How much does AI consulting cost?

Published AI consulting prices run from about $2,500 for a scoping roadmap to six figures for a mid-market build, and monthly retainers start around $2,000 and go up past $15,000. At Spej, the AI Roadmap is starting at $10,000. Design & Build is starting at $20,000. The AI Office is starting at $10,000 a month. The final price depends on scope.

The spread inside each range is mostly scope. A cheap audit is often a questionnaire and a slide deck, and an expensive one includes interviews, workflow mapping and something that runs. Below are the published numbers we could find, what each stage should hand you, and how Spej prices its own work.

By Alex Krutik, Co-Founder and COO, Spej. Former CFO with Big 4 experience.

Published AI consulting prices by stage

Prices as published on November 3, 2026. Tavoko defines mid-market as roughly $50M to $500M in revenue. Tavoko also cites Frogslayer's figure of $30,000 to $120,000 as a typical first-year spend for small and mid-market companies.

What each stage should hand you

The price only means something next to what you walk away with. Here is what we'd expect at each stage, using our own engagements as examples.

Roadmap

You should get your opportunities compared by expected value, a recommended build with scope, costs and timing, and an agreed baseline and way to measure the result. A good roadmap also gives you permission to stop. Ours counts an evidence-based recommendation not to build as a successful result. For a 45-person performance marketing agency, the roadmap work included a tool audit across every department, a benchmark of AI adoption across 12 comparable agencies, and a reporting automation design that identified 137 hours a month in savings opportunities (case study). For a national mechanical contractor, a readiness assessment found six high-ROI automation opportunities and produced one working prototype before the company committed to a larger investment (case study).

Build

You should get a working system built around your existing tools, training for the people who will use it, and the code and documentation in your name. At a commercial plumbing and drainage manufacturer, Spej built a system that identifies every affected product and updates its pricing using the manufacturer's existing business data. $300K+ returned so far. 95% fewer pricing errors. Price updates went from 3–5 weeks to near real time (case study).

Retainer

You should get maintenance of what's live, the next agreed improvement, training and a regular report. At a regional P&C insurer with 200 to 500 employees, Spej worked as the retained AI Office: governance, a roadmap of 12 initiatives prioritized by ROI and complexity, a monthly steering committee with the C-suite, and first working value in 60 days (case study).

What pushes the price up or down

Most of the cost and risk sits in work that doesn't demo well. If a quote skips it, the price is low because the scope is missing.

• Data cleanup. Messy data costs you every time AI reads it, because the model spends effort working out what each piece of data is before it gets to the task. We wrote about this on the Spej Substack in The best way to vet an AI partner is to ask about the ugly work. On our October 16 live stream, I put it this way: "the data cleanup still has to be there. It's just a lot easier to do if you know what you're doing with AI."
• Permissions and access. Once AI is reading your files, someone has to decide who can see what. The people who own the data should make those calls.
• Number of systems. Each system the build connects to adds design, testing and support.
• Security and compliance review. Regulated businesses need controls designed in and reviewed with their security and compliance leads before the build starts, which adds time.
• Who on your side owns it. Projects with a named internal owner and regular check-ins move faster. Our builds run a weekly 30-minute check-in for that reason.
• Fixed scope or open-ended. Fixed scope with an agreed price is easier to budget and to hold a firm to.

How Spej prices

Each stage is its own decision, with its own agreed scope and fee, and you choose whether to continue before the next one.

1. Free readiness call, one hour. We walk through five pillars (Commitment, Guardrails, Data, Capability and Impact) and give you an honest recommendation to explore an implementation, address gaps first, or hold off.
2. AI Roadmap, starting at $10,000. Detailed discovery, a business case and a plan. The final price depends on scope.
3. Design & Build, starting at $20,000. Scope, price, target and measurement period are agreed upfront. Design & Build puts one use case into production in four weeks, includes training, and you own the code and documentation.
4. AI Office, starting at $10,000 a month. Scope, monthly fee and commitments are agreed together. The AI Office puts one initiative into production every month and sends a one-page monthly report.

Design & Build is starting at $20,000. The AI Office is starting at $10,000 a month. The final price depends on scope. More answers are on the FAQ.

The guarantee, in full

Before each paid stage, we agree on what success means and how we'll measure it. If we don't deliver the agreed outcome and you meet the agreed responsibilities, the full fee for that stage is waived or refunded. It requires an agreed baseline, outcome, measurement period, reliable access to accurate data, and fulfillment of the responsibilities defined in your agreement. The guarantee applies to the fee for the relevant stage, and full terms are agreed before work begins. If reliable measurement can't be agreed, we can't offer an outcome guarantee. It isn't a guarantee of a fixed return on every AI investment.

For you as the buyer, that means the baseline has to exist before the work starts. If nobody can say how long a task takes today or how often it goes wrong, the first job is measuring it.

Where AI consulting money gets wasted

• Licenses without adoption. Buying seats doesn't change how work gets done. A tool nobody uses is a failed project, whatever the code looks like, so training and a measured before and after belong in the build budget.
• Building before deciding what to measure. Write down the business lever the project is supposed to move before you buy anything. We covered how on the Spej Substack in Before you buy any AI, write down the lever it's supposed to move.
• Skipping the "don't build" option. Some of the best roadmap results are a recommendation to hold off or fix something else first. A firm that never says that is selling builds.
• Soft benefits nobody priced. Time saved and fewer errors are real, but only if you recorded the starting number.

How to compare three AI quotes

Put each proposal next to these questions:

• Does it name a baseline and a target, and how will you both measure it?
• What happens to the roadmap fee if you continue with a build?
• Who owns the code, documentation and accounts at the end?
• What data cleanup and permissions work is included?
• Who on your team has to be involved, and for how many hours?
• What does month two cost, after launch?
• What happens if it misses the target?

If you'd like a second opinion on where to start, book a free readiness call. MSPs and consultants pricing AI work for their own clients can see how we work with partners on /partners.

FAQ