Three assets make a launch kit: a positioning statement, a launch email and a sales one-pager, all written from one product brief. ChatGPT, Claude or Gemini draft them, Gemini Notebook can check them against customer evidence, and Canva or your email platform handles layout and sending. The step that protects you is checking every claim against real evidence.
Tool lists age quickly, but the kit does not. The example below is a four-person interior design studio launching a fixed-fee package, and the same sequence suits an HR consultancy, a mortgage broker or a financial adviser with a productised service. Each section gives the prompt, a sample result and the check that stops an invented claim reaching a customer.
One product brief that every asset is written from
Assistants write generic copy from vague input, so the brief is your control. Fill it in before opening any tool, and keep it to one page. Here is the studio's brief for a package called Room Refresh (illustrative):
| Field | Entry |
|---|---|
| Product | Room Refresh: a fixed-fee redesign of one room, scheduled over three weeks |
| Who it is for | Homeowners with one room that has never worked, not whole-house renovators |
| Problem, from customer interviews | They have lived with the room for years, do not know where to start, and fear the cost will keep growing |
| Alternatives they use today | Doing it themselves from online inspiration; hiring a decorator by the hour; buying furniture and hoping |
| What is included | Layout, furniture plan, shopping list with links, one visit, one revision round |
| What is not included | Structural work, buying on the client's behalf, changes after the revision round without a quote |
| Price and terms | $3,500 fixed for the agreed scope; changes quoted before work starts |
| Proof we hold | Nine completed projects, five written reviews, three before-and-after galleries |
| Proof we lack | Permission to name clients; any measure of time saved |
| Launch and action | Email to past clients and enquirers; book a 20-minute call |
Two rows do more work than the rest. "Alternatives they use today" gives the assistant something real to position against, instead of the vague "other designers". "Proof we lack" tells it, and you, where claims will be thin. Keep the brief in a Project (ChatGPT or Claude), or a Gemini Gem that becomes a skill from November 2026, so every later prompt starts from it.
The evidence file: what customers have said and done
The brief says what you offer. The evidence file records what you can prove, and it is the document every claim will be checked against. Gather it from your own records: a project ledger, written reviews, call notes, and the emails where clients explained why they hesitated. Give each item an ID so a claim can point to it. The studio's file looks like this (illustrative):
| ID | Evidence | Source | Permission to quote |
|---|---|---|---|
| E1 | Nine completed projects; seven finished within the original budget, two went over after the client asked for changes | Project ledger | Not needed |
| E2 | Five written reviews from past clients | Review emails | Two of five confirmed in writing |
| E3 | Four of nine past clients said fear of runaway costs made them hesitate | Call notes | Not quoted, so not needed |
| E4 | Before-and-after photos from three projects | Project folders | Three of three |
Strip names, addresses and any personal detail before uploading anything, and use a plan whose model-training switch is off or a business plan that does not train on your content by default. Gemini Notebook (formerly NotebookLM) suits this job because it answers from the sources you upload and cites them. Google's help page sets a limit of 500,000 words or 200MB per local upload, and Google's plan limits allow 50 sources per notebook on Free and 300 on Pro. In a Claude or ChatGPT Project, upload the same file next to the brief; a free Claude account can create up to five projects.
Positioning in three drafts, then a five-question test
A positioning statement has a fixed frame: for a certain customer with a certain problem, the product is a category of thing that delivers a benefit, and unlike the alternative, it does something different. Ask the assistant for three versions at once, so you have something to compare.
Using the attached brief and evidence file, write three positioning statements in this frame: For [target customer] who [problem], [product] is a [category] that [main benefit]. Unlike [the alternative customers actually use], it [difference].
