Dave AndersonThe argument · Marketing · 2026
You have six months.
Marketing is about to lose more jobs, faster, than any function in the company. Not to a reorg — to five people with production access to their own website.
You are either excellent with AI, or you are out of a job. I have stopped saying that gently, because every marketer I say it gently to hears “carry on.”
01 · The arithmetic
Marketing efficiency, improved by 4,035%.
Six projects. For each one, the headcount and duration I would have quoted after twenty years of running these functions, set against what it actually took. The conventional column is my estimate and labelled as one. The actual column is counted from git history and session records.
| Work | Conventional | Actual | Multiple |
|---|---|---|---|
| Rebuild the company website | 660 | 16 | 41× |
| Research paper to launch program | 270 | 14 | 19× |
| Campaign, concept to playable | 240 | 1 | 240× |
| Competitor comparison pages | 75 | 0.5 | 150× |
| 90-second product explainer | 48 | 0.125 | 384× |
| Webinar campaign, end to end | 20 | 0.125 | 160× |
| Total | 1,313 | 31.75 | 41.4× |
You are entitled to think my conventional estimates are generous. So halve them.
Halve every single one and it is still 20.7× — a 1,968% improvement. Cut them to a third and it is still nearly 14×. There is no version of this arithmetic that lands anywhere near the 10 or 20% gain the industry is comfortable talking about.
The gap between those two numbers is the whole story.
02 · The number I would lead with
The person on my team who ships the most code is not me.
Of the 275 commits marketing has made to production this year, 103 are mine. I could have left the other number out.
Andrew Brown, who runs growth
Me, the CMO who writes about this
It is the most useful fact I have, because it kills the easiest objection to everything above: that this is one enthusiast with an expensive hobby, a CMO with a side project and a flattering spreadsheet. It is not. It is a way of working that transferred to the person sitting next to me — and he is better at it than I am.
That is also why this is a six-month problem rather than a five-year one. If it only worked for the person who enjoys it, you could wait it out.
03 · What twenty minutes buys
A complete marketing program, built in about twenty minutes.
Not a brief for a program. Not a deck describing one. The program — live, on production, behind a merged pull request with my name on it. No meetings. No planning session. No developer. No creative round.
- The top five AI optimisations the campaign is aboutRequires SME knowledge I do not have
- A recurring 20-minute series, five episodes mapped
- Episode one's run of show, minute by minute
- Landing page copy and the built page, HubSpot form slot marked
- A four-email sequence: invite, reminder, last call, replay
- A blog post
- Slide designs, in progress as this went up
2–3 weeks.
Five people, minimum.
- A subject-matter expert, whose calendar sets the real timeline
- Someone to ideate
- Someone to write
- Someone to design
- Someone to build it
- Someone to run the schedule keeping the others in sync
The part I could not have written
Read these five and decide whether a CMO wrote them.
This is the technical core of the campaign — the countdown the webinar is built around. I can follow every line of it. I could not have produced any of it.
01
Complete visibility and forecasting of AI workloads
The multiplier. Without allocation, every fix below is a one-off that regresses because nobody owns the number.
Allocate every token, GPU-hour and managed-AI dollar to a team, a product and a feature. Forecast per workload. Alert on anomalies before the invoice does.
02
Model selection policies
Frontier models are the default because nobody set a different default.
Enforce which model tier each use case may call. Route classification, extraction and eval traffic to smaller models, enforced at the gateway, with evals so exceptions earn their cost.
03
Token efficiency
Interactive endpoints doing batch work pay full price for tokens a cache would discount.
Prompt caching on repeated system prompts, context pruning, and batch endpoints for work that was never interactive.
04
Bedrock commitment and capacity optimisation
Provisioned throughput bought for a launch spike rarely gets revisited.
Right-size provisioned throughput against p95 utilisation, move batch-tolerant jobs to batch inference, use cross-region inference profiles to absorb burst.
05
GPU and inference infrastructure right-sizing
Classic infrastructure hygiene. It just has not reached the AI stack yet.
Reclaim idle GPU capacity, autoscale endpoints to observed demand, checkpoint training onto spot.
The savings range attached to each of these in the campaign is deliberately left off this page. Those figures are directional industry ranges pending measurement against live data, and the campaign’s own gate says replace them before anything ships. Publishing them here as fact would break the rule the rest of this page is arguing for.
04 · The distinction
Almost every marketer is using AI. Almost none are AI-native.
The difference is not enthusiasm, or prompt quality. It is write access.
An assistant that drafts copy
Saves you an hour.
It makes you slightly faster at the job you already have.
An agent that ships
Replaces the workflow.
It reads the warehouse, builds the page, opens the pull request and measures what happened. That removes the job as it is currently constructed.
So the skill that matters is not “being good at AI.” It is working like an engineer. Version control. A single source of truth. Grounding before generating. Freezing a dataset before anyone writes a word against it. Shipping a commit instead of circulating a draft.
None of that comes from a marketing playbook, and I do not think it can be faked from the outside.
05 · The consequence
We do not need five brand designers. We need one, with AI.
This is the part that gets me disinvited from panels. It is not a story about tools levelling the playing field. It is a story about a small number of people in every marketing organisation becoming dramatically more valuable, and the rest becoming difficult to justify.
Story and brand
Brand lead, copy chief, content strategist, internal comms
One, with agents
Growth and digital
Web team, SEO agency, paid media agency, marketing ops, analyst
One, with agents
Product marketing
Two to three PMMs, plus a competitive intelligence analyst
One, with agents
Social and video
Social manager, producer, videographer, editor, motion designer
One, with agents
Operations
Marketing ops team, project manager, production coordinator
One, with agents
Multiply that across a department and you are not looking at a productivity gain. You are looking at a budget line that halves, twice, and does not come back — both people and programs.
06 · The part nobody is ready for
Agent sprawl is the next cloud bill.
Running this way, I could not tell you exactly how many agents are working on my behalf. Somewhere near a hundred. Not a hundred sessions — a hundred running things, doing work against live systems while I am doing something else.
And the imprecision is the point. This whole page argues for counted numbers over estimated ones, and I cannot put a firm number on my own agent count. Nobody handed me a budget for it. Nobody approved it.
I run marketing for a company whose entire product is finding the waste in cloud spend, so I recognise this shape immediately. It is 2013 and someone just discovered they can spin up instances on a credit card. The productivity was real. The bill, three years later, was also real — and by then nobody could remember who started what, or why.
Visibility
If you cannot say how many agents are running and who owns each one, you do not have an AI strategy. You have an unmetered utility.
Control
Any competent marketer can start ten agents this afternoon. There is no procurement gate in front of agent capacity the way there is in front of software.
Budgeting
No cost centre, no owner of record, no line a CFO can read. The productivity is real. So is the invoice, about three years later.
This is not a reason to slow down. It is a reason to instrument.
07 · What I would do on Monday
Stop evaluating AI tools.
Give your best marketer write access to something that matters, and see what comes back. If nothing comes back, you have learned something important about your team. If a merged pull request comes back, you have found the person your department should be rebuilt around.
Then count your agents.
If you cannot say how many are running, who owns them, and what they cost, you do not have an AI strategy. You have an unmetered utility and a very good quarter.
Counted from git history and session records, not self-reported. The full working log — including the four things this does not solve, and the launch where our own number appeared at three different values on three of our own surfaces — is published in full.
Six months