How much to charge for AI consulting
We found no dated, primary-source public dataset of what independent AI consultants charge private clients. Here is how to build a rate from your own costs, check it against dated federal data, and match the pricing model to the work.
AI Consultant Guide editorial teampublished updated 9 min read
contents
- How much should an AI consultant charge? Start with a rate floor
- What do federal rates for AI roles look like?
- Hourly, day rate, fixed fee, retainer or value-based: which pricing model fits?
- When should you charge hourly, and when a fixed fee?
- Should you price on the value you create?
- What should sit outside your fee, like model usage and taxes?
- How do you set a first rate and raise it?
length9 min
in brief
- Build the rate from a floor, not an average: the pay you want, plus self-employment tax (a 15.3% rate) and the benefits and business costs you now carry yourself, divided by the hours you can really bill 1,2.
- We found no dated, primary-source public dataset of what independent AI consultants charge private clients. The closest dated evidence is GSA's schedule data on ceiling (not-to-exceed) hourly rates for AI-titled roles: a median of $195 an hour across 165 entries on October 23, 2026, rising from $131 for roles asking 0 to 2 years of experience to $258 for 10 or more 3,4. These are maximum rates that companies list on federal contracts, not freelance rates or amounts paid 5.
- Match the model to how well you can scope the work. Federal rules use time-and-materials only when the extent of the work cannot be estimated accurately, and a firm fixed price when a fair price can be set at the outset 6,7.
- Be careful pricing on promised gains. One study found a 14% average productivity gain from an AI assistant for customer support agents, with minimal impact on experienced workers, and a trial of experienced developers found AI tools increased completion time by 19% 8,9.
- Keep model usage outside a fixed fee: Anthropic, for one, bills API use per token at rates that vary by model and says prices can change 10.
Set your price in two steps. First, work out a floor from your own numbers: the pay you want, the taxes and benefits you now cover yourself, and the hours you can really bill. Then choose a pricing model by how well the work can be scoped: hourly or a day rate when it cannot, a fixed fee when it can. We did not find a dated, primary-source dataset of what independent AI consultants charge private clients, so treat any "average" that does not name its source and period with caution. The closest checkable evidence is the hourly ceiling rates that companies list on U.S. government schedule contracts for AI-titled roles, which ran from about $56 to $485 an hour on October 23, 2026 3.
Below: the arithmetic for a rate floor, the one dated market check we could verify, six pricing models compared, and how to scope work so the price holds. For the market you sell into, see AI consulting niches; if you have not yet taken a first client, start with how to become an AI consultant.
How much should an AI consultant charge? Start with a rate floor
A rate floor is the lowest hourly rate at which your practice pays you what you need: (pay you want + self-employment tax + benefits and business costs) divided by billable hours.
- Pay you want. Start with a salary you would accept for the same work. For scale only, the U.S. Bureau of Labor Statistics (BLS) reports a median of $101,860 a year, or $48.97 an hour, for management analysts in May 2025, most of whom work as consultants on a contractual basis 11. That figure excludes pay for the self-employed 11. It is not an AI consulting rate or a billing rate.
- Self-employment tax. The IRS rate is 15.3%: 12.4% for Social Security and 2.9% for Medicare, and the Social Security part applies only up to an annual earnings limit. Employers handle these taxes for most wage earners; you calculate your own and may deduct the employer-equivalent portion when figuring adjusted gross income 1. People in business for themselves also generally make estimated tax payments 12.
- Benefits and business costs. A salary hides what a job costs: BLS reports benefits were 30.0% of private employers' compensation costs in June 2026, $14.07 of $46.89 per hour worked 2. You now buy your own health coverage, retirement savings, liability insurance, software and training. Use real quotes rather than copying the 30.0%, which is an average across private employers.
- Billable hours. Sales, proposals, admin and learning are unpaid, so you will not bill every working hour. Track your time; until you have data, test two assumptions.