Rules: use only facts from the brief or evidence file. Name the alternative from the brief. Do not use the words transform, stunning, premium or elevate. After each statement list the evidence IDs it relies on, or write NONE.A first result for the studio might look like this (illustrative):
| Draft | Statement | Evidence | Verdict |
|---|---|---|---|
| A | For busy homeowners who want a beautiful home, Room Refresh is a premium design service that transforms your space. Unlike other designers, we deliver stunning results. | NONE | Fails: any studio could say it |
| B | For homeowners with one room that has never worked, Room Refresh is a fixed-fee redesign of that room in three weeks. Unlike hourly decorators, the price is agreed before work starts and changes are quoted first. | E1, E3 | Keep as the master statement |
| C | Room Refresh redesigns the one room that never worked, in three weeks, for a fixed fee agreed up front. | E3 | Use as the headline |
Draft A is what an assistant writes when the brief is thin, and the rules in the prompt exist to stop it. Run each draft you like through five questions, scoring one point per yes:
- Could a competitor say it word for word? If so, score zero for this question and rewrite.
- Does it name the alternative a customer would otherwise choose?
- Does every phrase trace to the brief or an evidence ID?
- Would the customer recognise their own problem in the first clause?
- Can you say it in one breath to a stranger?
Keep any draft that scores four or more. Draft B scores five, Draft C four, because its alternative is implied rather than named. If your first three drafts disappoint, give a second assistant the same brief and compare; the better one is the one that names the alternative and stays inside the evidence. Public reviews of the alternative are a good place to find its weak points, as the tutorial on mining competitor reviews for positioning shows.
The launch email: one idea, one action, evidence tags in the margin
Write a launch email to past clients and enquirers for Room Refresh, from the brief and positioning Draft B.
Format: three subject lines under 50 characters, one preview text, a body under 150 words, one call to action (book a 20-minute call).
Rules: plain and warm, no exclamation marks. Every factual claim ends with an evidence ID in square brackets so I can check it. Include no customer quote unless it is pasted verbatim from the evidence file.The result, with the evidence tags still in place (illustrative):
Subject options: "One room, three weeks, one price" (32 characters) / "The room that never worked" (26) / "Room Refresh: a fixed fee for one room" (38)
Preview: Redesign the problem room with the cost agreed before we start.Hi [first name],
If one room in your home has never quite worked, Room Refresh is how we fix it: a redesign of that one room, scheduled over three weeks, for a fixed fee of $3,500.
You get a layout, a furniture plan, a shopping list with links, one visit and one revision round. Changes after that are quoted before we make them.
Seven of our nine past projects finished within their original budget, and the two that did not changed because the client asked for more [E1].
Book a 20-minute call and we'll tell you whether Room Refresh fits: [link]
[your name]
This is the second draft. The first said "Guaranteed on-budget results, every time", which the evidence file contradicts, since E1 shows two projects over budget. The tag forced the fix. The honest sentence is also the better one, because it admits what changed the two budgets and gives readers a reason to trust the other seven. Strip the bracketed tags before sending, and keep the tagged copy with the launch records.
On sending: Mailchimp's pricing page shows a Free plan with one audience of up to 250 contacts and 500 sends a month, Essentials from $13 a month and Standard from $20 a month, with generative AI features listed for Standard and above but not for Free or Essentials. For a small list, write in your assistant and paste into the platform. Gmail's sender rules also matter if your list includes Gmail addresses: keep your spam rate under 0.10%, offer one-click unsubscribe once you send over 5,000 emails a day, and keep the message light, because Gmail clips emails over 102KB. Judge the launch by replies and booked calls rather than opens. A follow-up sequence, from teaser to last call, is in the tutorial on launch email sequences with ChatGPT.
The one-pager: every block has a job
A sales one-pager is read in under a minute, often on a phone after someone forwards it. That argues for blocks with a fixed job and a word limit, not a design-first page. Here is the studio's plan, with the evidence each block draws on:
| Block | What it holds | Words | Evidence |
|---|---|---|---|
| Headline | Positioning Draft C | 15 at most | E3 |
| The problem | The customer's own situation, from the brief | 25 | E3 |
| What you get | Four bullets from "what is included" | 40 | Brief |
| How it works | Three steps, one per week | 45 | Brief (a promise) |
| Proof | The budget record, plus one review quote if permission exists | 40 | E1, E2 |
| Price and terms | $3,500 fixed; what is excluded | 30 | Brief |
| Next step | Booking link and contact | 15 | None |
That totals about 210 words, which fits a single page with room for a photo. Ask the assistant for the copy block by block with the word limits attached, and say which evidence it may use in each. Use before-and-after photos only from the projects in E4 where permission is recorded; a photo is a claim about what you did, and E4 is the proof that the client agreed.
