Divide the profit you need, plus overheads and a buffer, by the days you can bill, which is far fewer than 260: 179 in the example below, giving about $650 a day. AI should change what you charge per project more than your day rate: if it cuts delivery by 20%, price the outcome or raise the rate to hold income.
Speed doesn't automatically lower your income, but it does expose any pricing built on counting days. A freelancer who sells days and delivers in 16 instead of 20 earns 20% less on the same job, while one who sells the outcome keeps the fee. How much of each project AI can actually touch decides the rest, and it is less than most people assume.
The day-rate formula, built from the income you need
Work backwards from what you need to keep. Every number below is an assumption for an illustrative freelance consultant, in USD. Replace each with your own, and treat the 25% tax set-aside as a placeholder: your real rate depends on your income and the rules where you work, so ask an accountant.
| Step | What it is | Illustrative figure |
|---|---|---|
| 1 | Take-home pay you want after tax | $70,000 |
| 2 | Retirement and benefits you fund yourself (an employer would have paid these) | $6,000 |
| 3 | Personal need (1 plus 2) | $76,000 |
| 4 | Profit needed before tax: $76,000 divided by (1 minus 0.25) | $101,333 |
| 5 | Overheads: equipment $1,000, software $900, AI tools $650, insurance $900, accountant $1,200, workspace $2,400, marketing $900, learning $800 | $8,750 |
| 6 | Revenue needed (4 plus 5) | $110,083 |
| 7 | Buffer for late payers and quiet months, 5% | $115,588 |
| 8 | Billable days (next section) | 179 |
| 9 | Day rate: $115,588 divided by 179, rounded up | $650 |
Two choices in that table are worth defending. Step 2 exists because an employee's package includes things a freelancer must buy: retirement saving, sick pay and cover for holidays. Leave it out and your real income is lower than the target by the amount an employer would have spent. Step 7 exists because a client paying in 60 days is an interest-free loan from you, and some invoices arrive late or not at all. A buffer of 5% is a starting point; if one client is a third of your work, make it bigger.
The result, about $650 a day, is a floor, not a price. It is the rate below which the year doesn't add up. What you actually charge also depends on what the client would pay, which is why the value pricing section later matters more than the formula once you pass this check.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowBillable days: the number most calculators get wrong
A year has 260 weekdays, but nobody bills 260 days. The count falls in steps, and the steps are where calculators lie by omission.
| Deduction | Days | Days left |
|---|---|---|
| Weekdays in a year (52 weeks of 5 days) | 260 | |
| Public holidays (assumed 10) | 10 | 250 |
| Your own leave | 25 | 225 |
| Sick or emergency days | 6 | 219 |
| Selling: proposals, calls, networking | 20 | 199 |
| Admin: invoicing, bookkeeping, tax | 12 | 187 |
| Learning and tool upkeep | 8 | 179 |
That leaves 179 billable days, a utilisation of about 69% of weekdays. The sensitivity matters more than the figure itself. The same $115,588 spread over 150 days needs $771 a day, over 179 days $646, and over 210 days $550. Dropping from 179 days to 150 raises the rate by $125, and rising to 210 lowers it by $96, so the full range spans $221 a day. The day count is the assumption to test hardest, ideally against last year's diary rather than your hopes.
A part-time version of the same sum
Fixed overheads bite harder when you work fewer days. Take a freelancer working three days a week with a $40,000 take-home target and $3,000 of retirement saving. At the same 25% set-aside the profit needed is $43,000 divided by 0.75, or $57,333. Add the same $8,750 of overheads and a 5% buffer and the revenue target is $69,387. Three-day weeks give 156 weekdays. Take off 6 public holidays, 15 days of leave and 4 sick days, then 12 days of selling, 8 of admin and 5 of learning, and 106 billable days remain. $69,387 divided by 106 is about $655, so the rate lands at $660, almost the same as the full-time example even though the target income is far lower. Overheads that don't shrink with your hours are the reason, and it is why part-time freelancers often underprice when they scale every line down in proportion.
