Ep. 146: Where Investors See Opportunity with Josef Najm of Thompson Reuters Ventures

Full analysis

A corporate VC gives founders his map of where ad-tech is heading, and the map is more useful than most news breaks. Josef Najm of Thomson Reuters Ventures tells AdTechGod where TR is placing bets: clean data infrastructure, agent verification, and creator-economy tooling. Along the way he drops two claims that should make every measurement vendor sit up, and one framing that should scare any vendor whose moat is just code.

What's actually being decided here isn't a single deal. It's how operators read the direction of travel: is agentic media, first-party measurement, and "vibe-coded" ad-tech a real threat to your roadmap, or founder-conference talk? This is a Type 2 read for most of you, easy to adjust as facts firm up, so the move is to update priors, not to bet the year. Two of Najm's supporting facts (Publicis buying LiveRamp, Fox buying Roku) are flagged unverified in the materials, so treat the specific deals as claims, not settled.

The Market Analyst. Strip out the unverified deal names and the underlying pattern still holds: data assets are the thing agencies and enterprises are paying up for, not media-buying software. Publicis paying for identity infrastructure, if real, says the agency of the future is a data company with a service arm attached. That's the same logic pushing every holdco toward owned graphs. For an informed outsider: the value in ad-tech is migrating from the software that places ads to the data that decides where they go. If you run a measurement or identity business, the question your board should ask is whether you're a target in that consolidation or a casualty of it. Najm says the consolidation wave runs about two more years. That's a window, not a warning shot.

The Skeptic. Let me steelman the case against getting excited. "The majority of the Lumascape can now be vibe coded" is a founder selling his own thesis. Rebuilding a DSP's UI in a weekend is not rebuilding the supply relationships, the spend commitments, the seat integrations, or the trust that took a decade. Code was never the moat. And the Fox-Roku measurement argument assumes a combined entity would rather run its own currency than sell into a market that wants a neutral third party. Buyers distrust seller-graded homework. That's exactly why Nielsen, iSpot, and VideoAmp exist. First-party data makes you self-sufficient for your own optimization, not for the currency the other side of the table will accept.

The Operator. Tuesday morning, none of this changes your reconciliation. But the "know your agent" idea is the one that shows up as a real problem at 90 days. If your DSP starts letting agents place bids, who owns the errant spend when an agent misfires at 2am? Nobody has answered that, and Najm is right that it's a category waiting to be built. The second-order effect: your ops team becomes the liability backstop for automation you didn't write. That's the part the agentic-workflow decks skip. Get the audit trail and spend caps in place before you get the agents, not after.

The Customer / End User. Here the customer is the brand, and Najm makes the most grounded point in the episode: brands will not hand OpenAI direct access to their loyalty and retail systems. That reluctance is real and it's durable. Chief marketers do not want a frontier model rummaging through customer files it might train on. That hesitation is the opening for trusted middleware, and it's the same reason clean rooms got traction. For the non-specialist: companies want the benefits of AI touching their customer data without actually giving the AI the keys. Whoever sits in that gap, holding the data steady while the agent works through a permissioned layer, has a business.

The CFO. The line that should reset a few budgets is the team-structure call: one or two humans running pods of agents inside five years. Treat the timeline as aspirational, but the direction changes how you value a vendor. If headcount-heavy managed-service shops are the ones AI compresses first, you're overpaying for anything priced on bodies. The payback question for every renewal becomes: am I buying proprietary data and network effects, or am I buying labor and code that a smaller team with better tools reproduces cheaper next year? Fund the first. Squeeze the second at renewal.

Where they part ways. The Market Analyst and the Skeptic split on the same fact: does first-party data plus hardware kill independent measurement, or just make one owner self-sufficient while the market still demands a neutral referee? That's the real fork. The Operator and the CFO agree agents are coming but disagree on urgency, build the liability plumbing now versus wait until the tools are real. And everyone quietly bets against the "vibe-coded Lumascape" as a threat to anyone holding real data or real demand.

What it hinges on. Three beliefs. One, that data assets, not software, are what's being bought in this consolidation, which the LiveRamp signal supports if the deal is what Najm says. Two, that buyers still want neutral measurement even when a seller can grade its own work, which history says yes. Three, that brand reluctance to hand AI direct data access holds, creating a middleware lane. The council leans that data and trust are the defensible positions, and that "code is cheap now" is true but mostly irrelevant to whoever owns the data or the demand. Before acting: verify the Publicis-LiveRamp and Fox-Roku deals independently, because two of Najm's load points rest on them, and pressure-test whether any of your vendors are priced on labor that AI compresses.

Prediction: No major independent buyer (a top-10 advertiser or holdco) will adopt a single integrated media owner's first-party data as accepted cross-media currency by the end of the 2027 upfront cycle in mid-2027.

Confidence: Medium. Buyers structurally reject seller-graded measurement, and that pattern has held for decades.

Why: Najm argues a content-plus-hardware owner could self-serve measurement and drop third parties, but that only works for the owner's own optimization, not for what buyers will accept as neutral currency. The entire reason independent measurement exists is that the buy side won't let the sell side grade its own homework, and that incentive hasn't moved. For a seller's first-party graph to become currency, competing sellers and skeptical agencies would all have to trust a rival's numbers, which nobody has ever done voluntarily. The opposite outcome would require buyers to abandon a decades-old insistence on neutrality in under two years, and there's no evidence in this episode that they're even asking to.

Revisit by 2027-07-15: We're right if no top-10 advertiser or holdco has publicly adopted an integrated media/hardware owner's first-party measurement as accepted cross-media currency by the 2027 upfront. We're wrong if at least one does.