A memo about what comes next — for the brands, the agencies, and the people thinking about commerce in 2026
Issue 01 — May 2026
v11 — NEW WORLD
A new world
is being written.
The agencies built
infrastructure for the old
one.
↑ WHAT THE AGENCIES BOUGHT
Walled-garden optimisation infrastructure.
~$2bn
↓ WHAT THE CONSUMER DID
Started asking machines instead.
2bn / day
This is a memo about a substrate shift — from the platform-mediated commercial layer the agencies built for, to an LLM-mediated layer the agencies cannot yet operate in. The machine in the middle is jagged. The world it mediates is jagged. The unfashionable virtues — strategy, consistency, things humans actually talk about — are what the new substrate happens to amplify.
§ 01 — The diagnostic concept everything else depends on
The agencies bought construction equipment
for what is mostly gym work.
↘ Construction site
Machines magnify human output. Faster. Cheaper. More.
Banner variants. Translation. Asset resizing. Bid optimisation. The machine is brilliant here.
↘ Gym
The point is the human capability. Outsource the reps; defeat the exercise.
Strategy. Brand stewardship. Cultural reading. Knowing when an idea is regression to the mean. Most agency work.
Helen Toner · Georgetown CSET · NYT, Feb 2 2026: "On a construction site, machines magnify what humans can do. We go to a gym to increase our own capabilities."
§ 02 — One academic finding. The only one that matters here.
The same AI is brilliant on some work and broken on others.
The line between is invisible.
+25% / −19%
HBS · Org Science · Dell'Acqua, Mollick et al · March 2026. Field experiment with 758 BCG consultants, three offices. AI inside its competence: +25.1% productivity, +40% quality. AI outside its competence: −19% accuracy. Same tool. Same humans. The dashboard didn't flag the difference.
§ 03 — The bet, sized. The room it was placed inside.
$2bn of agency AI was built
to win inside a room four companies own.
$2BN
Annual AI investment from the major holding companies. Publicis CoreAI €1bn. WPP £300m + Google partnership. Omnicom Omni AI all-employee, $900m–$1.5bn cost cuts. Every dollar calibrated to extract performance from auction systems the agencies do not own and cannot audit.
META
Andromeda · Advantage+
SNAP/TIKTOK
Closed auction
Roughly $550bn of combined ad revenue sits inside this room. The bet, in one sentence: the walled gardens will continue to be where the customer is.
§ 04 — Four voices. One diagnosis.
The CMOs, the analysts, the ANA, and Wall Street
are saying the same thing.
The agencies are not listening.
↗ FORRESTER · OCT 2025 · UNNAMED HOLDING-CO CEO
By 2028, we'll double profits and halve the people.
Cited by Jay Pattisall. Forrester forecast: 15% of agency jobs eliminated in 2026. 15% of AI decision-makers report any EBITDA lift.
↗ BOB LIODICE · CEO ANA · ADWEEK · APR 2026
So much opportunity from AI — but it comes with… issues. We don't spend much time talking about that.
ANA represents $400bn in marketing spend. 2026 Word of the Year, two words: authenticity + agentic AI.
↗ BILL HOBBIB · CMO DEMANDSCIENCE · DEC 2025
Marketers see dashboards glow green, lead goals exceeded — yet "qualified" leads convert at a low rate.
Survey of 750 senior marketing leaders. Two-thirds say their dashboards show success that fails to translate into revenue.
↗ ARTHUR SADOUN · CEO PUBLICIS · FT · APR 2026
More disruption in 12 months than in 12 years.
3 Feb 2026: Omnicom −11%, WPP −10%, Publicis −9%, Havas −7% in a single day. Barclays: top "AI losers" in European media.
87%
↘ Of orgs report unreliable or inflated marketing intent signals
7%/31%
↘ Of consumers say visible AI marketing increases / decreases brand trust
56%
↘ Of B2B buyers cite generic content as a reason to disengage
2/3
↘ Of UK media spend now bypasses agencies, going direct to platforms (FT)
§ 05 — The room moved while the agencies were tooling for it
The agencies built a Ferrari.
