The distinction is not rhetorical; it is architectural. In our systems, AI proposes, structures, verifies, and accelerates. The expert frames, arbitrates, validates, and signs. Every deliverable that leaves the firm has been judged by a human who stakes their reputation on it.
Four guardrails for responsible integration
For organizations that want AI inside their analysis and decision functions, four guardrails strike us as non-negotiable:
- Confidentiality by design. Sensitive data never leaves the agreed perimeter and is never used to train models. This is verified contractually, not on trust.
- Traceability of contributions. Knowing what came from the machine and what came from the expert, so both can be audited.
- Human judgment at decision points. AI can build the case file; it does not rule on it. Every decision point in the workflow must name its human decision-maker.
- Team capability building. A tool that teams endure produces surface compliance. A tool they master produces durable capacity.
The strategic stakes for organizations
Soon the question will no longer be whether your partners, competitors, and evaluators use AI. They already do. The question is whether your organization will have structured that use according to its values and obligations, or whether it will improvise under pressure.
This is a governance program as much as a technology program. And it is exactly at that intersection that our AI-augmented advisory practice operates: designing systems where the machine amplifies your teams’ judgment without ever substituting for it.
React to this analysis
Comments are moderated before publication. No account required.
React by email