The table uses placeholder inputs, not market data. Replace each with your own.
| Lower billable share | Higher billable share | |
|---|---|---|
| Pay you want (placeholder, near the BLS median) 11 | $100,000 | $100,000 |
| Self-employment tax at 15.3% of that pay (simplified) 1 | $15,300 | $15,300 |
| Benefits and business costs (placeholder) | $30,000 | $30,000 |
| Revenue needed | $145,300 | $145,300 |
| Working hours (46 weeks at 40 hours) | 1,840 | 1,840 |
| Share of working hours you bill | 50% | 65% |
| Billable hours | 920 | 1,196 |
| Rate floor per hour | about $158 | about $121 |
Tax is simplified here: self-employment tax is figured on net earnings, its Social Security part has an annual limit, and income tax comes on top, so have a tax professional check your own version. Read the result as a floor, not a price: it leaves no profit, no cushion for unpaid proposals and no pay before your first client. Tracking your real billable share matters: the same costs give a rate about 30% higher at 50% billable than at 65%.
What do federal rates for AI roles look like?
We found no primary source that publishes what independent AI consultants bill private clients, and BLS wage data leaves out the self-employed 11. The most useful dated, checkable dataset we found is the U.S. General Services Administration's CALC+ Labor Ceiling Rates tool. GSA describes it as showing not-to-exceed ceiling rates awarded under its Multiple Award Schedule contracts, and its API data refresh every day 3,4. GSA says these ceiling prices represent maximum vendor pricing and that discounts at task order awards often occur 5.
On October 23, 2026 we queried the API for labor categories with "artificial intelligence" in the title. It returned 181 entries; after combining entries that repeat the same firm, contract, title, price and experience requirement, 165 remained, from 52 firms on 53 contracts 4. The median current ceiling rate was $195 an hour, the middle half ran from $143 to $260, and the range ran from $56 to $485. Rates rise with the experience a role requires:
| Minimum experience the labor category asks for | Rate entries | Median hourly ceiling rate | Middle half |
|---|---|---|---|
| 0 to 2 years | 22 | $131 | $110 to $162 |
| 3 to 5 years | 61 | $183 | $140 to $236 |
| 6 to 9 years | 46 | $219 | $171 to $269 |
| 10 or more years | 36 | $258 | $204 to $295 |
Source: our calculation from the CALC+ ceiling-rates query of October 23, 2026 (keyword "artificial intelligence"), which refreshes daily, so a search today will differ; the middle half is the 25th to 75th percentile 4. The limits matter as much as the numbers:
- These are maximum rates that companies list on federal schedule contracts, not what a sole practitioner bills a private client. The data do not show what any agency paid, and GSA notes that discounts at task order awards often occur 5.
- Only four entries carry "consultant" in the title, at $112, $174, $188 and $288 an hour. Most of the rest are engineer titles (102 of the 165), followed by scientist, expert, specialist, analyst and developer roles. Titles that do not include the words "artificial intelligence", such as a plain data scientist, are not in this query.
Use the table as a sanity check, not a price list. If your floor lands well above the middle half for your experience level, recheck your costs and billable hours, or look for a niche where clients pay for specialist knowledge, such as those in AI consulting niches.
Hourly, day rate, fixed fee, retainer or value-based: which pricing model fits?
Hourly and day rates sell your time, fixed fees sell a deliverable, retainers sell availability and value-based fees sell a result. Each moves risk between you and the client differently.
| Model | How it works | Fits when | Main risk to you | Guardrail |
|---|---|---|---|---|
| Hourly | Hours worked at a stated rate. In federal time-and-materials contracts the rate includes wages, overhead, administrative expenses and profit 6 | Scope is unclear or likely to change; advisory work; a first engagement | Faster work earns less. Federal rules note that time-and-materials gives the contractor no positive profit incentive for cost control or labor efficiency, so buyers must monitor performance 6 | Minimum block, written estimate, not-to-exceed cap |
| Day rate | A set price for a full or half day | Workshops, training, short on-site stints | Days stretch; travel is unpaid | Define the day in hours and deliverables; agree travel terms |
| Fixed-scope audit | A fixed fee for an assessment with a written output, such as a use-case review | The client is unsure what to build; you want a paid first step | Scope grows inside the audit | List deliverables and exclusions; cap interviews and review rounds |
| Fixed project fee | One price for a defined deliverable, not adjusted for your cost experience 13 | You can list deliverables and price known uncertainties; federal rules call a firm fixed price suitable when a fair price can be set up front 7 | You carry the overrun: federal rules describe maximum risk and full responsibility for costs and profit or loss 13 | Written spec, assumptions, change orders, milestone payments |
| Retainer | A monthly fee for a defined amount of advice, hours or iteration | Work continues after launch: monitoring, tuning, steady advice | Disputes over unused or extra hours | Define monthly scope, response times, rollover, notice |
| Value-based | A fee tied to a measured result | The result is measurable, a baseline exists and you control the levers | AI gains vary widely 8,9, and adoption is up to the client | Measure the baseline, cap the amount at risk, define the metric in writing |
When should you charge hourly, and when a fixed fee?