The road is gravel now.
2bn
↗ ChatGPT queries / day, mid-2025
55%
↗ Of Google searches show AI Overviews
+527%
↗ AI-referred sessions YoY (Adobe)
−25%
↘ Traditional search by 2026 (Gartner)
The LLM substrate has no auction. No targeting parameter. No impression metric. No retargeting pixel. No advertising slot for sale, in any reliable form, on Claude or ChatGPT or Perplexity or Gemini. The brand is not a customer of the LLM. The brand is content the LLM might cite.
↘ ChatGPT · Top citations
47.9% Wikipedia · 11.3% Reddit · 6.8% Forbes
↘ Google AI Overviews · Top citations
21% Reddit · 18.8% YouTube · 14.3% Quora
↘ Perplexity · Top citations
46.7% Reddit (above all others)
The new gatekeepers are Wikipedia, Reddit, YouTube, Quora. The surfaces the industry has spent 30 years treating as adjacent to the real work.
§ 06 — The system was already broken. The new customer journey is forcing the question back.
For fifteen years the question was: how do we acquire customers cheaper.
The new customer journey is asking: what does this brand stand for, and what does the world have proof of?
↘ THE OLD QUESTION · ANSWER 01
Performance marketing optimisation. Dashboards say success; the customers don't come back.
↘ THE OLD QUESTION · ANSWER 02
Channel arbitrage on cheap social CAC. Until everyone copied it. The cohort died.
↘ THE OLD QUESTION · ANSWER 03
AI marketing efficiency at scale. Same trade. Same trap. The slop is corroding the brand experience.
↘ THE INDUSTRY CONFESSING · DOLLAR SHAVE CLUB CEO · CITED IN THE ECONOMIST · APR 2026
We lost our edge. We lost our vibe.
Read it as one CEO speaking. Then read it as fifteen years of an industry speaking. Unilever bought DSC for $1bn in 2016 and sold it to PE in 2023 at a fraction. Casper went private at −70%. Glossier reportedly seeking a buyer. Allbirds renamed itself NewBird AI on April 15 2026 and pivoted to AI infrastructure. Customer acquisition costs rose 40–60% between 2023 and 2025 (Daniel Corsten, IE Business School). Each failure was diagnosed as a brand losing its edge. None of them were the diagnosis. The diagnosis is that the question itself was wrong.
The new customer journey — LLM-mediated discovery, agent-mediated transaction, machine-readable everything — is asking the question the industry stopped asking somewhere around 2010. What does this brand stand for? What does the world have proof of? Those are not questions a machine can answer. They are the questions the next section is for.
§ 07 — Why strategy wins. And why the machine cannot do it.
You cannot operate this room
without a brand strategy humans prove through action.
The machine cannot generate that proof.
It can only amplify it.
AI content has commoditised. The slop is corroding the brand experience the LLMs now read. The cost of producing AI marketing has fallen to zero. So has its differentiation value. Same models. Same outputs. Same brand. The LLMs are reading the average and serving it back as the answer.
↘ AUDIENCE 01
LLM
Retrieves you from Wikipedia, Reddit, news, brand-owned content. Reads what humans say about you. Slop is anti-signal.
↘ AUDIENCE 02
AGENT
Acts on behalf of the consumer. Books, buys, recommends. Parses your positioning into machine-legible criteria. Inconsistency disqualifies you.
↘ AUDIENCE 03
CONSUMER
Uses both. Sees the LLM's answer. Trusts or distrusts. Will discount any brand that feels machine-made.
Three audiences. None tolerate slop. All three reward the same thing: differentiated positioning, repeated consistently, proven through action that humans then talk about. The conversation becomes training data. The training data becomes what the LLM retrieves. This is human work. The machine cannot generate the proof. It can only amplify what is already true.
