A freelancer can use ChatGPT to turn a one-off client into a monthly retainer by pasting the project history, asking it to list the client's recurring needs, and drafting an offer with a fixed monthly scope, a number of hours, response times and a price you set yourself. The offer then goes out in a short email straight after delivery.
The offer should include four things: what is covered each month, how many hours or tasks, how quickly you respond, and the fee with its terms (start date, notice period, what happens to unused time). If any of those is missing, the retainer becomes an open-ended favour, which is the main reason freelancers dislike retainers.
Why a finished project is the best retainer lead you have
You already know the client's business, their tools and their habits. They have already trusted you with something that mattered. And most projects leave a trail of small recurring needs: updates, tweaks, reports, content, fixes, questions. Clients rarely ask for a retainer; they ask for "just one quick thing" six times and you quote each separately or do them free. A retainer gives those jobs a home, a price and a limit.
ChatGPT is useful because the hard part is not the idea but the paperwork: working out what the recurring needs are from messy history, sizing the hours, and writing a clear offer that does not sound like a contract. Those tasks are drafting and sums, which is what the model does well. The decisions stay yours: whether you want the client, the price and the limits.
Step 1: mine the project history for recurring work
Gather what you have: the original proposal, the email thread, your time log if you kept one, and any "quick favour" requests during the project. Remove client names and anything under confidentiality, replace the client with "Client", and paste. If you work on Plus or another individual plan, turn off "Improve the model for everyone" under Data controls first, because personal plans share chats for model training by default and business workspaces do not.
I'm a freelancer who has just finished a project. Below are the proposal, my time log and the extra requests the client made during the project (names removed).
1. List the recurring or likely ongoing needs this client has, each with evidence from the text.
2. For each, estimate monthly hours as a range, showing the sum, based only on my logged hours.
3. Flag anything that looks like one-off work rather than ongoing.
4. Suggest what I should ask the client to confirm.
Do not invent needs the text does not support.
The constraint "based only on my logged hours" keeps the model from inventing numbers. If you did not log time, give it your own estimates and say so.

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 nowA removals firm's website build, mined for monthly work
This is illustrative, with invented figures. Say you built a five-page website for a removals firm over six weeks, billing a project fee. Your log shows 52 hours. In the thread, the owner made these extra requests: change three prices on the quote page (twice), add a "price list" PDF, fix a booking form that broke after a plugin update, add a new service page for storage, and ask whether you could write two short moving guides a month for the site.
What ChatGPT returns
A useful analysis (illustrative) groups the history into four recurring needs:
- Content updates (prices, PDFs): three requests in six weeks at about 20 minutes of work each, but every one needed an email exchange and a check afterwards, so allow 1 to 2 hours a month.
- Maintenance (plugin updates, backups, a broken form fix): 1.5 to 2.5 hours a month, with a spike when something breaks.
- New pages (storage service): occasional, about 2 hours per page, perhaps one every two months, so about 1 hour a month on average.
- Content writing: two moving guides a month at about 2 hours each, which is 4 hours.
Total range: 7.5 to 9.5 hours a month if all four are included (1 + 1.5 + 1 + 4 at the low end, 2 + 2.5 + 1 + 4 at the high end). You might decide the writing is a separate, optional add-on, because it is a different skill and a different pricing logic.
Step 2: design the scope so it can be said in three lines
A scope that needs a page to explain will be argued over. Aim for three lines the client could repeat to their partner. For the removals firm (illustrative):
Each month you get: up to 6 hours of website work (updates, fixes and one new page when needed); monitoring and updates of the site's software; a short monthly note on what was done. Requests are made by email; I reply within one working day. Urgent faults (the site or booking form down) are looked at the same day.
Ask ChatGPT to tighten your draft: "Rewrite this scope in three lines of plain English. List anything a client could reasonably assume is included but is not stated, so I can decide." The second sentence catches gaps: does "updates" include photography, does "fixes" cover problems caused by the client's own edits, does the hosting bill sit with you or them?
Step 3: set the hours and the response times
Hours are the heart of a retainer, and they should include a buffer. A simple method: take your estimated monthly hours, add 20 per cent for admin and thinking time, and round up. Seven and a half hours becomes nine; you offer a block of six to nine hours depending on price. Underestimating is the usual trap, because small tasks cost more than their working time once you add the context-switching and the emails.
Response times matter as much as hours. Clients value being answered more than being served immediately. Offer something you can keep, in plain terms: reply to requests within one working day; work on non-urgent tasks within five working days; same-day look at faults reported before noon. Never promise faster than you can deliver on a bad week. If you take holidays, write when: "I'm away 20 to 30 December; the urgent-fault arrangement still applies."