Federal acquisition rules give a useful test, though private clients are not bound by them. They allow time-and-materials contracts only when the extent or duration of the work, or its costs, cannot be estimated with reasonable confidence 6. They treat a firm fixed price as suitable when a fair price can be set at the outset, including when uncertainties can be identified, costed and accepted by a contractor willing to take a firm price 7. Before you quote a fixed fee, ask three questions:
- Can you list the deliverables as things the client could inspect?
- Can you name and cost the main uncertainties: data access, data quality, security and legal review, and how well a model performs on the client's own cases?
- Are you willing to carry the overrun if you are wrong?
Three yeses support a fixed fee. A no points to hourly or a day rate with an estimate and a cap. The federal rules agree: a time-and-materials order carries a ceiling price that the contractor exceeds at its own risk 6.
One pattern handles the uncertainty: sell a fixed-fee diagnostic first. Its scope is yours to control, it ends in a written finding, and it tests feasibility. Then price the build from what it found. For any fixed-fee piece, write down:
- Deliverables the client can inspect, such as a working workflow, an evaluation report or a runbook, not effort.
- Assumptions and exclusions: data access by a named day, a reviewer who answers within an agreed time, and what is not included, such as new integrations or ongoing support.
- Done: acceptance criteria and a set number of review rounds.
- Payment and exit: milestone payments with an upfront share, a change-order clause, and a feasibility exit if the data cannot support the use case.
Should you price on the value you create?
Value-based pricing ties your fee to a result such as hours saved or errors avoided. It can suit work with a measurable result, but the evidence on AI gains is mixed. In a study of 5,179 customer support agents, an AI assistant raised productivity, measured as issues resolved per hour, by 14% on average and 34% for novice and low-skilled workers, with minimal impact on experienced and highly skilled workers 8. In a randomized trial of 16 experienced open-source developers on 246 tasks, allowing early-2025 AI tools increased completion time by 19%, though the developers had predicted a 24% reduction 9. The settings, tools and dates differ; they show the spread, not a rule.
So do not quote a percentage gain as a promise. Measure the client's baseline first, ideally in a paid diagnostic, and price from the cost you take on plus a range of value the client can verify. If you accept a success-based component, cap the part of your fee at risk, define the metric in writing, and make sure the client controls its half, such as data quality and staff adoption.
What should sit outside your fee, like model usage and taxes?
Model and API usage. Anthropic bills API use per token, with separate rates by model, and says prices and plans can change at its discretion 10. A fixed fee that includes usage makes you the insurer of a cost you do not control. Have the client hold the provider account where possible; otherwise bill usage at cost on its own line, with an estimate in the proposal and a monthly cap.
Taxes. The IRS says individuals, including sole proprietors, generally have to make estimated tax payments if they expect to owe $1,000 or more when filing 12. That is a U.S. rule; ask a tax professional for your amount.
How do you set a first rate and raise it?
Pick a target above your floor, check it against the ceiling-rate range for your experience level in the CALC table, and quote your next three proposals at the target, not the floor. Record each outcome and the stated reason. If every proposal is accepted without a question, the price may be low. If buyers accept your approach but lose on price, try a fixed-fee diagnostic to lower the cost of entry.
Raise rates for new clients first and give existing clients written notice. Rerun your floor each quarter with real cost and time data.
