How to choose an AI marketing consultant

Every disappointing AI marketing engagement I’ve seen starts the same way.

A company knows it needs help with AI but cannot clearly define what that help should look like. So it hires the person who makes the most convincing pitch. The consultant then delivers the same solution they sell to every other client. Six months later, the company is left with a tool nobody uses and a deck nobody reads.

In many cases, the consultant did exactly what they were hired to do.

The problem is that the company hired for the wrong job.

That is really a casting problem, and it is worth naming because the first part of the fix is surprisingly simple. Before you take a single sales call, your team can usually determine what kind of AI expertise you actually need in one focused Zoom meeting.

The harder part comes next: evaluating the person who claims to have that expertise.

That is difficult because you are being asked to judge a skill you may not have yourself, in a category that barely existed a few years ago. There is no standard accreditation, no universally accepted credential, and not even a shared definition of what an “AI marketing consultant” is supposed to do.

Most vendors a CMO hires can be evaluated against years of experience and pattern recognition. You know what good agencies, strategists, developers, and media partners generally look like.

AI is different. The category is too new, the titles are too broad, and the signals of expertise are still noisy.

So this guide is designed to solve both problems.

First, it breaks the broad title of “AI marketing consultant” into the five distinct jobs hiding underneath it. Then it gives you a practical way to evaluate the person sitting across from you, even if you do not have the technical expertise to judge everything they know.

The problem is not access anymore

HubSpot's 2026 State of Marketing report, a survey of more than 1,500 marketers worldwide, put AI adoption among marketing teams at 86%, up from 67% the year before and 41% the year before that. Content Marketing Institute's most recent B2B research, a survey of 1,015 marketers, found nearly the same thing from a different angle: 95% of organizations use AI-powered applications, and 89% use AI tools to create or refine marketing copy.

Then the results get interesting. 58% said content quality improved. 21% saw no change. 12% said quality got worse. On performance, only 39% reported improvement, 34% saw no change at all, and 22% said it was too early to tell.

Near-universal adoption, three years running. Deeply mixed results. Whatever your company is short on, it is likely not access to the tools.

Deloitte's 2026 State of AI in the Enterprise report, a survey of 3,235 business and IT leaders across 24 countries, points at the same gap from a different angle. Worker access to AI tools grew 50% in a single year, and two-thirds of organizations say they are seeing real gains from it. But only 34% are using AI to meaningfully transform how the business runs. Just 30% are redesigning core processes around it, and 37% are still using it only at the surface, with little change to how work actually gets done.

That is the true market condition. Most organizations are at the point where everyone knows just enough to be dangerous. People can get Claude or ChatGPT to write a blog post. What comes out is generic, and it sounds like every other company's blog post, because knowing how to open the tool is not the same as knowing how to work with it. Bad AI output is worse than no AI output.

"AI marketing consultant" describes at least five different jobs

Here is what companies actually turn out to need once you get past the opening request. Find your row before you go shopping.

1. Education and upskilling. Your people have access and are producing mediocre work with it. They are not going to get better on their own, because the gap is skill, not permission.

• What good help looks like: hands-on training built around your real work, not a generic AI 101 webinar. Coaching over months, not a workshop.
• What it costs to hire the wrong role here: you buy a strategy engagement, receive a beautiful roadmap, and hand it to a team that still cannot execute any of it.

2. Executive support and team buy-in. One champion is carrying the whole thing. Leadership talks about AI in all-hands meetings and does not use it. Everyone else is waiting to see whether this is real.

• What good help looks like: change management, executive coaching, and a sponsor who visibly changes their own behavior.
• What it costs to hire the wrong role: you buy a plan the organization quietly declines to follow.

3. A strategic plan grounded in your current state. There is activity everywhere and no order of operations. Nobody can say which use cases matter most, which come first, or how you will know if it worked.

• What good help looks like: an honest assessment of where you are, a prioritized set of use cases tied to business outcomes, a sequence, and a measurement approach.
• What it costs to hire the wrong role: you buy tools and training pointed at problems that were never worth solving.

4. A shared way to move use cases, prompts, and context across the company. Two people are excellent with AI. What they know lives in their heads and their personal chat history. Nothing repeats, nothing compounds, and when they leave, it all leaves with them.

• What good help looks like: documented context, shared prompt and skill libraries, and workflows a new hire can pick up in a week.
• What it costs to hire the wrong role: you keep paying to rediscover things your own company already figured out.

5. Confidence that people can use AI safely. Nobody is certain what is allowed. So they either avoid it or quietly use personal accounts on company data, which is worse.

