Advisory
AI Marketing Consultant: What AI Should Run, and What It Should Never Decide
You bought three AI tools this year. Output went up. Pipeline did not. That is the normal result, because buying tools does not produce marketing: the value comes from deciding which work a machine runs and which decisions stay with a person, and then running that division of labour every week.
This is not another tool list. It is an operating model a two-person team can actually run, published below as a table you can inspect and disagree with. The fixed-scope version, where the deliverable is the operating model itself, is the $1,000 SMB Growth Plan. Retained work, where the function is run rather than described, starts from $3,000 per month.
What an AI marketing consultant does
An AI marketing consultant designs how a marketing function uses AI, decides which work is automated and which stays human, builds the review process that sits between the two, and then either runs that operating model or hands it over documented. It is a marketing role that uses AI tooling, not a technical role that builds AI systems.
Two adjacent things get confused with it, and both are legitimate purchases for different problems. The first is an agency selling AI-written content by the yard: volume at a low unit price, with the strategic decisions still sitting wherever they sat before. You get more artefacts and the same positioning. The second is an AI implementation consultancy building models, integrations and internal tools. That is engineering, measured in systems shipped rather than in market position changed.
This role sits between them. It designs the marketing operating system, decides what runs through it, and holds the standard on what comes out. The output is a function that produces differentiated work at a pace a small company can sustain, not a stack of subscriptions.
Two different things people mean by AI marketing
Almost every confused conversation about AI marketing is two people using one phrase for two different purchases. They are related, but they are bought for different reasons and measured differently.
One
Using AI to produce and run marketing
Research, drafting, repurposing, list building, sequencing, analysis and reporting, done with tooling instead of with headcount. The reason to buy it is capacity and speed at a size where hiring five specialists is not an option. That is this page.
Two
Getting your brand cited by AI assistants
Whether ChatGPT, Claude, Perplexity or Google AI Overviews can reach your pages, understand them, and quote you when your buyer asks the question you sell against. The reason to buy it is visibility in a channel that did not exist three years ago.
That is the AI visibility advisoryThey compound. A company that publishes structured, specific, first-hand material benefits twice: the production is cheaper because the first drafts are machine-assisted, and the same material is the kind an assistant will quote. But they are still separate purchases with separate success measures, and buying one while expecting the other is the most common disappointment in this category.
The line: what AI runs, what a human decides
Almost nobody publishes this line, which is why the positioning claim behind a lean marketing function is usually unfalsifiable. Here it is, in the form used in my own work, so you can check it against what a seller tells you.
| AI runs | A human decides |
|---|---|
|
|
The failure mode, stated plainly. The biggest risk from AI in B2B marketing is not clumsy prose. It is a confident claim that is not true, published under your name, to buyers who will check: an invented integration, a competitor comparison nobody verified, a statistic with no source, a compliance certification you do not hold. Prose gets edited. A false claim gets quoted back to you in a sales call.
If your real question is whether assistants quote you rather than how fast you can produce, start with the AI visibility advisory instead. It has a diagnostic that will tell you something about your own domain in a few minutes.
The lean AI-assisted marketing team
The five-specialist team existed for a reason. Someone had to write, someone had to design, someone had to run the site, someone had to build and clean lists, someone had to pull the numbers. Most of that was production, and production is exactly the part that has become dramatically cheaper. What has not become cheaper is knowing what should be produced and whether it is any good.
So the honest version of the claim is narrow. One senior operator with good tooling can cover strategy, writing, site changes, list building, sequencing and reporting for a small B2B company. In a normal month that means a handful of substantial pages or posts, one live outreach or email sequence, continuous messaging and site work, and the analysis that decides the next month. It does not mean forty assets. Anyone quoting forty is selling volume.
The boundary that keeps this credible: AI does not do relationship work. Customer interviews, partner conversations, analyst and press relationships, sales calls, community, events and the awkward follow-up that turns interest into a meeting all still cost human hours at the same rate they did in 2019. If your growth depends mostly on those, a smaller team is the wrong purchase and you should hire people. Work written up in detail, without invented metrics, includes how this worked for a data-protection vendor.
This is one person with tooling, not an agency with a bench, so the number of concurrent engagements is capped on purpose. Ask about current availability and the next start date in your first message.
The stack, by job not by brand
Tools change every quarter; the jobs do not. So the stack is organised by the job to be done, and the named products in each row are the ones actually used day to day rather than an aspirational list.
Research and synthesis
A general-purpose assistant with a long context window, pointed at source material you supply rather than at the open web, so the output can be traced back to something.
Drafting
One drafting surface with your brief, voice notes and prior work loaded as context. The quality of the brief decides the quality of the draft; the tool only decides the speed.
