The concept of 'full-stack AI company' is electrifying investor circles. But behind the jargon lies a more ancient and fundamental question: who controls the value chain? For SMEs and mid-caps, the answer doesn't necessarily involve replacing existing players. It involves the platform.
Published on 2025-12-18 by Nathalie Lamborghini Dumas.
What "full-stack AI" really means
The term has been circulating for months in pitch decks and investor notes. A "full-stack AI company" doesn't sell software to an accounting firm: it becomes the accounting firm. It doesn't provide a tool to an agency: it replaces the agency. AI enables enough operations to be automated that a startup can operate the service itself, end to end, with a fraction of traditional headcount.
In France, several players have embodied this logic for years. Dougs, created in 2015, is an accounting firm registered with the Order : not a software publisher. It operates the accounting of over 37,000 companies itself: automatic bank synchronization, AI categorization, balance sheet production, direct client relationship. The customer doesn't pay a software subscription: they entrust their accounting to Dougs, who produces it.
Alan, since 2016, does the same in health insurance. It's not a broker or a comparison site: it's an insurer with its own license, bearing the risk. Direct distribution without intermediaries, internalized contract management, 90% of reimbursements automated in less than 24 hours, telemedicine and prevention integrated via the AI assistant "Mo." Where a traditional mutual relies on a chain of brokers and managers, Alan has internalized everything.
These companies didn't wait for the concept to act. But their success illuminates what "full-stack" really means: it's not primarily a technology question. It's a question of value chain control.
The false dilemma: tool or replacement
The dominant narrative poses a binary choice: either you sell tools to existing players, or you replace them. This reasoning is seductive in fragmented or lightly regulated sectors. It is much less so in European realities, where SMEs and mid-caps are simultaneously operators, integrators, quality guarantors, legally and socially responsible.
For a leader, the real question is therefore not "Should I become a full-stack AI company?" It is rather: "What part of my business should I continue to delegate, and which should I take back in hand?"
Between the tool and the replacement exists a third way, largely underestimated: the operated platform. Not the marketplace in the trivial sense, but a strategic device that allows internalizing the critical building blocks (data, automation, orchestration) operating them first for one's own business, then structuring them to augment, equip or coordinate an ecosystem.
The platform: a realistic translation of full-stack
A platform is not just a marketplace. It stacks several levels of value creation. The infrastructure layer provides the physical or digital assets that keep the whole running. The intelligence layer structures data and generates insights for participants. The connection layer matches supply and demand, facilitates transactions. And the ecosystem layer creates collective impact that exceeds the sum of its parts. The more a company works these different layers, the harder its position becomes to replicate.
Let's take a concrete example. A perfume creation laboratory that uses standard software to manage its formulas remains a customer of a solution. Now imagine this same laboratory deciding to operate the creation service end to end: it receives a brand's brief, AI analyzes trends and proposes olfactory directions, perfumers refine the formulas, the platform identifies the best raw material suppliers, manages iterations with the client, and delivers the final juice ready to produce. The laboratory no longer sells just its expertise: it operates the entire creation journey. And by connecting independent perfumers, suppliers and brands on its platform, it becomes the orchestrator of its ecosystem.
Metro illustrates this approach with its Dish platform: the group continues its historical wholesale food activity while operating a digital platform that connects restaurateurs to a service ecosystem : online booking, inventory management, payment solutions, delivery. Every euro invested in the platform improves the competitiveness of the wholesale business. Every wholesale customer enriches the platform ecosystem.
What the full-stack AI discourse forgets
The architectures, tech stacks, autonomous agents: the dominant discourse describes these components very well. But it often leaves in the shadows what makes the real difficulty of transformation.
First, governance. You don't become "full-stack" by adding AI agents. You become it when AI becomes a structural component of the business model : which requires redefining responsibilities, trade-offs, and rules of the game with partners.
Second, sustainable hybridization between AI and humans. The most convincing cases (Alan with its doctors validating the Mo assistant's responses in less than 15 minutes, Dougs with its accountants available by chat) show that successful "full-stack" doesn't eliminate the human. It repositions them on what truly creates value.
Finally, the transformation of managerial roles. Piloting a platform is not piloting a production chain. It requires moving from direct control to ecosystem animation, from asset ownership to orchestration of third-party resources.
Redesigning the value architecture
What Dougs, Alan or Metro reveal, beyond their individual success, is a shift in value. In their respective sectors, margin is no longer primarily captured on the product or service sold. It is captured on the position occupied in the ecosystem: the one who orchestrates flows, structures data, connects actors.
This logic applies well beyond banking, insurance or accounting. In industry, logistics, B2B services, healthcare, agrifood (everywhere fragmented actors would benefit from being connected) the same opportunity exists.
SMEs and mid-caps that have business expertise, established relationships and accumulated data possess precisely what pure technology players lack: sector legitimacy. AI now gives them the means to convert it into a platform position : without needing to "become the next Alan" or having comparable resources.
Between structure and sovereignty
The real challenge for leaders is not "Will AI replace my business?" Rather: "Which parts of my business should I transform into a platform to remain master of my trajectory?"
This is where "full-stack AI" takes on its full meaning (not as a fantasy of American-style disruption, but as a lever of strategic sovereignty. Between selling commoditized tools and the illusory promise of replacing existing players, the platform offers a ridge line: that of deep, progressive, demanding transformation) where technology, organization and business model evolve together.