In a single week of April 2026, four independent voices in marketing wrote about four different problems and didn’t notice they were describing the same one.
Dan Renyi wrote about marketing teams that reset every time a new CMO walks in. Cari Lu wrote about a 2026 budget line no one has a name for yet. Eli Schwartz wrote about the people selling prompt tracking as the new rank tracking. And Chris Long pointed at ClickUp, which published roughly 2,000 AI-generated listicles outside its expertise and has shed an estimated 7 million organic visits since February 2025, per Ahrefs data he cited on LinkedIn.
Four problems. One problem. Marketing has arrived at the governance crisis engineering has been working through since 2024.
Signal one: the budget nobody planned for
Cari Lu put it cleanly in “Your 2026 Marketing Budget Has a Token Problem.” B2B marketing tech spend is rising two to three percentage points year over year, and the driver isn’t another SaaS subscription. It’s inference. Tokens. A variable cost line that behaves nothing like the software line it’s replacing.
The traditional B2B marketing split — roughly 45% headcount, 45% programs, 10% tech — is cracking. One commenter under her post said the quiet part plainly: “We’re used to handing our budget to media and creative agencies. Now we need to allocate budget to AI tooling that replaces agencies.”
Engineering lived this in 2024. CFOs stared at OpenAI invoices and asked why a cost center suddenly had a usage-based billing model. Finance teams learned, painfully, that tokens are not SaaS seats. Marketing is now two years behind that same learning curve, with the added complication that nobody on the marketing side has ever had to model a variable cost that scales with creative output.
A budget that scales with usage requires a governance layer that tracks usage. Engineering built observability for this. Marketing will have to.
Signal two: the compounding problem
Dan Renyi, writing at electricb2b, described the pathology every marketing leader has lived: every new CMO resets the stack. New agencies, new frameworks, new dashboards, new taxonomies. Institutional memory evaporates on a 24-month cycle. Product teams compound. Marketing teams cycle.
Renyi’s argument: AI finally lets marketing compound like product does. Persistent agents, memory layers, reusable prompts, codified playbooks — the same primitives that turned engineering from a craft into an industrial discipline could do the same for marketing.
But compounding requires preservation. Preservation requires ownership. Ownership requires governance. If your AI layer is a stack of ungoverned prompts in personal ChatGPT histories, it resets with the CMO the same way the old dashboards did. Worse, actually — at least the dashboards were in a shared tool.
The dark version of Renyi’s thesis: marketing now has the technology to compound and no habit of doing so. Engineering spent a decade building version control, CI/CD, code review, runbooks. These aren’t features. They’re the muscle memory that lets work persist. Marketing is being handed the ability to compound without any of the connective tissue.
Signal three: authority rotting in place
Eli Schwartz, in “AEO is Not SEO 2.0,” made a point so simple it has been almost universally ignored. LLM responses are personalized. Per user. Per context. Per session state. Per prior conversation. The answer you get isn’t the answer I get isn’t the answer your prospect gets.
Schwartz’s conclusion, which will ruin a small industry of new tools: “Prompt tracking as the AEO equivalent of rank tracking is fundamentally unreliable.”
Rank tracking worked because Google’s ten blue links were, more or less, a public artifact. You could measure them. AEO tools are selling the same dashboard for an artifact that doesn’t exist as a public object. Marketing leaders will pay for those dashboards anyway — because the alternative is admitting they don’t know what to measure, and nobody gets promoted for that.
Engineering had this moment with observability. The first wave of APM tools sold a comforting fiction that more dashboards equaled more control. The second wave sold governance: the ability to know which metrics were meaningful, which were noise, and which decisions each metric was allowed to justify. Marketing is about to buy the first wave. The second wave is where governance lives.
Signal four: shadow AI with consequences
Chris Long’s LinkedIn post on ClickUp was meant to be a warning about content farming. Read carefully, it’s a warning about authority.