For layout, Canva's pricing page lists a Free plan with one Brand Kit limited to three colours, Pro at $144 a year per person with five Brand Kits, and Business at $250 a year per person with 100. Its AI features draw on a monthly allowance: up to 20 Standard or Premium uses on Free, up to 200 Premium or 20 Ultra on Pro, and up to 400 Premium or 40 Ultra on Business. Standard AI tools on paid plans do not count against the allowance. A one-pager needs very few AI uses, so the Free plan is enough to begin. If you would rather skip a design tool, paste the copy into a Google Doc or Word template and export a PDF.
Whichever route you take, do four checks before the PDF leaves the building. Open it on a phone, because most forwarded one-pagers are read there. Scan any QR code with a second device. Check that the price matches the email word for word. And confirm the file name does not carry an internal draft label.
Competitor-facing material is a different document. If sales reps want comparisons, keep named competitors in an internal sheet, as in the tutorial on B2B battle cards and one-pagers with ChatGPT, and keep the customer-facing page about the problem and the alternative's category.
The claims check that runs before anything is published
AI drafts are fluent, and fluent copy hides unsupported claims. The check takes ten minutes and happens once, across all three assets, after the drafts are final:
Here are three drafts and the evidence file. List every factual claim, number, superlative, comparison and customer quote in the drafts. For each, give the evidence ID that supports it, or NONE, and say whether it is a past fact, a promise about the future, or an opinion. Do not fix anything yet.Then decide for each row whether to keep it, narrow it to what the evidence shows, or delete it. These are the studio's rows (illustrative):
| Draft claim | Evidence | Verdict | Rewrite |
|---|---|---|---|
| "No surprises, ever" | E1: two of nine projects went over | Reject | "Fixed fee for the agreed scope; changes are quoted before work starts" |
| "Clients love the result" | E2: five reviews, two with permission | Narrow | "Five past clients wrote reviews" plus the two permitted quotes |
| "The fastest redesign available" | None | Delete | Nothing |
| "Finished in three weeks" | A promise, not a past result | Keep if capacity allows | "Scheduled over three weeks from the first visit" |
| "We could finally use the room," says a past client | Written by the assistant | Delete | Use a real quote from E2 with permission |
Two distinctions do most of the work. A promise about the future ("scheduled over three weeks") is a commitment you control, so it needs capacity rather than evidence, but never state it as a past result unless the ledger shows it. A superlative ("fastest", "best", "leading") needs a comparison you can prove, and for a small company it almost never exists, so delete it.
The invented quote deserves its own warning. Ask an assistant for "social proof" and it will happily write a testimonial from a customer who does not exist, which is fabricated evidence. The prompt rule against quotes is not a style preference. For where the legal line sits, see the tutorial on AI-written testimonials, and for a method of sourcing every line of marketing copy, the tutorial on backing up every claim AI writes.
When the product is brand new and the evidence file is thin
The studio had nine past projects to draw on. A product with no customers yet has an evidence file that is nearly empty, and the temptation is to let the assistant fill the gap. Do not. A new product can still make three kinds of statement honestly: what it includes, what you commit to, and what customers said in interviews before launch.
So the copy leans on design decisions and policies. "Fixed fee for the agreed scope" is a policy. "Changes are quoted before work starts" is a policy. "Scheduled over three weeks" is a commitment. None needs a track record, only the discipline to keep it. What a new product cannot say is anything about outcomes, such as "saves time", "clients stay within budget" or "rated highly".
Build the first evidence by running five early customers through the product at a reduced fee or free, and say so ("founding clients"). Interview each at the end with the same three questions, record the answers, and ask for permission to quote on the spot. Five interviews will not support a statistic, but they will give you real language, real objections and perhaps two quotes. Never borrow evidence from a different product or an earlier business and present it as proof for this one.
Before you have customers, put an assistant to better use on the brief than on the proof. Ask it to read the brief as a sceptical buyer and list the five questions that buyer would ask before booking a call. The answers become the FAQ line on the one-pager and the first lines of your call script. If you are still deciding whether the product is worth launching, test the idea with real customers before you write any of this.
