Opella says it has used AI workflows to produce more than 21,000 marketing assets, and is now working with Microsoft on a 14-agent system to review healthcare claims and disclosures. The project puts a useful question at the centre of enterprise AI: can faster content production keep its checks in step?
Microsoft AI Watch analysis
What happened
Consumer healthcare company Opella, which makes brands including Allegra, Bisolvon and Dulcolax, developed an in-house content operation called Wonka using a Microsoft generative AI engine. Consumer Goods Technology reports that the effort has grown from a small Paris team to more than 85 people across six production hubs. Opella says it has increased output thirtyfold and lifted engagement to 50%.
Opella and Microsoft are developing Lumen, a verification platform comprising 14 specialised AI agents. The system is intended to check product claims, medical data, citations and legal disclosures, and to recommend distribution tactics. Opella says human experts will retain approval. Those figures and descriptions are claims reported by Consumer Goods Technology, not independently measured results.
Why it matters
This is a concrete example of AI moving beyond drafting into the review and routing of health-related marketing. If Lumen works as intended, it could help a large content operation check more material without asking people to manually inspect every item from scratch. That matters in a sector where a polished but inaccurate claim is more than a copy-editing blemish.
The scale-up also explains why verification is part of the story, not a reassuring afterthought. Opella estimates that 100 million pieces of health content enter the digital ecosystem each day, with at least half incomplete, misleading or inaccurate. That is the company’s estimate, but the practical challenge is clear: more output can mean more material to check.
Our read
The interesting move is pairing content-generation workflows with a specialised checking system, rather than treating speed as the only scoreboard. Fourteen agents sound like a sizeable review team; the real test is whether they catch errors reliably and make expert review more useful. In healthcare, “human approval” is only meaningful if people have the evidence and time to challenge the machine.
What to watch
- Whether Lumen is deployed beyond development and how Opella measures its accuracy.
- What human reviewers can override, and how disputed claims are handled.
- Whether the reported output and engagement gains hold up as the system expands.
Discussion spark: Should companies use AI agents to check health-related marketing, or should those checks remain primarily with human experts?
Sources and evidence
- How Opella, Microsoft Use Agentic AI to Build Enterprise Content Supply Chain – Consumer Goods Technology (10 November 2026, 15:49 UTC)
not affiliated with or endorsed by Microsoft