Repurposing
Turning a substantial piece into derivative formats: summaries, social versions, email sequences, slide outlines, FAQ blocks that also feed structured data.
List building and enrichment
Sourcing and enriching target accounts and contacts, with verification steps, because an unverified list burns a sending domain faster than bad copy does.
Outbound and sequencing
Sequence drafting and variant testing, with the targeting logic and the actual offer written by a person who understands what is being sold.
Analysis, monitoring and publishing
Reporting pulls, visibility monitoring across search and assistants, and the publishing pipeline that gets approved work live without a developer.
Two rules keep a stack this size from becoming a subscription graveyard: nothing enters without replacing something, and nothing stays that only one person can operate. Ask any consultant to name their tools per job and to say which one they dropped last quarter and why. The second question separates a working practice from a capability deck.
Does AI-assisted content hurt your search and AI visibility?
Every buyer asks this, and the honest answer has three parts. First, search engines are in the business of judging whether a page is useful and original, not of detecting authorship. The observable behaviour is consistent with that: thin, derivative pages lose visibility whether a person or a model produced them, and specific, first-hand pages hold it.
Second, the reason so much AI content fails is that it is undifferentiated, not that a model wrote it. A page assembled from the same ten sources everyone else summarised contains no information the web did not already have. It has no reason to be ranked and no reason to be quoted. That failure is a briefing failure, and it would have produced the same result if a junior writer had done the summarising.
Third, AI assistants raise the bar rather than lower it. They cite sources that state specific, checkable things: original data, clear definitions, dated claims, named methods, real prices. That is an argument for first-hand material and against volume, which is the opposite of how most AI content programmes are set up. The longer treatment of where the human line sits editorially is in AI-generated content vs human-written content, and the mechanics of being found by assistants are in the AI search visibility guide.
What is not claimed: that any of this guarantees rankings or citations. Nobody controls a search algorithm or a model's output. What can be done is to publish material worth citing, make it reachable, and measure what happens.
The engagement
One senior consultant, working monthly, with AI tooling underneath rather than a bench of juniors. You get the person you evaluated: the same person writes the strategy, sets the review standard and does the work. Deliverables are written and yours to keep, in your own accounts, from the day they are produced.
What it costs
At the $1,000 SMB Growth Plan level the deliverable is the operating model itself: the division of labour, the workflows, the prompts and the review steps, documented for your team to run. Retained work, where the AI-assisted function is actually run each month rather than described, starts at $3,000 per month at the Channel Growth level and is scoped individually above that. A fixed-scope $600 Visibility Starter also exists for companies that only want the website and social strategy. These are engagement prices, not hourly rates.
If the gap is marketing leadership rather than marketing production, the same person sells that as a fractional CMO engagement, and the AI-assisted operating model is included inside it. The version of that argument written for small teams is a fractional CMO running a small AI-assisted team.
Common questions
What is an AI marketing consultant?
An AI marketing consultant designs how a marketing function uses AI - which work is automated, which stays human, and what the review process is - and then runs or hands over that operating model. It is a marketing role that happens to use AI tooling, not a technical role that builds AI systems.
Will AI replace my marketing team?
AI replaces production capacity, not judgement or relationships. A small team with good tooling can now produce what a five-person team produced in 2022, but deciding what to say, what is true, and which channel to abandon has not become cheaper.
Is AI-generated content penalised by search engines?
Search engines evaluate whether content is useful and original rather than who or what typed it, so the failure mode is undifferentiated content, not machine assistance. Content that contains first-hand information, specific numbers and a clear point of view performs regardless of how the first draft was produced.
What can a one-person AI-assisted marketing function realistically deliver in a month?
Realistically: a handful of substantial pages or posts, one live outreach or email sequence, ongoing site and messaging changes, and the analysis to decide what happens next. Anyone quoting forty pieces of content a month is describing volume, not marketing.
Do you build custom AI tools or agents for us?
No. This is marketing operations - workflows, prompts, review steps and the tooling to run them - not software engineering. Custom development needs your engineering team or a specialist firm.
How is this different from hiring an AI marketing agency?
An agency's AI offer is usually production capacity sold by volume, with the strategic decisions still made by whoever briefs them. This engagement is the opposite: the senior decisions are the product, and AI supplies the production underneath them.
Will this help us get cited by ChatGPT and other AI assistants?
Indirectly, because structured, specific, first-hand content is what assistants cite. If being cited is the actual goal, the dedicated AI visibility advisory covers that directly, including how to audit and measure it.
Start with the division of labour, not the tool list
Send a short note about what marketing looks like today and which tools are already in use. You get a written reply on what should be machine-run, what should not be, and what the first month would cover. If the answer is that you need people rather than tooling, you are told so.
Month to month after the initial term. No lock-in.
Related reading