• What good help looks like: a policy written to permit specific things rather than forbid vague ones, clear data handling rules, and a review standard for anything client-facing.
• What it costs to hire the wrong role: adoption stalls for a reason nobody will say out loud in a meeting.

Most mid-market companies I talk to need number three first. The others get much easier once it exists, and much harder without it.

Diagnose yourself before you take a sales call

Ask your own leadership team five questions. The answers tell you which row you are in.

1. If I asked five people on this team to name our top three AI use cases, would I get the same three?
2. Do we have anyone who is genuinely good at this, and does anything they have built exist outside their own head?
3. When someone here uses AI on client work, do they know what the review standard is?
4. Has any leader on this team changed how they personally work in the last six months?
5. Can we name a number or metric that would move if this went well?

• If question one produces five different answers, you have a strategy problem and you should stop reading tool vendor websites until it is solved.
• If question two comes back empty, you have a knowledge problem.
• If three or four are shaky, you have a governance or sponsorship problem, and no amount of tooling will touch either.
• If five has no clear answer, you have a measurement problem, and you will have no way to distinguish real business impact from AI activity that merely looks like progress.

How to evaluate someone when you cannot evaluate the work

You do not need to become an AI expert to run these. Every one of them is answerable by a real practitioner in about ninety seconds.

Ask what runs on AI inside their own business, and ask to see it live. Their own operation is the tell here, not a client's. Anyone selling this should be their own first case study, and they should be able to open a laptop and show you something running. A consultant who cannot demonstrate their own system is selling you a project they have never personally survived.

Ask what AI does badly right now. A practitioner will have a long, specific, slightly irritated list, because they hit those limits every week. A presenter will give you one vague caveat and pivot back to possibility. That pivot is telling.

Ask what they got wrong on a past engagement and what it cost. In a field this young, expensive mistakes are the actual credential. Anyone who has been doing real work for more than a year has at least one scar. The ones who claim they do not are either new or skirting the truth.

Ask them to describe your operation back to you. This is the one that separates depth from vocabulary. After two conversations, a good consultant should be able to describe how work flows through your marketing team more clearly than your own team describes it. Depth shows up as specificity about your business, not fluency about theirs.

I spent seven years at Salesforce as the person who got brought in to solve the tough problems. One of those engagements was with the digital marketing team at one of the largest beauty retailers in the country. That team was excellent. They were also so deep in the day-to-day of their own business that they had never had time to learn what their marketing software could actually do, which meant their strategic options were quietly narrower than they realized.

I did not know their business better than they did. Not close. What I brought was enough knowledge of the capability to show them options they did not know were available, which let them build more ambitious plans than they had been building. That is the entire value of an outside expert, and it is a useful test. If a consultant is trying to prove they understand your industry better than you do, they are auditioning. If they are showing you options you did not know you had, they are working.

Ask what month thirteen looks like. The honest answer describes your team running things without them. If the answer is a renewal, you now know the business model, and the business model will shape every recommendation you receive.

One more thing worth stealing from the enterprise buyers. The sharpest Fortune 500 marketing leaders I worked with treated outside expertise as a speed decision. They had no interest in whether their team could theoretically figure it out, because the answer was usually yes and the timeline was usually eighteen months. Mid-market buyers tend to do the opposite. They try to prove it can be done in-house first, spend a year finding out, and start the real work from behind.

What you are actually paying for

Rates in this market vary enormously for what looks like identical scope. That spread is not random, and the question that resolves it is simple: am I paying for hours of execution or for decisions about sequence?

Execution is worth less than it used to be and is getting cheaper every quarter. Judgment about what to do first, and what to skip entirely, is the part that has held its value. When you compare two proposals with very different numbers, find out if you are buying execution or true expertise.

Then watch where the money actually goes. The most common way budget disappears in these engagements is software. Vertice's most recent analysis of corporate software spend, drawn from more than $75 billion it processes annually, found 14% of applications sitting entirely unused and another 51% underutilized, meaning the company is using less than half the licenses it pays for on those applications. That was true before AI, and AI has made it worse, because there is now an expensive platform for every problem and a compelling demo attached to each one.

I have watched clients get sold a platform positioned as the answer to everything, and I could see the ending before the contract was signed. It becomes another line item that people find confusing and use at a fraction of capacity. The tool was never the constraint, nor the answer.

If any part of a consultant's compensation depends on which software you buy, ask about it directly and early. It does not disqualify them, but it does mean the recommendation existed before they met you.

Warning signs, and the failure each one predicts

Each of these behaviors maps to a specific outcome you can expect.

