"An ERP project is disruptive: I know how to relocate a plant; changing the company's processes is a different matter." That line from the Groupe Le Triangle's managing director, shared during this case study at the Tech For Industry Show, sums up the ambition of the Axone project. Alongside Philippe Geoffroy, Head of Application Services (Nagarro), Thibault Chevalier, Director of Transformation (Groupe Le Triangle) delivered a remarkably candid account of transforming a family-owned mid-sized company in hypergrowth.

Founded in the 1990s by a poultry farmer turned steel-frame builder — then a pioneer, starting in the 2000s, of solar panels on agricultural buildings, helping farms with limited investment capacity finance their buildings — Le Triangle has posted uninterrupted growth of 15 to 30% per year. Today: around 1,000 employees, €450 million in revenue, 2,000 to 2,500 end-to-end projects a year — customer need, design, factory production, logistics, installation. A hybrid group between construction and industry, with a goal set by its founder: €1 billion by 2030.

Two workstreams structure the transformation: industrializing a business built almost entirely on custom manufacturing — 2,500 unique projects a year — and unifying a group of around thirty operating companies with a fragmented IT system, nearly all of which touch every project. With a customer-driven prod: "farmers are very patient; big retailers much less so."

The hardest part, Thibault Chevalier admits: executives and employees "no longer realize how big they've become." At half a billion in revenue, authorities, customers, suppliers and employees expect a level of execution that fragmented processes — manageable "when we were all within earshot of each other" — no longer allow. Add to that a classic mid-sized-company phenomenon: the feeling that "this kind of project isn't for us, it's too big." Hence a double task: demystify, and get people to understand the absolute necessity of improving operations to stay on the founder's trajectory.

Installing panels is no longer the point: batteries, storage and resale management, self-consumption — the business itself is becoming more technical. It's now impossible to "separate technique and technology from operations."

Even before the ERP, the finding: a large share of project hiccups are self-generated — sales, engineering, production and logistics don't speak the same language and are unaware of each other's constraints. The work began with input and output data, responsibilities, and boundaries between project-carrying companies and support companies. Result: less friction, more harmony — and a simple goal: knowing "at any moment of the day where things stand," and, at the end of a project, whether money was made. "Today, it's an obstacle course."

The starting point shows how far there was to go: around a hundred Sage 100 databases, with no shared third-party or item reference database. The consequence: a lack of visibility into projects, a source of uncertainty and "analytical shortcuts" that put employees in difficulty they shouldn't be in. And a time-consuming prerequisite: most processes weren't written down — the five-person transformation team spends most of its time formalizing them, "one by one, going door to door," through individual interviews and workshops: "there are billions of unwritten rules."

The project shifts organizations from "Jean-Claude is the one who knows" to a data-driven management logic. Uncomfortable for some managers — transparency "unearths" what was hidden, power structures shift, some have left — but liberating for managers and employees, who "are aware of what's dysfunctional": you just need to show them, very concretely, what tomorrow's daily routine will look like. All the way to production, not yet computerized today, where new data entry is accepted if the gain outweighs the constraint. The method: proximity, regular internal surveys, and "being a bit of an octopus" — also picking up unofficial information.

The choice: SAP Public Cloud with the extended suite, favoring standard best practices — "a mid-sized company doesn't reinvent the wheel" — while pushing certain capabilities to their limits, such as the native configurator to model an entire custom building and all the resulting flows. On the deployment side, the "ambitious" option: a big bang across all 13 operating companies in Q2 of next year — production, logistics, sales and advanced configuration all at once, since highly integrated projects don't lend themselves well to a phased rollout. Outside the initial scope: around a dozen recently acquired SMEs and financial companies, to be integrated later. A single production site limits the risk.

Among the expected benefits, beyond visibility and breaking down silos: attractiveness (demonstrating to partners and candidates the ability to invest) and preparing for the future — quality structured data and efficient processes, "because that's what will later allow AI building blocks to be added intelligently, not on sand."

From the learning expedition to Hannover Messe and SAP Sapphire, Thibault Chevalier keeps one key takeaway: LLMs and agentic AI "give analytical power back to users" — no more extracts and cubes reserved for financial controllers, replaced by natural-language queries and automations that eliminate repetitive screens. With two requirements in return: employees whose profiles will need to evolve, and stronger governance — since delegating to agentic systems requires knowing how to define "what a correct answer is," where human judgment follows incomplete, unwritten rules. And a repeated warning against the "magic wand effect" of AI marketing: "someone good will get better; if you're not good to begin with, you won't work miracles." A telling anecdote about the new reality: the team uses an advanced prompt (Claude) to query millions of pages of SAP documentation, pre-arbitrate choices, and "challenge the consultants" — to Nagarro's great delight (or not).

This account offers a rare level of candor about the real difficulties of an ERP transformation at a family-owned mid-sized company: rebuilding never-formalized processes from scratch, getting organizations used to informal knowledge to accept data transparency, and deliberately choosing a big-bang rollout over a phased one. Groupe Le Triangle's central message: data and process foundations come before AI itself, which can only add value on top of an already solid base.

Watch the full session, along with the other Tech For Industry Show replays, on our dedicated page.