Right now.
A snapshot of what I'm actually doing — operating, advising, building, thinking, reading. Updated as the work ships. September 2026.
This is a /now page.
Season 2 finale is out — Where the Surplus Settles16 Fridays, one question.
Two seasons, 16 polls, 786 ballots. Season 1 asked who can be held to an agent's action; Season 2 asked where the surplus settles once agents can act. Every week's ballot and both finales are in the archive below.
Past seasons.
-
S2
Where advantage, money, and power move once agents can act. 8 Fridays · 274 ballots · Nos. 09–16The finale — Where the Surplus Settles → Discussion ↗- W1 Jul 31
When every advertiser has a capable buying agent, what creates the edge?
Proprietary data closed at 78% of 42 ballots — Faster execution, Privileged access, and Clearer mandate split the rest at 9%, 7%, and 4%. Season 2 opened with a thesis that advantage moves to the quality of the mandate once intelligence commoditizes; the room didn’t buy it yet — they still named the asset that won the last cycle.
- W2 Aug 7
When agents decide what gets considered, what becomes the new shelf space?
Retrieval results and Paid recommendations tied at 31% each of 16 ballots — Training data and Default integrations trailed at 18% apiece: the room split evenly between the two levers that already exist to buy visibility, not the newer training-data or default-integration routes.
- W3 Aug 14
What will autonomous buying agents ultimately buy?
Outcomes closed at 57% of 73 ballots — the clearest majority of the season so far — with Impressions at 27%, Audiences 8%, Attention 6%. The room endorsed the week’s thesis: the unit of trade follows the result, not the delivery. A quarter still defend the impression, which is the constituency every outcome-priced model has to win over or route around.
- W4 Aug 21
What will buyer and seller agents negotiate hardest?
Success criteria edged Price 34% to 32% of 55 ballots — the tightest finish of the season — with Audience definitions and Liability at 16% each. Read as a block, two-thirds of the room said the hardest negotiation is no longer the price line but the things that were never really settled before agents had to write them down: what counts, whose audience, who owns the miss.
- W5 Aug 28
Which intermediary function remains most valuable when agents transact directly?
Assuming risk took 42% of 19 ballots, more than the two information functions combined were worth on their own: finding counterparties and verifying results tied at 21%, and negotiating price finished last at 16%. A week earlier a third of the room called price the hardest thing agents will negotiate. Asked what an intermediary is worth paying for, almost nobody picked it. Hard and worth paying for turn out to be different properties, and the room separated them.
- W6 Sep 4
When every seller can claim performance, what wins the bid?
Verifiable outcomes took 54% of 37 ballots — more than exclusive supply, proprietary data and lowest price drew combined. The two scarcity plays the industry has run on for twenty years finished at the bottom: exclusive supply 10%, proprietary data 13%. Lowest price came second at 21%, which is the more interesting number: asked what wins a bid once everyone can claim performance, the room expects the fallback to be price rather than inventory or data. A week earlier the same audience said risk-bearing was the intermediary function worth paying for. Both weeks point the same way — once prediction is abundant, the premium moves to whoever can be held to the claim afterwards.
- W7 Sep 11
If an agent chooses the brand, who owns the customer relationship?
Nineteen ballots split almost four ways: the consumer 31%, the brand and the agent platform tied on 26%, the data holder last on 15%. First to second is about one vote at this sample, so the reading is that the room did not converge, not that consumer ownership won. The split is the result. Four senses of the word own were on the ballot (intent, fulfilment, interface, record), each with a legitimate claim, which is why a question that looks like a disagreement is really a definition problem. The data holder finishing last is the sharpest correction to Week 6: the room had just said verifiable outcomes win the bid, then declined to hand the holder of the record the relationship that comes with it.
- W8 Sep 18
Where will agentic market power concentrate?
Thirteen ballots, the smallest room of the run: executing trades 53%, verifying outcomes 23%, setting objectives 15%, controlling access 7%. Asked to rank its own season, the room put execution first, which is the case law's answer rather than the essay's. The two options that finished last in Week 1 (clearer mandate 4%, privileged access 7%) finished last again as setting objectives and controlling access. On thirteen ballots the order holds and the margins do not. The finale's reply: surplus settles where refusal binds, and execution's binding refusals, first look and last look, are the ones a court has just ordered stopped.
-
-
S1
From standards to the standard that matters. 8 Fridays · 512 ballots · Nos. 01–08The finale — The Answerer of Record → Discussion ↗- W1 Jun 5
If AI can understand “anything,” do we still need standards?
an 87% landslide — the market’s least ambiguous answer: AI raises the value of standards rather than replacing them.
- W2 Jun 12
What will be the first AI-native form of fraud?
68% picked authenticity frauds we already have — synthetic audiences + attribution. The one genuinely agentic failure mode, agency laundering, came last. The blind spot is accountability, not authenticity.
- W3 Jun 19
At what point does an AI recommendation become an AI decision?
A 65% majority put the line at the withdrawal of human review — not automation (4%) or money moving (24%); only 6% said it never crosses. The room read the recommendation-to-decision boundary as an accountability handoff, not a technical one.
- W4 Jun 26
In an agentic market, what becomes the most valuable asset?
Trust closed at a 52% majority — Data second at 33%, Compute and Distribution tied at 8% each, 52 votes. The reflex answer (Data, the asset that won the last cycle) lost decisively to the one this cycle actually makes scarce: whose agent is allowed to act.
- W5 Jul 3
Which metric breaks first when agents become the primary buyers?
Three fragile inference metrics clustered at 28–35% while ROAS sat alone at 5% — the room re-ranked the cluster all week and never moved ROAS.