§ End — What the new substrate happens to amplify
AI slop dies.
Strategy wins.
Humans matter.
In a jagged world,
only humans can map the terrain.
END · A MEMO ABOUT WHAT COMES NEXT · MAY 2026
v12 · BILLBOARD
Notes & sources
- Helen Toner — Interim Executive Director, Georgetown Center for Security and Emerging Technology (CSET); former OpenAI board member. The gym/construction-site metaphor for when to use AI was given by Toner in The New York Times on February 2, 2026, and elaborated on her own LinkedIn and Substack ("Rising Tide"). Source for the diagnostic frame in §01 and the close: "On a construction site, machines are essential for magnifying what humans can do. But we go to a gym to increase our own capabilities." Used here as the load-bearing distinction between execution work (construction) and judgment work (gym) — the property the entire piece is built around.
- The Economist — "From Allbirds to Glossier, millennial brands have lost their mojo," April 21, 2026. Source for: Allbirds rebrand to NewBird AI on April 15, 2026; Casper taken private at −70% of value; Glossier reportedly seeking buyer; Unilever sale of Dollar Shave Club majority stake to PE in 2023; Daniel Corsten (IE Business School) on customer acquisition costs rising 40–60% between 2023 and 2025; Rhode (Hailey Bieber) sale to e.l.f. for as much as $1bn; Camille Moore on incumbents copying the DTC playbook.
- 2026 performance marketing data — synthesis of multiple 2026 surveys reporting that 87% of organisations describe their marketing investment signals as unreliable or inflated; 56% of B2B buyers cite generic content as a reason to disengage; 54% of marketing decision-makers say AI-generated content is hurting brand distinction. Sources include IAB The AI Gap Widens (2026), DemandScience 2026 State of Performance Marketing Report (cited above), Basis Technologies coverage of "AI slop" optimisation contamination, and Marketing Dive marketing predictions for 2026.
- Consumer AI trust data 2026 — only 7% of consumers say visible AI marketing increases brand trust; 31% say it actively decreases trust; 63% say AI makes them value human-made things more. Aggregated from 2026 consumer-attitude surveys covered by eMarketer (Shoppers Aren't Impressed by AI-Generated Marketing), Razorfish (Five Consumer Trends Rewriting the Brand Playbook 2026), and Adweek/DoubleVerify AI advertising trends reporting.
- DemandScience — 2026 State of Performance Marketing Report: Exposing the Marketing Data Mirage. December 2025. Survey of 750 senior marketing leaders at $100m–$5bn+ organisations across tech, financial services, manufacturing, healthcare, professional services, conducted by independent research firm TrendCandy. Source for: two-thirds of leaders saying dashboards show success that fails to translate to revenue; AI-generated content findings; 30% budget waste figure for organisations with frequently misleading metrics. Bill Hobbib (CMO DemandScience) named as quoted source.
- Averi / Digital Applied — State of AI in Marketing 2026. March 2026. Source for: only 19% of content marketing teams track AI-specific KPIs; 94% of marketers using AI; 88% using daily; 81% having no measurement framework for AI output quality.
- Supermetrics — 2026 Marketing Data Report. Survey of 435 marketing professionals across US, UK, Germany, Australia, Singapore, October–December 2025. Source for: 80% feeling pressure to adopt AI; only 6% having fully embedded AI; 40% struggling to prove ROI; 52% not owning their data strategy.
- The CMO Survey (Christine Moorman, Duke Fuqua) — early 2026 wave. Source for: customer retention overtaking acquisition as the top CMO priority for the first time in the survey's history; "notable reversal of the historical pattern in which acquisition and brand investment tended to lead"; sharpest decline in marketer optimism since the pandemic.
- CMSWire — The CMO Survival Guide for 2026: Budget Pressure, AI Accountability, and the Metrics That Matter. April 2026. Source for the practitioner quote: "Boards want double-digit cost cuts in 12 to 24 months… so the project sponsor can look like they've failed on AI even when metrics like churn or NPS are improving." Also the framing: "CMOs are being asked to defend AI spending while simultaneously proving that AI improves the customer experience rather than degrading it."