Keeping one voice across the three assets
Three assets written in three chats drift apart. Fix that with a short voice note kept beside the brief in the Project: three sentences describing how you speak, two sentences you have actually written to customers, and a list of words you never use. The studio's note says: plain and warm; write to one person; prefer "you" to "clients"; never say stunning, bespoke or curated. Paste your own emails, not marketing copy, as the samples, because the assistant imitates what it sees.
Paid brand-voice tools exist. Jasper's pricing lists Pro at $69 a month for one seat ($59 on annual billing), with two Brand Voices and a seven-day trial. Writer's starter plan covers up to five users, but its price is not printed on its own site, so confirm it at sign-up. They earn their fee when several people write for the company and need one enforced voice. For one or two writers, a Project instruction or a Gemini Gem does the same job; ChatGPT's custom instructions cap at 5,000 characters, so keep the voice note short. The comparison is laid out in the tutorial on brand voice tools.
Which tool for which step, and what each costs
The kit needs far fewer tools than most roundups suggest. Prices are USD list prices as published in autumn 2026, and vendors change them, so confirm on the vendor's page before you buy.
| Step | Options | List price | Pay when |
|---|---|---|---|
| Brief, positioning, email and one-pager copy | ChatGPT, Claude or Gemini | ChatGPT Plus $20 a month; Claude Pro $20 ($17 billed annually); Google AI Pro $19.99 | You hit a usage, file or project limit |
| Evidence file and claim extraction | Gemini Notebook, or a Claude or ChatGPT Project holding the file | Notebook allows 50 sources per notebook on Free and 300 on Pro | The file outgrows 50 sources |
| One-pager layout | Canva, or Google Docs or Word | Canva Free $0; Pro $144 a year per person; Business $250 a year per person | You need several Brand Kits, or the AI allowance runs out |
| Sending the email | Mailchimp or your current platform | Mailchimp Free (250 contacts, 500 sends a month); Essentials from $13; Standard from $20 with generative AI | Your list passes 250 contacts or you need automations |
| Brand voice | Project instructions; Jasper or Writer | Jasper Pro $69 a month a seat ($59 annual); Writer: confirm at sign-up | Several people write for the company |
| Claims register | A spreadsheet | $0 | Never |
A solo owner can run the whole kit for about $20 a month, using one assistant plan with the free Canva and Mailchimp tiers. A two-person team that adds a second assistant plan, two Canva Pro seats ($288 a year, so $24 a month) and Mailchimp Standard at $20 lands near $84 a month. Either figure is small next to the cost of an email that makes a claim you cannot support.
A launch fortnight on a page
| Day | Task | Where AI helps | Human sign-off |
|---|---|---|---|
| -14 | Brief and evidence file | Turns your notes into the brief; summarises interviews | Owner |
| -12 | Positioning drafts and the five-question test | Three drafts from the brief | Owner |
| -10 | Launch email and one-pager copy | Drafts with evidence tags | Owner |
| -7 | Claims check; permission requests for quotes and photos | Extracts every claim into a table | A second person |
| -5 | Layout and email build; test sends | Layout suggestions | Owner |
| -3 | Read-through on a phone by someone outside the project | None | Second person |
| -1 | Schedule the send; test the booking link and QR code | None | Owner |
| 0 | Send | None | Owner |
| +3 | Follow-up to people who have not replied | Drafts the follow-up | Owner |
| +7 | Review replies, calls booked and questions asked; update the evidence file | Groups replies by theme | Owner |
Notice that the human sign-off column never says "AI". The assistant drafts and sorts; a named person decides, and for anything that makes a claim about results, a second person reads it fresh. The timeline also puts the claims check a week before the send, which leaves time to ask a past client for permission or to find a missing number.