What this page cannot tell you. We found no primary dataset of what independent AI consultants charge private clients. The only rate data we could verify is from one market, U.S. federal schedule contracts, and it shows ceiling rates that companies list, not prices paid. The worked example uses placeholders, and model prices and AI evidence move quickly.
Questions readers ask
What is a good hourly rate for a new AI consultant?
We found no verified benchmark for what new independent AI consultants charge private clients, so work from your own floor: the pay you want, self-employment tax, benefits and business costs, divided by realistic billable hours. With the placeholder inputs on this page that came to about $121 to $158 an hour, before any profit or cushion. As a cross-check only, the U.S. General Services Administration's federal schedule data on October 23, 2026 showed a median hourly ceiling rate of $131 for AI-titled labor categories asking for 0 to 2 years of experience and $183 for 3 to 5 years. Those are maximum rates companies list on federal contracts, not freelance rates or amounts paid.
Should I charge hourly or per project for AI consulting?
It depends on how well you can scope the work. Hourly or a day rate suits work whose extent is unclear or likely to change. A fixed fee suits work where you can list the deliverables, price the main uncertainties and accept the overrun risk. A fixed-fee diagnostic followed by a capped or fixed build is one way to handle both.
Should I charge for a discovery call or an audit?
A short call to check fit is usually a cost of selling. A diagnostic that ends in a written finding, such as a use-case review or an assessment of data readiness, is work with a deliverable, and it should have a price. Writing down the deliverables and exclusions keeps a paid audit from turning into free consulting.
Should I pass model and API costs through to the client?
Keep them out of your fee. Anthropic, for example, bills API use per token, at different rates by model, and says prices can change. Where possible, have the client hold the provider account. Otherwise bill usage at cost on a separate line, with an estimate in the proposal and a monthly cap that needs the client's approval to exceed.
How much should I set aside for taxes as a self-employed AI consultant?
There is no single percentage, and this answer covers the United States only. The IRS sets the self-employment tax rate at 15.3%, made up of 12.4% for Social Security and 2.9% for Medicare; the Social Security part applies only up to an annual earnings limit, and income tax comes on top. The IRS says individuals, including sole proprietors, generally have to make estimated tax payments if they expect to owe $1,000 or more when they file. A tax professional can set the right amount for your situation.
Notes and sources
- 1Internal Revenue Service, “Self-employment tax (Social Security and Medicare taxes)”, 2026-06-27. Checked . Back to text
- 2U.S. Bureau of Labor Statistics, “Employer Costs for Employee Compensation Summary - 2026 Q02 Results”, 2026-09-09. Checked . Back to text
- 3U.S. General Services Administration (buy.gsa.gov), “Pricing: Hourly Labor Ceiling Rates”, 2026-10-23. Checked . Back to text
- 4U.S. General Services Administration, GSA Open Technology, “Digital Experience (DX) CALC+ Quick Rate API”. Checked . Back to text
- 5U.S. General Services Administration (buy.gsa.gov), “Pricing Intelligence Suite”. Checked . Back to text
- 6Federal Acquisition Regulation, acquisition.gov, “FAR 16.601 Time-and-materials contracts”, 2026-03-13. Checked . Back to text
- 7Federal Acquisition Regulation, acquisition.gov, “FAR 16.202-2 Firm-fixed-price contracts: Application”, 2026-03-13. Checked . Back to text
- 8National Bureau of Economic Research, “Generative AI at Work (NBER Working Paper 31161)”, 2023-11. Checked . Back to text
- 9arXiv, “Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity”, 2025-07-12. Checked . Back to text
- 10Anthropic, “Plans & pricing”. Checked . Back to text
- 11U.S. Bureau of Labor Statistics, “Management Analysts: Occupational Outlook Handbook”, 2026-08-27. Checked . Back to text
- 12Internal Revenue Service, “Estimated taxes”, 2026-09-25. Checked . Back to text
- 13Federal Acquisition Regulation, acquisition.gov, “FAR 16.202-1 Firm-fixed-price contracts: Description”, 2026-03-13. Checked . Back to text
How this guide was made. Written by the AI Consultant Guide editorial team from the sources listed above. General information, not financial, tax or legal advice. See how we source and funding and recommendations.