Everyone says artificial intelligence will 'revolutionise' businesses. What's less discussed is that many projects never move beyond the demonstration stage. The problem isn't the power of algorithms, but the gap between stated ambition and what the organisation can actually absorb.
Published on 2025-09-30 by Nathalie Lamborghini Dumas.
Everyone says artificial intelligence will "revolutionise" businesses. What's less discussed is that many projects never move beyond the demonstration stage and never truly touch day-to-day work. The problem isn't the power of algorithms, but the gap between stated ambition and what the organisation can actually absorb: ways of working, systems, skills, partners, decision-making processes.
This article is about that: setting an AI ambition at the right level for your company, so that projects move from presentations into real life.
Classic scenario: you're presented with a spectacular use case, "auto-optimised" prospecting, augmented customer service, "intelligent" forecasting, real-time steering. On paper, everything flows. In your reality, data is scattered across business tools, CRM, spreadsheets; practices vary from one team to another; many key decisions are still made by phone or in informal meetings.
An experiment is launched on a tightly scoped perimeter, with strong support from the provider. Technically, it works. But as soon as it comes to scaling, everything creaks: insufficient or poorly structured data, no rituals for exploiting results, teams that have neither the time nor the reference points to integrate the tool. The problem isn't AI itself, it's having imagined AI as if the company were already something other than what it actually is.
Most conversations start with: "What could we do with artificial intelligence in our company?". For an SME or mid-cap, a more useful question is: "What type of AI project can our organisation truly absorb in the next twelve to eighteen months?".
This shift forces you to look at very concrete elements:
AI ambition then becomes not a list of abstract possibilities, but a choice within a realistic space, defined by your current context and your likely trajectory. Experience elsewhere shows the same thing: the companies that advance aren't those stacking the most "intelligent" tools, but those aligning their AI projects with the robustness of their data, the maturity of their practices, and their teams' ability to adopt new uses.
To simplify, we can distinguish three levels of AI projects in an SME/mid-cap, across all sectors.
First level: AI within existing tools. This means activating or adding AI capabilities in systems you already use: writing assistance in customer relations, request triage support, recommendations in the CRM, automatic meeting summaries in project management tools, risk analysis in financial tools. It's unspectacular from the outside, but highly effective when use cases are well chosen.
Second level: AI in decisions and rituals. The goal is to prepare structuring decision moments: sales committee, team meeting, priority arbitration, executive committee. AI produces scenarios, prioritisations, pre-structured analyses that teams discuss and validate: priority prospect lists, workload distribution between teams, cash flow scenarios, possible action plans in response to demand changes. The main difficulty is no longer technical but managerial: evolving how these moments are prepared and run.
Third level: AI in the business model. This means leveraging data and new capabilities to offer new propositions: subscriptions including monitoring services, data-augmented consulting offers, client portals or platforms, usage-based models rather than one-off services. This level directly touches your value proposition, your pricing, your client relationship. It requires the first two levels to be solid enough for the promise to hold.
The challenge for a leader isn't to choose "the highest level" on principle, but to position each project at a level consistent with the company's current state.
In an executive committee, a few questions suffice:
The answers aren't a judgement. They simply indicate which types of projects have a reasonable chance of succeeding in your context, today.
Every company has its "internal gravity": what moves fast, what moves slowly, what blocks, what accelerates. An AI project that acts as if this gravity doesn't exist has little chance of moving beyond the prototype stage.
Conversely, a project designed from your actual processes, your actual systems, your actual decisions has a much greater chance of being adopted and producing visible effects. Whether your goal is to optimise operations, secure your business, better serve your clients, or later create new services and business models, the starting point remains the same: what your organisation is ready to absorb today.