- Spencer Stuart — 2025 CMO Tenure Study. Source for CMO tenure at S&P 500 firms falling to 4.1 years (lowest in more than a decade); 62% of departing CMOs moving to equal or larger roles. Cited via CMSWire.
- Gartner — finding cited by CMSWire that marketing budgets flatlined at 7.7% of revenue heading into 2026.
- ChartMogul / Kyle Poyar — The SaaS Retention Report: The AI Churn Wave. March 2026. Analysis of approximately 2,700 B2B SaaS, 600 B2C SaaS, and 200 AI-native companies. Source for: AI-native products under $50/month seeing 23% gross revenue retention vs B2B/B2C SaaS at 43–45% — 20 points worse. Cassie Young (Primary Venture Partners) coining "gross retention apocalypse."
- Bob Liodice (CEO, ANA) — interview with Adweek, April 2026. Source for: "There is so much incredible opportunity from AI… but it also comes with… issues. We don't spend as much time talking about that in our industry." Liodice steps down at end of 2026 after 30 years; ANA represents 20,000+ brands and roughly $400 billion in marketing spend.
- ANA 2026 Marketing Word of the Year — announced December 2025: for the first time, two words selected, "authenticity" and "agentic AI."
- Ad Age — March 2026 reporting on agency AI tooling. Source for the framing used in §06: "Clients looking for unbiased AI solutions from holding companies may be in for a rude awakening."
- Colleen Kirk (NYIT) — research published in Journal of Business Research, 2025. Source for the "AI-authorship effect" and the "moral disgust" finding when consumers believe emotional marketing is AI-written.
- Woodside Capital Partners — The Next Ad War Is Being Fought Inside LLMs. April 2026. Coverage of Profound's Zero Click SF 2026 conference; the framing of LLMs as "the new shelf"; the data point that 12% of S&P 500 companies are already on Profound; Parag Agrawal's projection that agents will explode web usage by 1,000x or more.
- Profound (James Cadwallader, founder/CEO) — Zero Click SF 2026 conference, April 2026. Source for AEO category framing and S&P 500 adoption figure.
- Gartner — forecast that traditional search volume will drop 25% by 2026 as AI answer engines grow.
- Frase / Amsive / HubSpot AEO research — published Jan–April 2026. Source for ChatGPT 2 billion queries/day, 883 million monthly users; Google AI Overviews appearing in 55% of all Google searches; AI-referred sessions to websites growing 527% YoY through mid-2025; HubSpot's own data showing AEO conversion at 3x organic; insurance site converting at 3.76% from LLM traffic vs 1.19% from organic; ecommerce at 5.53% vs 3.7%.
- Amsive / Profound research — citation source breakdown across LLMs: ChatGPT cites Wikipedia 47.9%, Reddit 11.3%, Forbes 6.8%; Google AI Overviews from Reddit 21%, YouTube 18.8%, Quora 14.3%; Perplexity 46.7% Reddit-dominant. The user-generated-content layer as the new gatekeeper.
- PitchBook (cited via Woodside) — AI-in-advertising acquisitions in 2025 totalled ~$13bn (+61% vs 2024); ~$20bn of VC funding (+57% vs 2024).
- Dell'Acqua, McFowland III, Mollick, Lifshitz-Assaf, Kellogg, Rajendran, Krayer, Candelon, Lakhani — Navigating the Jagged Technological Frontier. Organization Science, March 2026 (HBS Working Paper, Sept. 2023). Source for the +25.1% inside-frontier and −19% outside-frontier figures, the persuasion-bombing finding, and the original Centaur / Cyborg taxonomy.