Putting the model in a spreadsheet
A calculator website gives you one number. A spreadsheet lets you change an assumption and watch the rate move, which is the only way to learn which assumption matters. Lay it out with inputs in one column and formulas below them.
B2 Target take-home pay 70000
B3 Retirement and benefits 6000
B4 Tax set-aside 25%
B5 Overheads (total) 8750
B6 Buffer 5%
B7 Billable days 179
B9 Profit needed =(B2+B3)/(1-B4)
B10 Revenue needed =(B9+B5)*(1+B6)
B11 Day rate =ROUNDUP(B10/B7,-1)
B13 Days per project, old pace 20
B14 Net saving from AI 20%
B15 Days per project, new pace =B13*(1-B14)
B16 Rate that holds project income =B11*B13/B15
B17 Projects a year at old pace =B7/B13Check it against the table above: B9 returns 101,333, B10 returns 115,588, B11 returns 650, B15 returns 16 and B16 returns 812.5. If you would like an assistant to help build or debug the sheet, the walkthrough in the ChatGPT for Excel guide covers the formula-writing side.
An assistant can also review the structure, as long as the prompt has no client details and asks for review rather than recalculation.
Here are the assumptions behind my freelance day rate: take-home
target, retirement, tax set-aside, overhead list, buffer and
billable-day count. List anything missing from the overheads and
any assumption that looks optimistic. Do not recalculate. I am
checking the structure only.Expect the review to add items most first-time models miss: the cost of replacing equipment on a cycle, cash-flow cover for slow-paying clients, bank and payment fees, professional indemnity or liability cover beyond the basic policy, sick pay, and the cost of any days between projects. It will also question a 69% utilisation if you have no pipeline evidence. Take the missing categories and leave its guesses at amounts alone: the figures are yours to decide.
What AI speed does to the maths: one project, four responses
Start with how much of a project AI can touch. Take an illustrative 20-day project where 12 days are production (drafting, analysis, building) and 8 days are everything else: scoping calls, client meetings, review rounds, handover. AI speeds up production, not the client's calendar. If it makes production 40% faster, the gross saving is 12 times 0.4, or 4.8 days, which is 24% of the project. But checking AI output takes time too. Add back 0.8 days for review and rework and the net saving is 4 days: 20 days become 16. That is the 20% used throughout this tutorial, and it is a more honest basis for pricing than any headline claim that AI makes you twice as fast.
Where the production days sit depends on the trade. For a writer it is drafting and editing, for an analyst building and checking models, for a designer producing concept variations, for a developer writing and testing code, and for a consultant drafting decks and reports. Client calls, review rounds, waiting for approvals and project management stay where they were. Measure your own split from your last three projects before you assume any figure: if production is 40% of the days rather than 60%, the same 40% speed-up saves a third less.
At the $650 day rate, the old fee is 20 days times $650, or $13,000. Here are four ways to respond to the 16-day delivery.
| Response | Fee per project | Effective day rate | Annual picture at 8.95 projects |
|---|---|---|---|
| A. Keep $650 a day and bill the 16 actual days | $10,400 | $650 | Revenue falls from about $116,350 to about $93,080 |
| B. Raise the day rate to hold project income | $13,000 (16 days at $812.50) | $812.50 | Target met if clients accept a rise of 25% |
| C. Quote the outcome at the old price | $13,000 | $812.50 | Target met; the client pays for the result and gets it sooner |
| D. Keep $650 and win more projects | $10,400 | $650 | Needs about two more clients, and winning them costs unpaid days |
The annual figures come from capacity. At 179 billable days and 20-day projects you complete 8.95 projects a year, which is $116,350 at $13,000 each and roughly matches the revenue target. At 16 days the same year holds 11.19 projects, an increase of 25%, but only if the clients exist. Response A is what happens by default if you change nothing: the savings go to the client without a conversation.