Quick sum: three tiers
Ask ChatGPT to do the arithmetic, then check it yourself. Say your hourly rate is $70 and you want a retainer discount of roughly 10 per cent for committed work. The prompt: "Hourly rate $70. Create three monthly tiers: 3, 6 and 10 hours. Apply a 10 per cent committed-work discount to each. Show the monthly price, the effective hourly rate, and the annual value." Output (illustrative):
| Tier | Hours | Standard value | Monthly price | Per hour | Annual |
|---|---|---|---|---|---|
| Light | 3 | $210 | $189 | $63 | $2,268 |
| Standard | 6 | $420 | $378 | $63 | $4,536 |
| Extended | 10 | $700 | $630 | $63 | $7,560 |
The arithmetic is simple, so verify it yourself in a spreadsheet. The real decision is whether $63 an hour is acceptable to you after counting admin. Some freelancers price a retainer at or above their normal rate instead, on the grounds that guaranteed availability is worth paying for; either approach is a legitimate choice, and it is yours. For a broader view on how clients weigh freelancers against automation, see this comparison on cost.
Step 4: the terms that prevent arguments
Five short clauses stop most disputes. Ask ChatGPT to turn your decisions into plain-English wording, then have a lawyer or your usual template check it before it becomes a contract, because I am not offering legal advice and rules vary.
- Start date and minimum term: for example three months, then rolling monthly.
- Notice: 30 days either way.
- Unused hours: lapse at the month end, or one month's carry-over cap.
- Extra work: quoted at your standard rate before starting.
- Payment: invoiced at the start of each month, due within 14 days.
Paying at the start of the month is worth asking for. It reduces chasing, and the invoicing wording can be drafted along the lines in this tutorial on invoices, proposals and chasers.
Step 5: the email to send after delivery
Timing and tone decide whether the email works. Send it within two or three days of the handover, once the client has had time to look at the work but before the project fades from mind. It should not read like a sales pitch. It should read like a helpful next step based on what you noticed.
Write a short email (under 160 words) to a client I have just delivered a website for. Mention one specific thing from the project (the booking form plugin that needed fixing twice). Suggest a monthly arrangement covering updates, fixes and one new page when needed, for 6 hours a month at $378. Say how requests work and the reply time. Offer a 15-minute call, not a contract. Friendly, no pressure, no jargon. Sign as [your name].
A realistic draft (illustrative) reads: "Hello [first name], the site is live and the booking form has been stable since Friday. While building it, we handled a few small updates that kept coming up: price changes, a plugin fix and the new storage page. If it helps, I can cover that kind of work each month: up to 6 hours for $378, with replies within one working day and same-day attention if the site or booking form goes down. No contract needed to talk it through; shall we have a quick call on Thursday? [your name]."
What to edit: delete any invented specifics, check the price and hours match your offer, and cut a line if it sounds like a brochure. Remember that the concrete detail (the booking form) is the persuasive part, and it has to come from your real history.
Further examples: three other freelancers
A freelance copywriter
The project was a set of landing pages for a yoga studio. Recurring needs from the history: seasonal offer emails, one new class page per term, and tweaks after each timetable change. Scope: four emails and one page a month, plus up to 90 minutes of edits, replies within two working days. The sum is mostly writing time, so ask ChatGPT for the offer wording but write the sample in your own voice. Keeping your own voice when AI drafts for you covers that side.
A freelance bookkeeper
The project was a catch-up of a landscaping company's books. Recurring needs: monthly reconciliations, a payroll check and quarterly reports. Here the retainer is largely fixed work, so you might price per month at a flat figure rather than by hour, with a response time for client queries. ChatGPT helps list the monthly checklist from your catch-up notes, and you remove anything client-specific.
A freelance designer
The project was a brand pack for a tutoring centre. Recurring needs: flyers each term, social templates and occasional signage. Scope in design assets rather than hours: up to four items a month, with one revision round each. The rounds are managed best with the approach in this tutorial on feedback and revision plans.
Three ways to price a retainer, and which suits whom
Hours are not the only model. Ask ChatGPT to lay out the options for your kind of work, then choose. The three common structures:
| Model | How it works | Suits | Risk |
|---|---|---|---|
| Hours block | A set number of hours a month at a fixed fee | Mixed, unpredictable tasks such as website changes | Arguments over how time was counted |
| Fixed deliverables | A named list, such as four emails and one page | Content, design and reporting work | Clients asking for a fifth item |
| Availability | A fee for priority access and fast replies, with work billed on top at a reduced rate | Specialists clients call when something breaks | Clients feel they pay for nothing in quiet months |
The removals-firm site fits an hours block. The copywriter fits fixed deliverables. A freelance developer who rescues failing systems often fits availability. Describe your work to the model, ask which suits, and treat the answer as a prompt for your own thinking rather than a verdict.
Answering the client's objections
Expect four objections. Paste each in a message, with your offer, and ask for two draft replies, one short and one with more reasoning. Always check the reply does not promise more than the offer.
"It's too expensive for what we need"
Do not discount the same scope. Offer the lighter tier or a shorter list. A reply (illustrative): "That is fair. The 3-hour block at $189 covers the price changes and plugin updates, which were most of the small work this quarter. We can add a new page as a one-off quote when needed." The point is to reduce the scope, not your rate.
"We'll just ask you when we need something"
This is honest, and it is how you ended up quoting each job. The reply should name the benefit: planned work gets scheduled first, replies are guaranteed, and the price is lower per hour than ad-hoc quotes. If the client still declines, note a date to ask again after the next small job, and do not argue.
"Can we try one month?"