- Randazzo, Lifshitz-Assaf, Kellogg, Dell'Acqua, Mollick, Candelon, Lakhani — Cyborgs, Centaurs and Self-Automators: The Three Modes of Human–GenAI Knowledge Work. SSRN working paper, December 2025. Source for the Self-Automator category and the upskilling-vs-newskilling-vs-neither finding from a study of 200+ BCG consultants.
- Karpathy, A. — post on X, 25 July 2024, popularising Jagged Intelligence. The verifiability filters in §V draw from his subsequent essays and talks.
- Mollick, E. — I, Cyborg: Using Co-Intelligence. One Useful Thing, March 2024 (and subsequent posts on Centaurs, Cyborgs, and the Jagged Frontier).
- WPP plc — FY2025 results, Form 6-K SEC filing, and Elevate28 strategic plan. Announced 26 February 2026. Source for headcount fall (108,044 → 98,655), £300m annual AI investment in WPP Open, 76,000 monthly active users (90% of client-facing staff), £500m savings target by 2028, three-phase plan, FTSE 100 exit December 2025. Reported by VideoWeek, Futureweek, The British Eye, The Drum.
- WPP / Google — Strategic AI partnership announcement, October 2025. Source confirming WPP's £300m annual AI investment continues into 2025 and beyond.
- Digiday — WPP Media launches its "Manhattan project" tech solution into a crowded AI market. June 2025. Source confirming £300m (~$403m) AI investment, scale of build effort.
- The Drum — Is WPP living proof that Big Tech's multi-billion bet on AI will pay off? February 2025. Source for "3,500 people working on WPP Open, platform updated 30 times a day, more like a tech company than a traditional agency."
- Publicis Groupe — FY2025 results press release, 3 February 2026 and CoreAI launch, 25 January 2024. Source for €1bn total committed CoreAI spend, €900m M&A budget for 2026, 2.3 billion consumer profiles in CoreAI database. Reported by AI Certs News, Campaign Asia, Marketing Dive, Digiday, LBBOnline.
- Omnicom Group — Q4 2025 and Q1 2026 earnings calls; Form 8-K filings; restructure announcement following IPG acquisition, 1 December 2025. Source for Omni AI deployment to all client-facing employees end 2025, agentic-AI rollout in Q1 2026, $900m cost synergy target for 2026, $1.5bn target by mid-2028, 4,000 redundancies, retirement of DDB/FCB/MullenLowe brands, R/GA exit. Reported by Storyboard18, Campaign, Infotech Lead, HR Grapevine USA.
- Futureweek — WPP Chief Exec Leaves Amid Industry AI Pressure. September 2025. Source for the Mark Read departure and the immediate context of the Elevate28 plan.
- UK advertising workforce data — cited in AI Certs News, February 2026, on the largest annual exodus of UK ad-sector staff on record, particularly among younger workers.
- Fishbowl (advertising bowl) — anonymous practitioner posts, December 2025–April 2026.
- Marketing Brew — How Meta's AI push is changing ad creation. April 2026. Source for Andromeda system mechanics, Advantage+ adoption levels (60–70% of agency Meta spend), agency media-buyer quotes on client resistance to AI creative, Meta's stated end-of-2026 target for full ad-creation automation.
- Yellowjack Media / digital advertising trade press — Google AI Max GA, 15 April 2026; Performance Max driving 62% of Google ad clicks (Feb 2026 data).
- Entrustech / industry analyst summaries — Meta projected to surpass Google in global digital ad revenue in 2026 (~$243.5bn vs $239.5bn). Meta + Google + Amazon controlling more than 60% of global digital ad spend.
- Campaign US / PR Week / Storyboard18 — coverage of 3 February 2026 holding-company sell-off following Anthropic's Claude "Cowork" plugin release. Source for Omnicom −11.15%, WPP −10.13%, Publicis −9.24%, Havas −7.48% one-day declines.
- Barclays — buy-side investor survey, published 3 February 2026. Ranked WPP, Omnicom, Publicis the top three "AI losers" in European media.