Response D deserves a harder look because it sounds free. Filling the freed days means about 2.2 more projects. If each win costs three unpaid days of selling, the sum becomes 35.8 freed days divided by 19 days per extra project (16 delivered plus 3 selling), or about 1.9 projects. Revenue then recovers to about $112,700, within $3,700 of the old year. On paper D nearly works. In practice it needs about a fifth more clients, and most freelancers who rely on D discover the constraint was never delivery capacity but demand.
B and C are closer than they look, and C is easier to sell. Raising a day rate by 25% invites a comparison with other freelancers at $650. Quoting a fixed price for an outcome removes the comparison, because the client is no longer buying days. The tutorial on whether to charge clients less when AI speeds up your work covers the fairness argument from the client side, and consulting pricing models after AI compares fixed fee, retainer and value pricing in more depth. If you want to size the saving before you price it, how much time AI can save a solo consultant each week helps you estimate the production share honestly.
Does the market agree? Checking the rate against your last quotes
The formula gives a floor, and the market sets a ceiling that you can only find by looking at what happens when you quote. Pull your last ten quotes and sort them into won, lost on price, lost for another reason and no reply. Reading the pattern takes ten minutes.
- You won nine of ten with no negotiation. You are probably underpriced. Raise the next five quotes by 10% and watch what happens.
- You lost most to a cheaper rival, repeatedly. Either the value isn't visible in the proposal or the clients you are reaching are the wrong ones. Cutting your rate is the last fix to try.
- You lost on reasons other than price. Timing, scope or trust. The price isn't the problem, and changing it won't help.
The numbers here are illustrations, and the rate they produce isn't a benchmark for any trade. Nothing in this sum tells you what clients in your field pay; your own quote history does that. If the formula says $650 and your last six quotes at about $500 all closed without argument, the formula has told you something about your pricing, not about the market.
Why the tool bill is the smallest number in the sum
The overhead line for AI tools in the example is $650 a year. Here is where that figure comes from, using list prices from the vendors' own pages on 1 November 2026. Prices change often, so treat the table as a template to refill.
| Tool | List price | Per year |
|---|---|---|
| ChatGPT Plus | $20 a month | $240 |
| Claude Pro | $17 a month on annual billing ($200 up front), or $20 monthly | $200 |
| Google AI Pro | $19.99 a month | $240 |
| Gamma Plus | $9 a seat a month on annual billing, $108 a year | $108 |
| Otter Pro | $8.33 a month on annual billing, $16.99 monthly | About $100 |
A lean stack is one assistant at $20, Gamma Plus at $9 and Otter Pro at $8.33: $37.33 a month, about $448 a year. A fuller stack with two assistants (ChatGPT Plus and Claude Pro on annual billing) plus Gamma and Otter is $54.33 a month, about $652 a year, which is the $650 in step 5. Set that against the saving on one project: four days at $650 is $2,600, four times the year's entire tool bill. The arithmetic says that agonising over a $17 subscription while leaving the pricing model untouched is looking at the wrong number.
Two cautions keep the stack honest. Subscriptions overlap: if you use one assistant for 90% of work, the second is a luxury. And free tiers carry real limits, set out in what the free tiers really allow for freelancers, so a $20 plan often pays for itself in avoided friction before it saves any billable days.
Value pricing: the alternative to selling days
Value pricing means the fee is tied to what the work is worth to the client rather than how long it takes. It suits AI-assisted work because the faster you deliver, the wider the gap between your cost (days) and the client's benefit (a result). It needs a measurable outcome and a client willing to share the number.