ClickUp — a productivity SaaS company — published approximately 2,000 AI-generated listicles on topics adjacent to, but outside, its domain expertise. “Best CRMs,” “best project management for agencies,” dozens of permutations. Ahrefs data cited in Long’s post shows roughly 7 million estimated organic visits lost since February 2025, alongside Google’s ongoing adjustments to how it weighs AI-generated content on domains without topical authority.
The interesting question isn’t whether AI content is allowed. It’s who authorized those 2,000 URLs, and under what policy. My guess: nobody authorized them the way engineering would authorize a production deploy. Someone had access to a CMS, a decent content generation pipeline, and quarterly traffic targets. The governance layer — does this match our editorial policy, does this fit our authority surface, who owns the rollback — either didn’t exist or was overridden.
Shadow AI is not new. Engineering has been fighting unsanctioned Copilot installs and rogue OpenAI keys since 2024. What’s new is marketing shadow AI that can ship 2,000 URLs in a quarter. The blast radius is different. The governance gap is the same.
The same fabric, translated
The pattern across all four signals is worth stating directly, because marketing leaders keep treating each as a separate fire.
A function has acquired a high-leverage, high-variance tool. The tool has variable cost (token problem), persistent value if preserved (compounding problem), uncertain measurement (AEO problem), and unclear authorization boundaries (ClickUp problem). These are the four questions governance answers: what does it cost, how do we keep what we learn, how do we know it worked, and who was allowed to ship it.
Engineering has been answering these questions since roughly the second half of 2024. Not perfectly. Often painfully. But with enough accumulated pattern-matching that a working fabric now exists: cost observability, prompt and context versioning, eval and telemetry discipline, policy and approval flows. We’ve written about this fabric in the context of agent deployment in marketing and advertising’s own arrival at governance. This piece is about the operating system underneath: budget, ops, authority, shadow.
The fabric transfers. The names change. Cost observability for engineering is spend traceability for marketing: per-campaign, per-agent, per-creative, per-prompt. Prompt versioning becomes playbook versioning: the asset is a reusable creative brief with an attached context window, not a code module. Evals become content evals: does this output match brand voice, claims policy, legal constraints, channel-specific rules. Approval flows become editorial gates with teeth.
None of this is exotic. It is, almost boringly, the same governance work engineering has already done, applied to a function that hasn’t yet built the muscle.
Why the two-year lag matters
Engineering’s lag cost was measurable. Incidents. Rollbacks. Regulatory findings. Occasional public embarrassments. The companies that moved on governance in early 2024 spent 2025 compounding; the ones that waited spent 2025 cleaning up.
Marketing’s lag will look different but cost similarly. The ClickUp pattern will repeat at larger companies with more to lose. The token budget line will be the subject of a few bad earnings calls before anyone builds controls around it. The AEO dashboard industry will generate a layer of decisions based on measurements that don’t hold up. And the institutional knowledge the new generation of AI tools could preserve will evaporate with the next CMO transition, because nobody set up the compounding mechanism.
Renyi’s best line, loosely paraphrased: marketing finally has the chance to compound. Whether it does depends entirely on the governance layer that nobody in marketing is yet being asked to build.
The reckoning isn’t that marketing has a governance problem. It’s that marketing has engineering’s governance problem, eighteen months late, with fewer people who have lived it. The good news is the playbook already exists. The bad news is the playbook isn’t labeled “marketing.”
This analysis synthesizes How GTM-as-Product Changes the Game (April 2026), Your 2026 Marketing Budget Has a Token Problem (April 2026), AEO is Not SEO 2.0 (April 2026), and Chris Long’s ClickUp Cautionary SEO Tale (April 2026).
Victorino Group helps marketing teams build governance infrastructure that lets AI compound instead of reset. Let’s talk.
All articles on The Thinking Wire are written with the assistance of Anthropic's Opus LLM. Each piece goes through multi-agent research to verify facts and surface contradictions, followed by human review and approval before publication. If you find any inaccurate information or wish to contact our editorial team, please reach out at editorial@victorinollc.com . About The Thinking Wire →
If this resonates, let's talk
We help companies implement AI without losing control.
Schedule a Conversation