In January, Omnicom stood on a stage at CES and showed off a rebuilt, AI-powered Omni, the platform it has spent two years describing as the center of everything: the merger logic, the client pitch, the reason a holding company deserves to exist in 2026 at all.

By June 9, roughly 50 U.S. employees in the Omni Platforms Product & Engineering division were laid off. Over June and July, according to documents reviewed by Adweek, at least 468 more, across the U.S., U.K., India, and Malaysia, most of them the product and engineering people who build and run Omni, were transferred to Endava, a publicly traded IT services contractor.

Read that again. Not a tools vendor. Not an acqui-hire. The people who maintain the thing Omnicom calls its differentiation now get their paychecks, their performance reviews, and their marching orders from a company whose entire business model is renting engineers to whoever signs the master services agreement.

Credit where it belongs: Adweek's Kendra Barnett did the hard part. She got the documents, she got affected employees talking, and she put on the record something Omnicom moved quietly over two summer months and clearly hoped would stay quiet. That is real reporting, and this piece does not exist without it. Hat tip.

But the story stopped at what happened. The framing, "Omnicom offloads staffers," treats this as a workforce story. It is not a workforce story.

It is an equity-narrative story. And the narrative just confessed.

John Wren spent $13.5 billion buying Interpublic. The combined company has eliminated roughly 10,000 positions since the deal was announced and promised Wall Street north of $750 million in annual synergies. And the argument for why any of this pain was worth it, the argument made in every earnings call, every town hall, every trade-press victory lap, was Omni. Proprietary AI. Unified data. A technology moat that Publicis and WPP supposedly could not match.

Seven months after the deal closed, the moat's builders are on a contractor's org chart.

Here is the thing about moats: you do not outsource them. Nobody outsources their moat. Google does not transfer the search ranking team to Accenture. Netflix does not hand its recommendation engineers to Infosys.

When a company moves the people behind its claimed differentiation to a services firm, it is telling you, in the language of org charts rather than press releases, what it actually believes that asset is.

Overhead.

Omnicom's statement to Adweek is a small masterpiece of the genre: this is a multi-year partnership to expand engineering capacity, and Omnicom keeps product direction, AI innovation, data science, client relationships, and intellectual property.

Fine. Omnicom owns the sheet music. The orchestra works for Endava now.

IP ownership is a legal fact. Product control is an operational fact, and operational control lives in people: the ones who know the undocumented workarounds, the client-specific integrations, the data pipelines held together with hope and cron jobs, the ones who get paged when a model deployment fails at 2 a.m. during a client's biggest campaign flight.

Those people either just left the building or now bill through a third party whose staffing decisions, utilization targets, and margin requirements Omnicom does not control.

And there is a second problem, the one that should make every Omnicom competitor smile: Endava is not captive. It is a NYSE-listed services firm. Its engineers, its playbooks, its accumulated knowledge of how to build agency AI platforms are, in principle, available to anyone with a budget.

The more of Omni's execution lives at Endava, the harder it becomes to answer the only question a sophisticated client asks about "proprietary" technology: what exactly is proprietary here? The models? The data? Or the commercial wrapper around outsourced engineering?

Omni is not back-office software. It touches audience data, planning inputs, optimization decisions, measurement outputs, the evidence trail used to justify how client money gets spent. So the questions are not HR questions. They are governance questions:

Omnicom may have clean answers to every one of these. Clients should insist on seeing them in writing rather than accepting "we retain the IP" as a substitute, because that phrase answers a question nobody asked.

Strip the branding and this move is legible in about four seconds. Converting fixed internal labor into vendor spend produces near-term restructuring savings, flatters the synergy math, and presents a leaner post-merger cost base. That is not a technology strategy. That is financial engineering with a keynote slide.

Which is the real story, and the uncomfortable one. Omnicom's public position is that Omni is central, irreplaceable, the future. Omnicom's operating decision is that the people who built it were reducible cost. Both of those cannot be fully true. Either the AI moat was never a moat, or Omnicom just handed the keys to it to a contractor to hit a synergy number.

The employees Adweek spoke to already figured this out. They questioned the wisdom of separating a core product from the team that supports it. They built the thing. They would know.

How we reported this: This analysis is based on Kendra Barnett's August 2026 reporting for Adweek, which first disclosed the transfer figures, along with Omnicom's public statements, earnings communications, and SEC filings. ADOTAT has not independently reviewed the transfer documents, and the 468 figure is Adweek's minimum count. ADOTAT sent questions to Omnicom and Endava before publication and will update this piece with any response. ADOTAT has no financial relationship with Omnicom, Endava, or Adweek, no subject received advance review, and this piece is funded by readers, not advertisers.