They lead with a tool. You will get a purchase instead of a plan, and the plan will be reverse-engineered to justify the purchase.

They cannot name what AI is bad at. They will over-promise, and your team will absorb the correction as extra work six months from now.

They resell what they recommend. The recommendation was made before your discovery call.

Their adoption plan ends at training day. Usage will decay within a quarter. Deloitte's numbers already tell you this: access is not the constraint, and a single training session does not change behavior.

Prompt engineering is the product. Better prompts pointed at work that does not matter is a faster way to produce output nobody needed. The skill is real. Sold as the whole answer, it is one of the more dangerous purchases in this market, because it looks like progress and generates no result.

Their case studies name vendors instead of outcomes. If the proof is "we implemented X at Y," there was no measurement discipline on that engagement and there will not be one on yours.

When you should not hire anyone

Three situations where the honest answer is to keep your money.

Your marketing fundamentals are broken. No clear positioning, no defined buyer, no measurement. AI amplifies whatever is already there. If what is there is unclear, you will produce unclear work faster, at scale, and you will have spent money to do it. You would find significantly more benefit from hiring a fractional CMO or fractional marketing team.

You only need training and you have someone internal who can run it. If your gap is training/education and you have a capable person with time protected for it, buy them a budget and a deadline instead of a consultant.

No executive is willing to own it. Sponsorship is easy and costs nothing. Ownership means someone senior has their name on the outcome. Without that, you are buying a document.

A scorecard you can take into the meeting

Ten yes-or-no questions. Score every firm you talk to, including mine.

1. Can they show me an AI system running inside their own business, live?
2. Did they name at least three things AI currently does badly, without prompting?
3. Can they describe a past engagement that went wrong and what it cost?
4. After two conversations, can they describe our workflow accurately?
5. Does their proposal name business outcomes rather than tools?
6. Does the plan include what happens after training day?
7. Is their compensation independent of which software we buy?
8. Can they tell me what month thirteen looks like without us?
9. Do their case studies contain real, measured outcomes?
10. Did they tell us anything we did not want to hear during the sales process?

Seven or better is a serious option. Five or six is a possible backup option. Anything under five, keep looking.

How Marketri measures against this framework

Fair is fair. Here is where we sit on our own scorecard.

We are strongest at row three (strategic planning). Building a strategic plan from an honest read of where a company actually is, tied to the use cases that matter for that specific business, is the work we do best and the work we believe most companies need first. Rows four and five follow from it, and we do both.

We are not the right call if what you primarily need is a general upskilling program. That is a real need and there are firms built around it. It is not our main focus. And if the missing piece is executive buy-in that does not exist yet, no outside firm can guarantee that for you. We can help a sponsor lead it, but we cannot be the sponsor.

For question one of the scorecard, here are some of the things we run internally at Marketri:

• An AI system that runs deep SEO and AEO audits, builds content plans, and rewrites existing pages, which replaced several hours of manual work per cycle.
• Research agents that keep us current on our clients' industries, saving a few hours a week.
• An operations agent that analyzes our internal utilization data and delivers individual scorecards to each employee through Slack.
• Custom-built Claude skills that generate brand-consistent visuals and assets.
• A research agent that works up every warm lead that comes through our website, which took an hour of manual work per lead.

In regards to the mistakes we have made over the last few years, we have plenty. We started too complex and without the strategic plan we now insist our clients build first, so we pivoted several times early and lost weeks and months doing it. I wanted everything custom, which turned out not to be feasible, and that slowed both our progress and our team's willingness to trust the direction. And we badly underestimated the human side. Adoption was much slower than I expected. It only started to move when we stopped shipping capability and started providing real support to the people who were supposed to use it.

That last one cost us the most, and it is something that I want you to avoid at all costs.

The question worth asking

If you ask a consultant one thing, ask what your team will be able to do without them in a year.

The answer I want you to get, from us or from anyone: you have a strategic plan built around your actual use cases, processes and assets in place that people find easy enough to use that they actually use them, and enough understanding of how AI runs in your company that you can make the next decision yourself.

Another tool is usually not the answer. A strategic plan specific to your business is what will move the needle.

When you're ready to discuss your company's AI needs, book an initial conversation with us to get started down the right path.

Sources: HubSpot, 2026 State of Marketing (survey of 1,500+ global marketers); Content Marketing Institute, B2B Content and Marketing Trends: Insights for 2026 (survey of 1,015 B2B marketers); Deloitte Insights, The State of AI in the Enterprise, 2026; Vertice, analysis of unused and underutilized SaaS applications.