Say yes, with the three-month minimum replaced by a one-month trial at the standard rate (no discount), converting to the retainer price after month two. A one-month trial tests the relationship cheaply for both sides.
"Why not use AI for these tasks?"
Some clients will ask this. An honest answer (yours to give) explains where you use AI to be faster and where your judgement and responsibility matter. See this tutorial on telling clients about AI for wording options.
The monthly note that justifies the fee
Retainers get cancelled when clients cannot see what they are paying for. A short note at the end of each month fixes this and takes about ten minutes. Keep a running list of tasks with their time as you go, paste it (anonymised) at month end, and ask for a client-friendly summary with three headings: done, coming next, and suggestions. Illustrative output for the removals firm:
Done in October (5.5 of 6 hours): updated three prices and the price list PDF; applied two plugin updates and checked the booking form; added the storage page. Next: refresh the home-page photos once you send them. Suggestion: the contact page still shows an old opening time; shall I change it?
Note the suggestion line. It is where new paid work comes from, and it shows you are paying attention. Check that the hours match your log, because the client may compare them to their invoice.
Reviewing the retainer at month three
Put a review in the calendar when the retainer begins. At month three, compare the hours used with the block, the number of urgent requests and your own sense of the relationship. Ask ChatGPT to analyse your three monthly notes and the log: "Which task types took the most time? Which requests fell outside the scope? What changes to scope or price would you suggest, given these figures?" Treat the suggestions as input. If the client consistently needs 8 hours on a 6-hour block, offer the larger tier before you start working unpaid. If they need 3, a smaller one avoids a cancellation. A review conversation of ten minutes keeps the arrangement honest on both sides.
Where retainer offers go wrong
- The scope is vague. "General support" invites unlimited requests. Replace with named task types and a monthly limit.
- The price ignores admin. Hours billed rarely cover emails, planning and monthly notes. Add the 20 per cent buffer.
- The offer is sent too late. Three months after delivery, the client has found someone else for small jobs.
- The model invents a need. If ChatGPT lists "SEO monitoring" and nothing in the history mentions it, remove it. The client will ask what it is.
- No exit. A 12-month term with no review leaves you resenting the client. Start with three months.
Pitch, call and sign-up numbers for the first quarter
Track three numbers over the first quarter: the share of delivered projects that you pitch (aim to pitch every one that has a recurring need), the share of pitches that lead to a call, and the share of calls that become retainers. Small samples mislead, so judge by the first ten pitches. Also track your actual hours against the block each month. If you consistently exceed it by more than 20 per cent, raise the price or narrow the scope at the next review; if you use under 70 per cent, the client may feel they are overpaying and cancel, so propose a smaller tier before they ask. For a view from the buyer's side of ongoing arrangements, this tutorial on AI consulting retainers shows what clients ask before committing.
Whatever ChatGPT drafts, you read every line, set the price and sign the offer. It speeds up the thinking and the writing; the judgement about who deserves your monthly hours stays with you.
Retainer questions freelancers ask
When is the best moment to pitch a retainer?
Right after delivery, while the client is pleased and the work is fresh, and again at the first moment something small comes up that you would have to quote separately. Pitching before delivery feels like upselling; pitching months later feels like a cold email.
Should the retainer hours roll over?
Default to no rollover, or a cap of one month's unused hours. Unlimited rollover builds a liability you cannot plan around. Say clearly in the offer what happens to unused hours so nobody argues about it later.
Can ChatGPT choose my retainer price?
No. It can do the sums from your hourly rate, hours and buffer, and compare options you give it, but it has no knowledge of your market, your costs or your client's budget. Set the price yourself and use the model for arithmetic and wording.
What if the client says they only need work occasionally?
Offer a smaller block, such as three hours a month, or a pre-paid bundle with a longer expiry. If even that is too much, agree a light-touch check-in and a priority booking arrangement instead.
Further reads
- How to Hire an AI Automation Freelancer on Upwork or Fiverr — How buyers hire freelancers on marketplaces.
- AI Consultant vs Agency vs Freelancer: Which Should You Hire? — How clients compare you with an agency.
- AI for Freelance Designers: Faster Concepts Without Rights Issues — Rights questions that matter once work is ongoing.
- Consulting Pricing Models After AI: Fixed Fee, Retainer or Value? — Which pricing model survives AI-cut hours? Worked fixed-fee and retainer sums, a value-pricing test, scope limits and wording for telling clients.
- Freelance Day Rate Calculator: Set Your Rate When AI Speeds Up Delivery — A step-by-step freelance day rate calculation in USD, with billable days, overheads, AI tool costs and four ways to price when AI speeds up delivery.
- Website Maintenance Plan: Packages a Web Agency Can Add With AI — Three maintenance tiers with tasks and price bands, tool costs per site, and the minutes AI saves on reports and update checks.
- 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.
- How Can AI Help Freelancers? Proposals, Invoicing and Late Payers — A three-stage workflow for freelancers: proposal from call notes, clean invoices, and a day-by-day chaser ladder, with the messages filled in.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ChatGPT pricing page, OpenAI business data privacy page and OpenAI help-centre data controls article (checked 4 November 2026).