- Jamie MacEwan / Enders Analysis — senior media analyst quoted in Campaign US, February 2026, on the structural exposure of agencies to platform consolidation.
- Financial Times — April 2026 reporting on the future of marketing. Source for Arthur Sadoun's "more disruption in 12 months" quote, Mark Read's "not where agencies make the money any more" comment, and the data point that more than two-thirds of UK media spend now goes directly to tech platforms, bypassing media agencies entirely.
- Financialcontent / MarketMinute — coverage of Omnicom's $5bn share repurchase authorisation, 18–19 February 2026; UBS price-target revisions following the announcement; the planned 9–11% share-count reduction by end of 2026.
- Seeking Alpha — analyst notes on WPP stock down 70% after losing Mars and PepsiCo as clients (February 2026).
- Forrester — Predictions 2026: Marketing Agencies (Jay Pattisall, principal analyst). October 2025. Source for "agencies will resign their agency in 2026," 15% agency-job-loss prediction, 8% holdco headcount cuts in 2025, 78% of top 80 digital media agencies on PE/VC money, and the unattributed quote "By 2028, we'll double profits and halve the people."
- Forrester — Predictions 2026: Technology & Security (Sharyn Leaver, chief research officer). October 2025. Source for the 25% deferral of AI spend into 2027 and the "AI hype period ends" framing.
- Forrester — Three Questions That Will Define AI In 2026 and the State of AI Survey 2025 (1,400+ AI decision-makers). Source for 15% EBITDA-lift figure and 48% of firms having cut headcount due to AI.
- Forrester — 2026 B2C Marketing Predictions. Source for the 30% display ad budget drop forecast as consumers leave the open web.
- The Drum — Forrester predicts 15% agency job losses in 2026. 3 October 2025. Coverage of the Pattisall report and the holding-company-CEO quote.
- Interactive Advertising Bureau (IAB) / Sonata Insights — 2026 study showing 82% of advertising executives believe Gen Z is positive about AI-generated ads, while only 45% of Gen Z respondents report positive sentiment. Cited in PAGE Magazine (February 2026) and Digiday.
- Gartner — 2026 consumer survey, US: 50% of consumers prefer brands that don't use generative AI in customer-facing messaging. Cited in Klaviyo Consumer Trust report.
- Klaviyo — 2026 AI Consumer Trends Report. Source for the 13% complete-trust figure, 32% trust-penalty figure, the "1 in 5 see slop weekly" data point, and APAC vs. US vs. EU regional trust breakdown.
- EIN Presswire / Mojo Creative analysis (April 2026) — synthesis of multiple consumer studies showing 20–35% engagement penalty for content perceived as AI-generated, the 50% → 19% collapse in consumer enthusiasm for AI between 2024 and 2026, and the 54% AI-fatigue figure.
- PAGE Magazine — Gucci's AI Backlash Signals A Deeper Trust Crisis In Advertising. 27 February 2026. Coverage of the 25 February Gucci AI campaign backlash on the brand's own Instagram comments.
- Ipsos — Consumer Tracker survey, January–February 2026. 63% of US adults say ads in AI search results would make them trust those results less. Cited in eMarketer.
- Marketing-Interactive (March 2026) — Klaviyo APAC analysis: only 5% of APAC consumers fully trust AI-generated brand content (vs. 12% US, 16% Europe).
- Digiday — Backlash grows against AI slop, but marketers remain unfazed. November 2025. The disconnect between consumer sentiment and marketer behaviour.
- Kate O'Neill / KO Insights — The Authenticity Premium: Why Consumers Are Rejecting AI-Generated Content. January 2026. Synthesis of the AI-authorship effect literature.
- Nuremberg Institute for Market Decisions — 2025 study on AI-labelling effects on advertising attitude and purchase intent. Cited in O'Neill, K., The Authenticity Premium.
- Journal of Business Research — 2025 studies on AI-authorship effect; consumer judgment of AI-written emotional marketing as inauthentic, triggering "moral disgust."