AI consultant vs AI automation agency: which one a growing business needs

Put "AI consultant" and "AI automation agency" next to each other on a search page and you will find firms using both words for very different work. Some sell a few weeks of interviews and a roadmap. Some sell a running system. Some sell both and price them separately. For a business of 10 to 250 people the choice matters, because the wrong one costs a quarter and a budget line and leaves you holding something nobody uses.

The way through it is to ignore the label and ask one question: what will I hold at the end? Here is what each kind of help actually hands over, a decision guide, and the questions that sort them out on a first call.

Four kinds of help, sorted by what you hold at the end

The strategy-only AI consultant

What you hold: a document. A list of opportunities ranked by value, a roadmap with phases, perhaps an AI usage policy, a vendor shortlist or a training plan. Delivered as a report or a slide deck, often presented to the leadership team.

At the top of the market this is expensive. When we read published price lists in August 2026, Blue Orange Digital charged $25,000 to $50,000 for its assessment phase and Winder.AI published strategy sprints at £10,000 to £50,000. Smaller firms sell the same shape for much less: £495 to £3,500 for a paid diagnostic in the UK, in our survey of 20 agency price lists.

The good version is genuinely useful. A larger organisation with competing departments, a board asking for a written AI policy, or regulated data that needs a governance decision before anyone touches it, needs this work done by somebody senior and neutral.

The implementation agency

What you hold: a system running in production on your own data, built in accounts in your company's name, with documentation, a warranty and usually an offer to keep it running. The advice still happens, it just happens in the first conversation and the written scope rather than in a separate engagement.

The bad version builds whatever it was asked to build, including things that should not have been built. The good version tells you when the honest answer is no.

The freelancer

What you hold: a working system, often good value, and the risk that one person is the only one who understands it. We have written up the trade-off in detail in agency vs freelancer vs in-house. The short version is that skill is rarely the problem and continuity often is: three of the six broken systems we had scoped for rescue when we wrote that post started with a builder who stopped replying.

The in-house hire

What you hold: a person inside your business, on your systems every day, and a salary that runs through quiet months. Our working figure for a fully loaded US automation engineer, explained in the same post, is $13,500 to $15,000 a month. It is the right answer when there is a full-time queue of automation work. For one project it is the wrong one.

Why advice that ends in a working system matters

This is our opinion, and it is the reason we describe ourselves the way we do: AI consulting should end in a working system, not a slide deck. Three things go wrong when the advice and the build are separated.

The estimates are untested. A roadmap says "automate invoice intake, six weeks, high value". Nobody who wrote it has opened your inbox and seen how many of the invoices arrive as phone photos of paper. The person who builds it finds out in week one, and by then the business case in the deck is fiction.

The hard part is invisible from a workshop. In the systems we run and the broken ones we are called into, the trouble is almost never the model. It is a login that expired with no alert, a spreadsheet that gained a column, a step that assumed a field would never be empty. Why AI automations die in production goes through the list. None of it shows up in interviews. All of it shows up in the first month of running.

Nobody owns the outcome. A strategy consultant is paid when the report is delivered. If the recommendation turns out to be wrong, the cost lands on whoever was asked to build it, usually after the consultant has gone. When the same people advise and build, a bad recommendation costs them, which is the kind of incentive you want.

None of this makes strategy work worthless. It means that for a smaller business, the most useful advice is the advice that gets tested on real work within weeks, by the people who gave it.

Do you need an AI consultant at all?

For most businesses of 10 to 250 people, not a separate strategy engagement. The test is whether you can already name the work. If someone on your team spends hours each week retyping enquiries into the CRM, chasing documents, building the same report or answering the same five questions, you do not need a roadmap to find it. You need someone to look at that job, tell you whether automating it pays back, and build it if it does. The work done by hand every week is a good place to start the list.

You probably do need strategy work first if:

• several departments each want AI money and someone has to decide between them;
• your board, a regulator or a large customer has asked for a written AI policy;
• you handle health, financial or legal data and nobody has decided what may leave your systems;
• you are choosing between large vendor platforms on a multi-year contract.

If none of those apply, a paid strategy phase is mostly a delay with an invoice attached.

A simple decision guide

Your situation	Who to hire	What you should hold at the end
You can name two or three jobs done by hand every week	Implementation agency that scopes before it quotes	One of those jobs running in production, in your accounts
You have no idea where AI would help	A free consultation with a firm that builds, or a short paid diagnostic with the fee credited	A written list with the arithmetic shown, and a fixed price for the first build or a plain no
Board or regulator wants an AI policy	Strategy consultant	A policy document your team can follow
One small, well-defined workflow, tight budget	Freelancer or a small agency sprint	A working workflow, documentation, accounts in your name
Systems already running, nobody looking after them	Agency with a care retainer	Monitoring, a fix window and a monthly report
A full-time queue of automation work	In-house engineer	A person on your team

Questions to ask either kind

These sort a consultant from an agency, and a good one from a bad one, inside one call:

1. What will I hold at the end? A document, a prototype or a running system. Get the answer in writing.
2. Who does the work? The person on the call, or someone you have not met.
3. Have you ever told a client not to build something? A firm that has never said no is selling, not advising.
4. Is there a price, or at least a range, before the call? Published pricing means the scope has been thought through.
5. If there is paid discovery, what is credited, against what, and by when?
6. Whose name is on the accounts? Yours, every time.
7. What happens in month two? Warranty, support window and the monthly running cost.

Ten questions before hiring an AI agency has the longer list, with the answers that should end the call.

Where we sit

We are an AI consulting and automation agency, and we built the business around the argument above. The advice comes first and is free: a 30-minute consultation with the engineer who would do the work, followed within one to two business days by a written proposal with a fixed price, or a plain note that automation would not pay back for you. Then we build it, run it and look after it. Our prices are on the pricing page, and if you want to talk through a specific job, book the free consultation.