The Consultant's Workbench

The new-style consultant no longer sells packaged diagnosis. He moves in with the client, builds the tool that replaces the generic SaaS product, and carries the responsibility of keeping it running. A precise, bounded economic shift, one that opens a window for territories outside the big metropolitan centers.

For three decades, the consultant sold one thing: diagnosis, dressed up and delivered. Frameworks, matrices, recommendations. The value lay in the quality of the analysis and the elegance of the presentation. Execution was left to the client, to its IT teams, or to a third-party integrator. Between the one who thinks and the one who builds ran a clean line, and it structured the entire profession.

That line is shifting. Not everywhere, not for everything, but across a specific zone, one worth mapping out before going further.

Where the line moves — and where it doesn't

What's changing is economic far more than ideological. Building a custom tool cost enough, until recently, that the calculation almost always favored generic SaaS. Better to rent a tool used by ten thousand companies than fund a development team for six months to answer one company's own need.

With generative models, the cost of building and iterating a custom tool has dropped sharply. What took six months now takes two weeks. What required a team now requires one person who can both read a business need and hold their own in a code editor.

It's worth bounding right away what this doesn't mean. Within an SME's stack of SaaS products, the zone under attack covers only the middle layer: reporting and consolidation, business workflows, specialized assistants, interfaces between existing systems, local automations. That layer is precisely where the standard rubs worst against the real work, and where every company ends up paying for a generic subscription to do roughly what it wanted in the first place. That's where the consultant's workbench becomes credible. Outside that zone — ERP, statutory accounting, messaging, the critical tools at the core of the information system — SaaS remains structurally superior, and for good reason.

What SaaS carries that AI doesn't lower

SaaS doesn't win only because it costs less to build. It wins because it carries a burden easy to forget when you look only at the price of creation: corrective and evolutive maintenance, security updates, regulatory compliance, user support, continuity in case of failure, integrations kept up to date with the whole array of third-party tools. That burden is real, it's ongoing, and it has a cost.

Generative AI lowers the cost of building and iterating a tool spectacularly. It does not, by itself, lower the cost of the responsibility that follows deployment. That responsibility still falls on someone — in the model I'm describing here, on the consultant who delivered the tool. That's what separates a prototype demoed in a meeting from a tool that runs for three years without waking its creator at two in the morning. The new-style consultant doesn't sell a prototype. He sells responsibility carried over time, within a scope he chose because he can actually carry it.

Immersion replaces the engagement

The practice begins with immersion. The consultant settles in at the client's premises, observes the flows, watches what employees actually do at work. That observation is no longer limited to interviews and workshops; it can draw on tools that measure, at the level of individual workstations, which tasks eat up time and which ones repeat. This mapping of activity, cross-referenced with an understanding of the business, brings into view which gestures deserve a tool of their own — and which don't.

This kind of instrumentation calls for explicit discipline, because the risk of drift is real. The observation is consented to, announced to the teams before it's put in place, aggregates rather than names individuals, and its purpose is confined to identifying tasks worth automating. It is not used to evaluate people. It stops once the mapping work is done. Instrumentation that strayed outside that frame would be monitoring, plain and simple, and the bond of trust with the teams — which is precisely what makes the immersion possible in the first place — would be broken.

What the consultant delivers afterward is no longer a document describing what ought to be built: it's a tool that runs. Where it makes sense, that tool replaces a SaaS subscription that had been weighing on cash flow without serving the actual work. It's sized to the client's own flow, not to the generic flow that SaaS imposes on everyone alike. The practical, custom-built tool replaces the generic slide machine. At the scale of an SME's budget, the sum of middle-layer subscriptions regularly runs to several percentage points of revenue; the recoverable share is immediate, and it can be counted.

The distinction worth holding onto

The deepest benefit sits less in the invoice than in the time handed back to employees: time they can invest in whatever justifies their presence in the company, rather than in manually wrangling a poorly fitted piece of software. That effect, to my mind, is the one that truly changes the nature of the gain; it moves the argument out of the accounting ledger and into the register of the work itself.

The market confuses two things that need to be told apart. A company's AI transformation is not the same kind of thing as training its staff in AI. You learn to use an assistant by using it, and generalist awareness seminars — the ones that repeat what everyone has already read — add little. Targeted upskilling, by contrast, built around the tools a company actually deploys, grounded in the teams' real work, keeps its full value; it may even determine whether the delivered tool gets used at all. It's ungrounded coaching that's the problem here, not coaching as such. Transformation itself demands something else: a rare competence, able to grasp both a business value chain and the chain of technical tools that can support it. That's the competence the new-style consultant has to offer, and it's not one that can be pooled into an online session.

The profession doesn't disappear, the job description does

The consultant doesn't disappear, any more than most professions do. But the job description changes radically. What he used to sell — producing frameworks, delivering analysis — loses relative value. What he sells now — immersion, on-the-ground diagnosis, tools built to fit the real flow of work, and the responsibility of maintaining them — demands skills that, until yesterday, weren't part of the profile. Understanding an organization and holding a chain of technical tools together have become the same qualification. A share of consulting will stay in commentary and will keep losing ground to those who deliver the tool along with the analysis.

For territories that aren't consulting strongholds, this reshuffling is a window. The new consultant needs to be close to the client, to know its actual constraints, to iterate fast. Those conditions favor local presence over metropolitan centralization, and when the tool being built runs on sovereign inference infrastructure, the advantage of proximity becomes an advantage of sovereignty as well.

This shift is not a minor one. It redefines what a consultant sells: the tool that does, in place of advice about what should be done, and the responsibility of keeping it standing. The workbench replaces the firm, and what comes off the workbench holds up at delivery the same way it holds up the next morning.