- W6 Jul 10
Who should be accountable for an agent’s decision?
The Agent Owner closed at 50% of 91 ballots — the season’s biggest room — double Shared Responsibility’s 25%, with vendor and user barely claimed: accountability concentrates on whoever deploys and constrains the agent, and the room read “shared” as the absence of an assignment, not a model of one.
- W7 Jul 17
What will agents optimize for?
Business Outcomes closed at 54% of 51 ballots — beating Lowest Cost (23%) and Highest ROAS (11%) combined, with User Satisfaction last at 9%: asked what agents should actually optimize for, the room chose outcomes over cost or return — the same room six weeks of accountability questions had already been persuading.
- W8 Jul 24
In 5 years, what will be the most important standard?
Trust & Accountability closed at 40% of 81 ballots, ahead of Agent Protocols (29%), Measurement (17%), and Identity (12%): asked which standard will matter most in five years, the room chose accountability over communication, measurement, or identity — the same convergence seven weeks of standards-to-stakes questions had been pointing toward all along.
-
Agentic Transformation
The most active work right now is helping companies move from AI pilots to governed agentic workflows: decision rights, tool access, evaluation, fallbacks, semantic infrastructure, and measurable business outcomes.
I'm also experimenting with a live AdCP Ecosystem prototype — a visual model for how agentic advertising workflows could coordinate signals, identity, governance, creative, buying, clean rooms, and measurement around a live brief.
Three lanes,
in parallel.
- Day jobGlobal Head of Enterprise · Samba TVBuilding the enterprise commercial motion across global markets — product, partnerships, and the buyer-side conversation about agentic measurement.
- FounderNo Fluff AdvisoryOperator-led advisory for AdTech, MarTech, data, AI, and SaaS companies entering or scaling in the US. Fifteen specialized playbooks: six vertical (Audio, Gaming, Travel, Performance & Native, Attention, Mobile App Growth) + nine operating (Growth Operating System, Org Design for Scale-Ups, Enterprise Data Collaboration, Retention & Expansion, Agentic Transformation, Commercial Productization, Commercial Architecture, Outcome Underwriting, Measurement Governance). Three engagement shapes: Audit, Sprint, Retainer. Newest practice area: AI Visibility & GEO Advisory — making vendors the source AI engines cite, every technique run on this site first. Selective.
- StandardsAdCP Founding Member · Signals & Measurement WG Co-LeadCo-leading the Signals & Measurement working group of the Ad Context Protocol — the open standard for agentic advertising. Shipping the reference signals adaptor in the open.
In the open.
- Borrowed TrustA state-of-the-market research paper on open-web native advertising in 2026: a supply-rights business sold as a performance product, the missing causal evidence, what the machines optimise, and 24 native vendors scored on documented capability with the Native Quotient. Published on-site rather than through an outlet.
- What the Machines Can SeeA state-of-the-market research paper on mobile app growth in 2026: what the acquisition engines optimize, what attribution and Apple's privacy postbacks can show, payback after store fees and web checkout, and eleven app-growth platforms scored on documented capability. Published on-site rather than through an outlet.
- How Has Your Data Strategy Changed With Agentic AI at Your Doorstep?Advertising Week AW360 column. Agents will sit between customer intelligence and marketing action — deciding, executing, measuring and adjusting — so the governance question becomes decision rights and action rights, not data access. Accessible data tells an agent what exists; decision-ready intelligence tells it what the data means, whether it may be used, for what, and what action it is permitted to take. DMP to CDP to clean room to warehouse-native each moved control closer to the enterprise; agentic on the warehouse adds decision and action control, with the audit record as what makes the accountability real rather than asserted.
- AI Trust: New Standards For The AI EraDMEXCO column. An LLM can infer what a field means, but it cannot prove the data was authorized, that a signal came from a legitimate source, or that an agent stayed inside its delegated authority. Standards exist so markets can coordinate, not so machines can read — so agent autonomy changes what standards are for rather than removing the need for them.
- Proof and StoryA state-of-the-market research paper on contextual advertising in 2026: what brand suitability, emotion AI and narrative targeting actually deliver against what the vendor field claims, where curation and CTV change the picture, and which capabilities the vendor landscape is missing outright. Published on-site rather than through an outlet.
All bylines → · Latest essay outside the Friday poll: Sixteen Fridays, One Question → · Latest No Fluff Weekly: Privacy-Safe Was the Wrong Word →
Open questions.
- Agent-mediated measurement — what attribution looks like when buyer agents and seller agents negotiate the impression in real time
- Vector-based audience matching — graded similarity scores replacing the audience yes/no binary
- Operator + advisor as a permanent posture, not a transition state
- How thoughtful organizations stay thoughtful at 1,000+ people
Selective
conversations.
- Board roles AdTech / MarTech / data — US, EMEA, APAC
- Advisory mandates Series B Scaleup (75–200 · $10–30M ARR) primary · Series A PMF selective
- Fractional GTM US market entry, BD partnerships
- Speaking DMEXCO, IAB events, podcasts, private salons
- Pro bono Skills-based consulting for mission-driven orgs (since 2015)
- Conversations About what AI is actually changing in advertising
On the shelf.
- Everything You Want Is on the Other Side of Hard
- Reshuffle: Who Wins When AI Restacks the Knowledge Economy
- We Who Wrestle with God: Perceptions of the Divine
Last updated September 2026 — stamped from the newest poll,
essay, or byline on this page, so it can't go stale by neglect.
Drift > 90 days means I owe the commentary an update — feel free to nudge me.
Subscribe via RSS →
Working on something similar?
These are current priorities. If they overlap with a live decision on your side, the scorecard routes you to the playbook closest to it.