An illustration. A client's reporting process eats 15 hours of staff time a week. At an internal cost of $35 an hour over 48 working weeks that is 15 times 48 times 35, or $25,200 a year. You can automate it in 8 delivery days with AI help. Your floor is days times rate: 8 times $650 is $5,200. Three packages sit above the floor.
| Package | What the client gets | Fee | Share of first-year value |
|---|---|---|---|
| Fix | The automated report and a handover document | $5,200 | About 21% |
| Fix and training | The above, plus a half-day team session and a 30-day check | $8,400 | One third |
| Fix and support | The above, plus quarterly updates for 12 months | $12,600 | One half |
On the middle package the effective rate is $8,400 over 8.5 days (8 of delivery plus the half-day session), or about $988 a day. That is where AI-assisted work earns its margin, and the client still pays well under a year's saving. The value figure has to be the client's own number, agreed with them. If you invent it, the fee looks arbitrary, and if they disagree with it, you have learned the real size of the opportunity. Fix the scope tightly, state what is excluded, and check that the outcome is within your control; if it depends on things you can't influence, quote milestones instead.
Telling existing clients: a rate note that doesn't apologise
New clients get the new pricing from the first quote. Existing clients need notice, a start date and a reason that isn't "AI makes me faster", because that sentence invites the discount you are trying to avoid. The honest reasons are scope, value and predictability. A fixed fee agreed up front is easier for a client to approve than an open-ended day count, and that benefit is real.
Subject: Rates from [date]
Hi [first name],
From [date] my day rate moves to $[new rate]. The project we
are working on now stays at the agreed fee. For future work I
price by outcome where I can, usually as a fixed fee agreed up
front, so you know the total before we start.
Happy to talk it through, and to put a fixed-fee option next to
a day-rate option on the next quote so you can compare.
Best, [your name]Offering both options on the next quote is the useful move. Clients who prefer day rates stay comfortable, and clients who like certainty move to fixed fees, where your margin sits. Whether to mention AI use at all is a separate question with its own answer, covered in should freelancers tell clients they use AI. Where a client explicitly asks about it, or a contract requires disclosure, say so plainly.
A quarterly check on effective day rate, not just the quote
The rate you quote is a wish. The rate you earn is the fee divided by the days you actually spent. Once a quarter, list your recent projects and compute it. An illustration with three projects:
| Project | Pricing | Fee | Days quoted | Days actual | Effective day rate |
|---|---|---|---|---|---|
| A | Fixed outcome fee | $13,000 | 20 | 16 | $812.50 |
| B | Fixed fee, scope crept | $6,000 | 9 | 11 | $545 |
| C | Day rate at $650 | $7,800 | 12 | 12 | $650 |
| Total | $26,800 | 41 | 39 | $687 |
The weighted figure, $687, is above the $646 floor, which is good news that hides a problem. Project B earned $545 a day because the scope crept by two days, and fixed fees punish loose scoping in a way day rates never do. So the check has two outputs: whether the blended rate clears the floor, and which project fell below it and why. A fixed fee works when the scope is written down tightly, extras are priced as change requests, and review rounds are capped. AI shortens production, which makes B-type overruns easier to fall into: the draft arrives quickly, the client sees it and asks for another direction.
Track three other numbers while you are there. Actual billable days so far this year against the 179 pace. The review time you actually spend on AI output, against the 0.8 days per project assumed here. And whether the tool bill still matches what you use. If review is running at twice the assumed 0.8 days, the net saving is 3.2 days (16%), a project takes 16.8 days, and the rate that holds project income is $13,000 divided by 16.8, about $774. The figure matters less than where it comes from: your measured review time, not a vendor's speed claim, sets it. If you'd rather map where AI really saves time in your delivery before you reprice, that is what my AI implementation consultation covers. For a sector example of the same logic, grant writer rates when AI drafts faster runs the numbers for a trade where the fee is often fixed by the funder's budget rather than the writer's days.
Day rates and AI: questions freelancers ask
Should I quote an hourly rate or a day rate?
A day rate removes the tracking burden and suits work where thinking time doesn't fit an hourly clock. Convert to hours only when a platform or contract demands it, and then divide by the hours you actually bill in a day, not eight. Charge a little over half a day rate for half days, because the context switch costs the rest. For AI-assisted work an outcome fee often fits better than either.
