AI consultant for a small business: what you should get and what it costs

An AI consultant for a small business should hand you three things: your workflows ranked by what they cost in staff time, working agents that carry the top ones, and a written plan for who maintains it all. Our builds that do this start at $15,000. The version to avoid is advice with nothing running at the end.

That one-paragraph answer took us a while to earn. The question behind it, do we need an AI consultant and what should one cost, is one of the most common questions small business owners ask in forums, and the answers they get are mostly from people selling consulting. We sell builds, so we have a position too. We'll state it plainly and show the math, and you can weigh it the same way you'd weigh anyone else's.

What a good engagement delivers

A 5 to 200 person company needs specific repeat work moved off specific people, and a record of how the company runs that survives when someone leaves. A transformation roadmap delivers neither. Judge any consultant, us included, on three deliverables.

A ranked list of workflows, priced in dollars. Every company has repeat work: the follow-up after each sales call, the brief before each meeting, the weekly numbers assembled by hand, the invoices chased, the same questions answered again and again. A consultant's first job is to put a dollar figure on each one, with the arithmetic visible, so you can see which three matter and which twenty don't. Hours ranked without rates mislead you, because 5 hours of a founder's week costs more than 8 hours of a support rep's.

Working agents, wired into your tools. Recommendations age badly, and an agent drafting the follow-up in your actual CRM after your actual sales calls keeps earning while a slide deck gathers dust. If the proposal ends at "adopt AI tooling" or a pile of SOPs for your team to execute someday, you're buying homework. The build should go live inside the engagement, in the tools your team already uses, and your team should be trained on it before the consultant leaves.

A maintenance plan with a name on it. Feeds break quietly. A calendar export changes format, an API updates, and an agent keeps producing output from stale memory while everyone assumes it's fine. Before you sign, you should know who watches for that after launch and what it costs, whether that's a retainer, your own team with handover documentation, or both. A consultant with no answer here is planning to be gone when it matters.

There's a fourth deliverable most engagements skip, and it's the one that compounds: the memory underneath the agents. Most of how a small company runs lives in one person's head. The pricing exceptions, the reason the big client gets invoiced differently, what was promised on last quarter's calls. When she leaves, it walks out the door with her. A company brain puts meetings, customer conversations and decisions in one searchable place, a source of truth the agents draw on and your team can query. Agents without it write generic email and guess at answers.

Advice-only consulting vs a build partner

The market splits into two kinds of firm, and the proposals look similar until you read what you own at the end.

	Advice-only consultant	Build partner
Core deliverable	Assessment, roadmap, vendor recommendations	Working agents live in your tools
Pricing shape	Hourly or phased, scope grows as findings do	Fixed quote with a written definition of done
Your team's job after	Execute the roadmap yourselves	Use the system; it's running on day one
Who owns the result	You own a document	You own the system and its data
How to verify value	Hard; quality of advice shows up months later	Watch the agent do the work this week

Advice-only firms earn their fee in some situations: a regulated industry where the hard question is what you're allowed to automate, or an enterprise choosing between seven-figure platforms. At 5 to 200 people the problems are usually repeat work eating hours and knowledge sitting in people's heads, and both of those are build problems.

The proposal tells you which kind of firm you're reading. Ask one question of every line item: what is running when this is done? If the answer is "you'll know what to do next," that's a strategy engagement priced like a build.

What it should cost

Published market numbers are scarce, and most pricing pages quote the firm's own fees. The clearest public benchmark is Upwork's, which lists AI engineers at $25 to well over $100 an hour and custom AI development from $5,000 to hundreds of thousands depending on complexity (Upwork's published cost page). Those are one marketplace's historical contract ranges, so treat them as a floor and a sanity check rather than a market rate. Agency and boutique pricing runs higher, and hourly engagements carry a built-in problem: the meter rewards slow work.

Our own pricing, so you have a concrete anchor: builds start at $15,000, fixed quote, written definition of done before anything is signed, delivered in the first month. That covers the brain and the first agents, live in your tools, with the team trained. An optional retainer from $2,500 a month covers new agents and upkeep. Nobody has to keep paying after the build; the system belongs to the client, in plain files on machines they control.

Whatever you pay, demand the payback math in the proposal. The method we use, the same one our audit tool runs, is hours a week on the workflow x the hourly rate of the person doing it x 52 weeks. As an illustrative worked example: a sales follow-up workflow eating 6 hours a week at a $75 sales rate prices at 6 x $75 x 52 = $23,400 a year. Repeat questions pulling 5 hours a week of a founder's time at $150 price at 5 x $150 x 52 = $39,000 a year. Those hours are placeholders to show the shape of the math; your numbers come from your week. When we ran the audit on 11 real companies, the 1 to 5 person teams landed between $53,000 and $116,000 a year in their top 3 workflows, the 16 to 50 person teams between $157,000 and $192,000, and the 51 to 200 person teams between $315,000 and $655,000, counting staff time only.

Against figures like those, a $15,000 build clears quickly, and a consultant who can't produce the equivalent math for their own quote is asking you to take payback on faith. We back ours with a 60-day savings review: 60 days after launch we run a cost review from the brain's own records, and if it finds under $15,000 a year in savings you agree are real, you get up to 2 months of the retainer free. On our own install at Rye Valley, the holding company behind Hockey Stick, the same approach removed $425,000 in annual expenses within about a month, and the team now ships 17x faster. It's one case study from our own books, so discount it as you see fit.

If you want your own numbers before talking to anyone, our free Leverage Audit estimates the 3 workflows where agents would save you the most, with the hours and dollars shown. It takes 2 minutes, and it gives you a baseline to hold any consultant's proposal against, including ours.

When you don't need one

We turn down work that fails these tests, so take them as more than politeness.

Your repeat work is already cheap. If the workflows you'd automate add up to a few hours a week at low rates, the math above won't clear a $15,000 build. Run the numbers first. A team of 3 with one painful workflow might be better served by an off-the-shelf per-seat tool and a Saturday of setup.

You haven't tried the obvious tools. ChatGPT, Claude and their team plans solve the single-player problems well: drafting, summarizing, research. If nobody on your team has pushed those to their limits, do that first. Where they stop is the plumbing: feeds that keep a memory current and agents wired into your CRM, calendar and inbox. You can build a chatbot yourself. The plumbing is what dies as a side project, and it's the legitimate reason to pay someone.

The problem is adoption, and nobody's bought in. If the team ignored the last three tools you rolled out, a consultant becomes an expensive way to buy a fourth. Fix the ownership question first: one person accountable for the workflow, then automate it.

Someone is selling you urgency. The pitch that your competitors are all doing this is a tell. The only reason to buy is arithmetic: hours back, knowledge kept, dollars you can check in 60 days.

Questions that separate the serious from the slideware

Take these into any sales call, ours included. Which three workflows will you automate, and what does each cost us today in hours and dollars? What is live by the end of month one? What do you need from our team during the build? Who owns the system, the data and the prompts when you leave? What breaks quietly after launch, and who notices? What happens if the savings don't show up?

A build partner answers all six in numbers and names. An advice-only firm answers the first one with "that's what the discovery phase determines," which means you're paying them to find out what the proposal should have contained. The consultant worth hiring is the one whose proposal you could hold them to in a 60-day review, because they wrote it expecting you to.