AI Is an Electrician's Job and Your IT Landscape Is the Grid.
During a talk at École Polytechnique, Arthur Mensch, CEO of Mistral AI, dropped a line worth pausing on:
"AI is a bit like an electrician's work. Our job is to turn megawatts into intelligence."
We found this remarkably clear-eyed, because it does something rare: it demystifies the subject entirely. Behind the "magical" chat interfaces, the spectacular demos, and the weekly announcements of yet another model, there's a physical, industrial reality. Flows. Consumption. Efficiency. Cables.
And when you take this metaphor seriously, it sheds light on a lot of what we see every day working with our customers.
The Current, the Grid, and the Power Strip
Let's run with the metaphor for a moment.
If AI is the electrical current (the megawatts of intelligence), then your IT landscape is the distribution grid. The cables, the circuit breakers, the panels, the outlets.
The current can be as powerful as you want. If the grid isn't well mapped, a bit outdated in places, or held together with extension cords, you tend to end up with:
- Short circuits. AI generates answers from inconsistent data because it's no longer clear which sources of truth to rely on.
- Transmission losses. The energy invested in AI projects dissipates in integration work, data cleaning, and reconciling reference systems.
- Dead zones. Entire parts of the organization stay in the dark because the grid simply doesn't reach them.
We all know what happens when you plug industrial equipment into an old power strip lying around. It's rarely spectacular in a good way.
A Question of Efficiency, Not Power
What struck us about Mensch's quote is the idea of efficiency. The goal isn't to have the biggest model possible. It's to have the most efficient one for the problem at hand.
That's an interesting shift in perspective for enterprises. Over the past two years, the trend has mostly been "plug AI into everything and see what happens." Companies connected LLMs to processes without always checking whether the process itself was well understood. They injected AI into workflows that hadn't necessarily been fully mapped.
That's not a criticism, it's a pretty natural reaction when a technology moves this fast. But the pattern tends to repeat: lots of proof-of-concepts, fewer making it to production. Lots of enthusiasm, ROI that's not always easy to measure.
And in quite a few cases, the reason is fairly straightforward: the current came before the grid.
When you think about it, the most useful questions before an AI project might not be "which model should we use." They're more like:
- What is our data? Where does it live? Who owns it?
- Are our information flows documented, or do they mostly rely on the knowledge of a few key people?
- Is our application architecture readable, or has it become a bit hard to navigate?
These are electrician's questions. Architect's questions. And they're probably worth asking before the data scientist's questions.
Before You Electrify, Map the Grid
No serious electrician starts by installing outlets. They start by reading the grid plan. And if the plan doesn't exist, they draw it before touching a single cable.
For AI in the enterprise, we think the logic is pretty much the same.
Before connecting an AI copilot to a CRM, it's worth asking: do we know precisely what data feeds it, where it comes from, and how it flows?
Before deploying an AI agent on support tickets, the useful question might be: do we have full visibility of the application chain between the ticket system, the customer database, the knowledge base, and the billing tool?
Before automating a workflow with generative AI, it can be worth checking: is this workflow documented somewhere, or does it mostly live in the heads of the team that runs it?
These aren't blocking questions. But answering them upfront tends to save quite a few surprises down the line.
The Enterprise Architect in All of This
The electrician metaphor also offers an interesting lens on the role of the enterprise architect.
They're sometimes seen as a compliance guardian, someone who documents the past. But in this framing, they become something more strategic: the person who makes sure the grid can handle tomorrow's load.
Not producing an exhaustive diagram for an annual audit. More like maintaining a living, shared, up-to-date vision of the IT landscape, so that every new connection is made with full awareness.
In practice, that comes down to a few fairly concrete things:
A living map, not a frozen document. If the grid plan is eighteen months old, things have probably moved since. Architecture benefits from reflecting reality continuously, connected to operational data.
A shared view, not an expert-only diagram. The electrician doesn't keep the blueprint to themselves. They show it to the site manager, the project owner, sometimes the client. Enterprise mapping benefits from being readable by the CIO and the executive board alike.
Impact analysis capability. When you want to plug in a new piece of equipment, the electrician checks whether the panel can handle the load. Same idea here: being able to quickly assess the impact of a new project on the existing ecosystem.
This Is Pretty Much Why We Built Boldo
At Boldo, we're building an enterprise architecture tool that starts from a simple observation: the grid plan should always be readable, always up to date, and accessible to all stakeholders. Not just the architect who drew it.
Intuitive by design. Business and IT read the same map, no training required. Because a grid plan isn't much use if only the electrician can read it.
Real-time collaboration. Mapping is built and updated continuously, not once a year during a master plan exercise.
Connected to your data. Boldo plugs into your existing sources (CMDBs, reference systems, service catalogs) so the map reflects the real state of the grid.
Built for impact analysis. Identify dependencies, assess risks, and check that the grid can handle the load before plugging anything in.
Arthur Mensch is right: AI is an electrician's job. And the first reflex of a good electrician is to check the grid plan before turning on the current.
👉 Try Boldo free for 14 days and make sure your grid is ready for what's coming.
