How to price AI consulting

1. Work out your floor rate before you pick a model

Your floor rate is what you need to earn plus business costs, tools, and model or API costs you absorb, divided by realistic billable hours. Start with the same planning period for both costs and hours. Subtract sales, admin, learning, and other nonbillable time from the hours available to you.

The floor is the minimum you need to cover your plan, not the price to quote. Estimate all the work a project requires, including meetings and revisions, then check whether the proposed fee covers that effort and its uncertainty. Revisit the calculation when your costs or available time change.

2. Understand the consulting pricing models

Choose a model based on how much you and the buyer know about the work. These consulting pricing models distribute uncertainty differently. Agree on checkpoints and approval rules before starting, whichever model you use.

Hourly or day rate

Fits exploratory work billed for time. You carry less estimation risk; the buyer carries more total-cost uncertainty. Define what counts as billable work and when to review spending.

Hourly with a cap

Fits an evolving brief with a spending limit. The buyer has a ceiling, but you must stop or renegotiate at the cap. Agree on checkpoints so unfinished work is not a surprise.

Fixed project fee

Fits a bounded deliverable. The buyer knows the fee; you carry overrun risk. Explicit exclusions and change approval protect both sides when assumptions change.

Value-based

Fits an outcome the buyer can value and both sides can test. You risk misjudging delivery effort; the buyer risks paying for value that depends on untested assumptions. Validate the baseline and responsibilities first.

Retainer

Fits recurring monitoring, updates, or reserved support capacity. You risk scope creep; the buyer risks paying for poorly defined availability. Set capacity, response windows, and exclusions.

Performance-based

Fits only when attribution, measurement, and your influence are clear. You risk delayed or disputed payment; the buyer risks incentives that favor the measured metric over useful behavior. Avoid it when external factors dominate the outcome.

3. Price the parts of AI work generic guides miss

Scope discovery as its own paid step when data quality, access, or feasibility is unknown. Its deliverable can be a workflow map, sample evaluation, risks, and a recommendation about whether to build. Price evaluation and test sets, failure handling, human review, data access, and permissions as delivery work rather than treating them as invisible extras.

Include documentation and handover in AI automation pricing. Decide who owns the model, API, and software accounts, who receives bills, and who can set usage limits. State whether running costs pass through at cost or are included within an agreed usage allowance, and what happens if usage exceeds it. Prefer client-owned accounts when the client will operate the system.

4. Use value-based pricing in consulting without guessing

Ask the buyer for their own numbers about the current workflow: time spent, frequency, rework, delays, and the cost of errors. Record the assumptions and how they were measured. Estimated value is the cost of the current workflow minus the expected cost of the proposed workflow, including operation and human review. Treat that estimate as a discussion to validate, not a result you have delivered.

Tie a value-based fee to an agreed outcome you can test and influence. Define the baseline, test conditions, acceptance criteria, and buyer responsibilities. If the data or permissions cannot support a meaningful test, propose discovery first. Never promise untested accuracy or savings, or charge against an outcome whose measurement neither side can explain.

5. Answer “what is your hourly rate?” by clarifying scope

When a buyer asks for your AI consultant hourly rate, explain what work it covers and ask about the inputs, integrations, review needs, and definition of done. You can share your calculated rate with those boundaries, but a scoped quote tells the buyer more about the decision than a range alone.

Published rate guides and surveys vary widely by scope, buyer, and region. Check current sources and ask buyers about budget rather than treating AI consulting rates as a universal benchmark. If the budget and responsible delivery do not fit, offer a smaller scope or a phased build with a decision point after discovery.

Once your offer is scoped, use the first-client guide to prepare buyer conversations.

6. Write the quote with explicit boundaries

List deliverables, exclusions, acceptance criteria, client responsibilities, milestones, and the payment schedule. State ongoing costs, account ownership, the support period, and how a change is requested, estimated, and approved. Acceptance criteria should describe observable behavior on agreed test inputs, including failures that go to human review.

Example scope wording: “Build a request review queue for one agreed input source and request type. Test it against the agreed sample set. Completion means requests reach the agreed queue and incomplete or uncertain inputs go to human review. External actions require approval. Additional sources and request types are excluded. Changes require a written estimate and approval before work begins.”

7. Define retainers and support after launch

Price a retainer from the recurring work and capacity you agree to provide: monitoring, reviewing failures, checking model changes, and updating prompts or workflows. Include meetings, evaluation, documentation updates, and response responsibilities in that estimate. Check the fee against your floor and the delivery capacity you reserve.

Set a review cadence, response windows, usage limits, and a support request process. Distinguish fixes within the agreed support boundary from new features or integrations. Explain who handles incidents outside your availability and who approves model changes. A retainer should describe ongoing responsibilities, not imply unlimited work or uninterrupted availability.

8. Review margin, raise prices, and know when to decline

After each project, compare the fee minus absorbed project costs with the hours actually worked. Compare that effective rate with your floor. Include discovery, sales effort attributable to the project, meetings, testing, and support so you can see where your estimate missed work.

Use what you learned in the next quote: adjust the scope, uncertainty allowance, model, or price. Discuss changes to ongoing agreements before they take effect. Decline work when the buyer will not agree to safe access, testable acceptance criteria, a workable budget, or clear responsibility for running the system.

Related guides

• How to become an AI consultant
• How to start an AI consulting business
• How to find your first AI consulting client
• Explore AI Service Lab