The same kit for three other firms
The sequence does not change with the business, but the evidence and the sign-off do. Here are three productised services with a first-draft positioning line each (all illustrative):
| Firm and product | First-draft positioning line | Evidence to collect first | Extra sign-off |
|---|---|---|---|
| HR consultancy: Retention Health Check | For companies with 50 to 200 staff whose good people keep leaving, a fixed-price review of why, ending in a written plan. | Anonymised exit-interview themes; client permission to describe the work | An HR adviser reads any claim about outcomes |
| Mortgage broker: First-time buyer service | For first-time buyers who do not know where to start, a broker who explains each step and compares lenders for you before you apply. | Your own count of lenders on your panel; reviews with permission | Your compliance adviser reads it before sending |
| Financial adviser: Retirement readiness review | For people within ten years of retiring who have never had their plans checked, a one-off review with a written summary. | Your own client-feedback records; the exact scope of the review | Compliance review, and no performance or returns language in AI drafts |
Regulated businesses should treat the claims check as the start of a review, not the end of one. An assistant has no idea which phrases your adviser's rules treat as promotions, so keep it away from forward-looking statements and let a qualified person approve the final copy. Even a mortgage broker's harmless-sounding "we compare lenders for you" is a factual claim about your panel, and it needs the count to back it.
After the send: feed the evidence file and keep a register
Launch week produces the best evidence you will have: replies in customers' own words, the questions they asked, and the objections that stopped a booking. Within a week, paste the replies into the Project or notebook (with names stripped) and ask the assistant to group them by theme. A question asked five times means the one-pager is missing an answer. A phrase customers repeat belongs in the next positioning draft. Ask each customer who liked the offer for permission to quote them, and record that in the evidence file with the date.
Keep a claims register beside the evidence file so that a reused claim is checked again, not trusted from memory. One row per asset is enough:
| Asset and version | Date | Claims checked | Changes made | Approved by |
|---|---|---|---|---|
| Launch email v2 | Day -7 | 5 | "No surprises" narrowed to the fixed-fee policy | Second person |
| One-pager v3 | Day -7 | 6 | "Fastest" deleted; invented quote removed | Second person |
| Follow-up email v1 | Day +3 | 2 | None | Owner |
Rerun the check whenever a number changes, when a claim moves from the email to a website page or an advert, and when a competitor you described changes its offer. If a claim is more than a year old, check it again even if nothing has changed, because the evidence may have aged out. A launch kit written this way costs a few hours and an assistant plan. It also leaves you with a brief, an evidence file and a register that make the next launch faster than this one.
Questions small teams ask about an AI launch kit
Do I need a paid AI plan to build a launch kit?
No. A free plan can draft all three assets. A paid plan at about $20 a month, such as ChatGPT Plus, Claude Pro or Google AI Pro, mainly buys heavier use and higher file and project limits. Start free, keep the brief and evidence in one place, and upgrade when you hit a limit rather than before.
How do I ask a past client for permission to quote them?
Send the exact words, say where they will appear and ask for a written yes. For example: '[first name], may we quote this line from your review in our Room Refresh email and one-pager: [the line]? We would show your first name only.' Record the reply and the date against the evidence ID, and remove the quote if they later change their mind.
Should the launch copy name competitors?
Avoid it in public copy. Naming a competitor turns your claim into a comparison you must prove and keep current, and a legal adviser should review it. Describe the alternative by category instead, such as hourly decorators or doing it yourself. Keep named competitors in internal battle cards, where a wrong detail costs you less.
Does the same kit work for a physical product?
Yes, with different evidence. The brief adds specifications, materials, delivery times and returns terms, and the evidence file holds test results, returns rates and reviews. Claims about safety, durability or the environment need the strongest proof. If you sell to consumers in the EU, generic green claims such as 'eco-friendly' have been banned since 27 September 2026 unless you can show recognised excellent environmental performance.
Further reads
- How to Build a Brand Voice Guide That AI Can Follow — A brand voice guide your assistant can follow
- How to Write Case Studies From Customer Interviews Using AI — Turn customer interviews into evidence you can publish
- How to Test a New Service Idea With AI Before You Launch It — Test the idea with customers before you write the kit
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Canva pricing page; Mailchimp pricing page; Google Gemini Notebook help page; vendor pricing and help pages for ChatGPT, Claude and Google AI plans. Prices are USD list prices as published in autumn 2026 and change often. The studio, evidence and results in the examples are illustrative.