What if a client asks for a discount because I use AI?
Price the outcome, not the effort. Show what they receive, by when and with what guarantee, and offer faster delivery or added scope rather than a lower price. If their contract or policy requires disclosure of AI use, comply with it. A discount for speed is a choice, not an obligation: the client is buying the result, not your hours.
Do I charge clients separately for AI subscriptions?
Treat subscriptions as overhead inside the day rate, as the worked example does with about $650 a year. Charge separately only for usage that scales with one client's project and is agreed in advance, such as pay-as-you-go credits for a large batch job. Small separate line items for chatbot subscriptions tend to invite haggling over your tools.
How often should I recalculate my day rate?
At least once a year, and again whenever your billable days change by more than about 15% or you change tools. Each quarter, work out the effective day rate per project (fee divided by actual days) and compare it with your target. If projects keep landing below it, the fix is usually scope control or fixed-fee pricing rather than a higher headline rate.
Further reads
- How to Price Videography Services: Day Rate, Project Fee or Package — Day rate, project fee or package for a creative trade.
- How Freelancers Use AI for Invoices, Proposals and Chasers — Invoices, proposals and payment chasers with AI.
- Can AI Tell You What to Charge? The Limits of AI Pricing Research — What AI pricing research can and can't tell you.
- AI Software for Copywriting: A Freelancer's Monthly Tool Budget — A freelancer's monthly AI tool budget, itemised.
- How Can AI Help Freelancers? Proposals, Invoicing and Late Payers — Proposals, invoicing and late payers with AI.
- ChatGPT for Freelancers: Turn One-Off Clients Into Monthly Retainers — Turning one-off clients into monthly retainers.
- ChatGPT Prompts for Freelancers: Scope Creep, Chasers and Updates — Prompts for scope creep, chasers and updates.
- How Much to Charge for Corporate Training: Day Rates With AI — The same day-rate maths run for corporate trainers.
- How to Price Photography Packages: A Cost-per-Shoot Model With AI — A cost-per-shoot sheet that counts every hour behind a session, then a three-rung package ladder and print add-ons priced from it.
- Best AI Tools for Freelance Writers: Research, Editing, Client Admin — Tools for freelance writers sorted by client-trust risk, with prices, sample prompts, a source ledger and wording for telling clients.
- Video Production Pricing Sheet: Build a Rate Card With AI — A rate card for video work built from your own costs: day rates, edit hours, revision rules and licence fees, with a worked brand-film quote.
- Virtual Assistant Pricing When AI Does the Drafting: Rates and Packages — How a virtual assistant can price hourly, by package or on retainer once AI cuts drafting time, with worked sums and a repricing message.
- Best AI for Freelancers: Check a Client Contract for Red Flags First — A clause-by-clause checklist for using AI to spot payment, IP, revision and liability red flags in a client contract, plus what still needs a solicitor.
- Franchise Disclosure Document With AI: What a Consultant Can Prepare — Which of the 23 FDD items AI can prepare from your data, and which ones a franchise lawyer must draft, approve and sign off.
- Franchise Operations Manual With AI: Structure, Drafting and Review — Build the manual's table of contents from a site visit, draft SOPs with AI, and prove each one with a franchisee walk-through.
- Freelancer Brief Template for Agencies: A Checklist and AI Draft — The sections an agency brief needs, a filled example with acceptance criteria, and an AI prompt that drafts the brief from your notes.
- Upwork AI Policy: Can Freelancers Use ChatGPT on Client Work? — What Upwork's terms actually say about generative AI, where client contracts override them, and disclosure wording to use before you start.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Claude pricing page, Gamma pricing page, Otter pricing page, Google Gemini plans page and OpenAI pricing page for tool list prices, read on 1 November 2026 and checked again on 4 November 2026. Income, tax, overhead, day-count and project figures are illustrative assumptions, not statistics.