AI consulting services from a team that also builds

An AI consultant helps leadership decide which work to hand to AI, what it will cost, and who will build and run it. A good engagement ends with a decision you can fund, backed by a test on your own data.

Our work has 4 parts: finding the work by interviewing the people who do it and sampling real records, ranking it (covered below), proving the top 1 or 2 ideas with working pilots, and writing the rules on who can use which tools with what data. The strongest candidates are often unglamorous: invoice entry, inbox triage, weekly reporting, the same customer question answered 30 times a day.

The difference from strategy-only firms is that we build. Roadmap estimates come from the people who would quote the production work, so a line that says 4 weeks and $15,000 is a number we would sign. If you build in-house or with another vendor, the build plan is written for them to pick up. You own the register, the roadmap and every pilot, all on your accounts.

We also say no. If a rule-based workflow in a tool you already pay for does the job, the roadmap recommends that, because it is cheaper to run and easier to maintain. Expect the roadmap to contain both kinds of project.

Score every candidate on the same 5 criteria, then fund the best ratio of payback to risk. That usually means starting with work that is frequent, well documented and easy to check. The criteria:

1. Hours per month, measured from system logs or a week of tracking.
2. Money effect: revenue protected or gained, such as faster quotes or inbound leads answered within 5 minutes.
3. Data readiness: whether the inputs sit in a system we can read, and how messy they are.
4. Integration effort: how many systems the work touches and whether they have usable APIs.
5. Risk: what happens when the output is wrong, who would notice and whether a person must approve it first.

Each criterion gets a score from 1 to 5 with a written reason. We still argue through the top 10 with you, because scores miss context: a process that looks perfect on paper might be redesigned next quarter.

Two rules keep the list honest. Anything customer-facing or financial starts with a person approving each output (see human-in-the-loop), and that step comes off only after the error rate has been measured. And the first project should pay back inside 12 months on conservative numbers. Ambitious projects can follow once the first one is live and your team trusts the method.

A first phase takes 3 to 8 weeks from kickoff to leadership readout. Where you land depends mostly on how many teams we interview and how long it takes to get pilots access to the systems they need.

A single-department phase usually runs like this:

Group-wide engagements run longer, mostly because interviews span sites and IT security reviews take time. The scope names the approvals that could slow things down.

Access causes more delay than anything else. If a pilot needs production data, someone on your side has to approve a read-only account or an export. We request it in the first week and keep a list of what is still blocked, with an owner next to each item.

After the first phase, the work moves into delivery: the roadmap names the first build with its price, and production builds typically take 2 to 8 weeks. For a lighter first step, the AI readiness audit covers discovery alone in 1 to 4 weeks (1 to 2 for a snapshot, 3 to 4 for a full audit), without pilots or governance work.

Our AI consulting services cost $12,000 to $40,000 per phase, fixed in writing before work starts. The price only moves if the scope does.

What moves a phase toward the top of the band:

• Business units. One department with 8 interviews is a different job from 4 units across 2 countries.
• Pilots. A phase without one costs less, but the roadmap then rests on estimates. We recommend at least 1.
• Systems access. Pilots that need data from an ERP, a core banking platform or a clinical records system take longer to set up safely.
• Governance depth. A short usage policy is quick. A full governance pack mapped to the NIST AI Risk Management Framework (opens in a new tab) or ISO/IEC 42001 (opens in a new tab) takes longer.
• Regulated data. Healthcare and financial services work adds review time with your compliance team.

Model usage during pilots (OpenAI, Anthropic and similar) is billed at cost with no markup, on its own invoice line.

For context, Leanware's October 2026 AI consultant cost guide (opens in a new tab) puts small AI strategy assessments at $5,000 to $25,000 and mid-size opportunity analyses at $25,000 to $100,000. Our full breakdown of rates and packages is in what AI consulting costs.

Skip a consulting phase if you already know the job you want done. If the brief is "answer inbound leads within 5 minutes and book them into the calendar", you need a build, and a strategy phase would charge you to rediscover what you already know. Go straight to AI automation, and we will tell you on the scoping call if something upstream needs fixing first.

It is also the wrong choice when:

• You have fewer than 10 staff. The process knowledge sits in 2 or 3 heads, and a workshop plus 1 well-chosen build will do more for less.
• There is no budget to build anything in the next 12 months. A roadmap that cannot be funded for a year will be out of date by the time it can.
• You want a vendor-neutral verdict on a large platform purchase. We will give you a view, but we are a delivery firm and we say so up front.

What is not included: cybersecurity audits, penetration tests, legal opinions on AI regulation and consumer app builds. We do none of those. Your counsel should review the governance documents before you adopt them. The production build sits outside the consulting fee too; it is scoped and priced separately, and you are free to give it to someone else.