Using AI to turn a string of customer WhatsApp or text…

A typical client project thread on WhatsApp runs 30–100 messages, contains at least one contradiction, and buries the actual deadline somewhere around message 47. If you've ever needed to organize client instructions from WhatsApp messages into something you can actually work from, this post walks you through a repeatable five-minute workflow: export the thread, paste it into an AI tool with a structured prompt, and get a clean brief. Done consistently, this single habit reduces scope disputes, eliminates re-reading old threads, and gives you a written record both you and your client can refer back to.

What you need before you start

ChatGPT{:target="_blank"} — a general-purpose AI that accepts long pasted text and returns structured output. The free tier (GPT-4o access included as of early 2025) covers this workflow entirely. If you're processing multiple briefs daily or want to save custom prompt templates, ChatGPT Plus{:target="_blank"} runs $20/month as of March 2026 — not required, but useful for the custom instructions feature covered later. Claude{:target="_blank"} (Anthropic) and Gemini{:target="_blank"} (Google) work equally well here; I'll note where they differ.

Time required: 2 minutes to export your first chat thread. Under 5 minutes total to run the full workflow once you have a saved prompt template.

Skill level: No technical background needed. You need a WhatsApp account, an AI tool open in a browser or app, and the ability to copy and paste. No integrations, no code.

How to export the thread in under two minutes

Open the WhatsApp conversation you want to export.
Tap the three-dot menu (Android) or the contact name at the top (iOS) to open chat settings.
Select More > Export Chat > Without Media. WhatsApp will generate a plain .txt file — this is a built-in feature{:target="_blank"}, no third-party app needed.
Send the file to yourself via email or save it to your phone's files app, then open it on the device where you have your AI tool running.
Select all and copy the text from the file. You're ready to paste.

For SMS or iMessage on iPhone, use the Print to PDF workaround: open the thread, tap the contact name, then use the share sheet to print — in the print preview, pinch outward on the page to save it as a PDF. Steps vary slightly by iOS version. Alternatively, iMazing{:target="_blank"} ($34.99 one-time license as of March 2026) exports iMessage threads cleanly as plain text. Android users can use SMS Backup & Restore{:target="_blank"} (free) for the same result.

GPT-4o and Claude 3.7 Sonnet both handle 100,000+ token context windows, which means even a very long export — several months of messages — can be pasted in a single go without truncation.

The exact prompt to organize client instructions from WhatsApp messages into a brief

Open your AI tool, start a new conversation, paste the entire chat transcript, then add this prompt directly after it:

Prompt:

The text above is an exported WhatsApp conversation between me and a client. I need you to turn it into a structured project brief I can use to start work.

Return the brief under exactly these headings:

Goal — What the client wants delivered, in one or two sentences. Deadline — The date or timeframe mentioned. If multiple dates appear, list them all. Budget — Any figures mentioned, or note "Not stated" if absent. Requirements — Specific instructions, constraints, preferences, or non-negotiables the client mentioned. Open Questions — Anything that was raised but not resolved, or any detail I'll need before starting work. Contradictions — Flag any places where the client said conflicting things (e.g., different deadlines, changed requirements). Do not resolve these silently — list them so I can clarify. Client Contact Preference — How and when the client prefers to be reached, if mentioned.

Write in plain, factual language. Do not add information that isn't in the transcript. If a section has no relevant content, write "Not mentioned."

What you should get back: a 200–400 word brief with all seven sections populated, drawn only from the transcript. Verify it by spot-checking two or three specific details against the original thread — the deadline, one specific requirement, and any contradiction flagged. If the AI has invented a detail that doesn't appear in the transcript, that's a hallucination; regenerate or edit manually.

The structured heading list in the prompt is doing most of the work here. An open-ended "summarise this chat" request produces narrative text that's hard to act on. Specifying the output format forces the model to categorise information rather than paraphrase it — and the Contradictions heading is the most important one, because without it the model will pick whichever instruction appeared most recently and silently discard the earlier one.

When something goes wrong

The brief is vague or misses key details. Root cause: the transcript contains a lot of off-topic chat (greetings, weather, personal updates) that dilutes the signal. Fix: add one line to your prompt — "Ignore small talk and messages unrelated to the project. Focus only on instructions, decisions, and requests." This alone tightens output significantly on social-heavy threads.

A contradiction isn't flagged — the brief just picks one version. Root cause: the Contradictions heading was omitted or the model treated the later message as an override. Fix: explicitly add the instruction "Do not resolve contradictions. List both conflicting statements with the date/time of each." WhatsApp exports include timestamps on every message, which gives the model the sequence it needs to report this accurately.

The export file is garbled or the AI says the text is too long. Root cause: very old Android exports sometimes include unusual encoding, and a small number of AI tools still have context limits below 50,000 tokens. Fix: if you're using a browser-based tool and hitting a limit, switch to Claude.ai or ChatGPT — both handle long transcripts reliably as of early 2026. If the encoding looks broken, open the .txt file in a plain text editor, select all, and re-copy before pasting.

Saving your prompt as a reusable template

Once you've confirmed the prompt works for your typical client type, save it somewhere you can paste it in ten seconds. Three options:

ChatGPT Custom Instructions (available on free and Plus): go to your profile > Custom Instructions and paste the prompt into the "How would you like ChatGPT to respond?" field. It will pre-load on every new conversation.
Claude Projects{:target="_blank"}: create a project called "Client Briefs," paste the prompt as the project instructions, and run every new transcript inside that project.
A notes app: a pinned note in Apple Notes, Notion, or even your phone's default notes app works fine. Copy, paste, done.

For businesses with recurring project types — a builder who always needs the same brief structure, a designer who always needs brand guidelines flagged — you can create variations of the prompt. A landscaper's brief looks different from a web designer's brief. The headings are yours to define; the core logic of the prompt stays the same.

A quick note on privacy before you paste

Pasting a client's messages into a consumer AI tool means that data leaves your device and passes through the tool provider's servers. For most trades and service businesses, this is a reasonable trade-off. For businesses handling sensitive client information — legal matters, medical history, financial records — it warrants more care.

OpenAI's data controls{:target="_blank"} allow you to turn off chat history, which means your inputs are not used to train the model. Claude and Gemini have equivalent opt-outs in their settings. Turning this off takes about 30 seconds and is worth doing regardless of what you're pasting.

If your business is in a regulated sector, the right move is an enterprise tier or API access — both of which include stronger data handling agreements. ChatGPT Enterprise and Claude for Enterprise both offer data processing agreements (DPAs) for this purpose. Those plans start at pricing that requires a direct quote from the vendor, so I won't estimate figures here.

What to do next

Make this the last step of every client intake: export the thread, run the prompt, save the brief, send it to the client with a one-line message asking them to confirm it's accurate. That confirmation is your written scope agreement — and it takes less than five minutes to produce.

If you want to take this further and connect the brief output directly to your project management tool, read how to use AI to move information between apps without code.

FAQ

Can I use this workflow with Google Messages or iMessage instead of WhatsApp? Yes, with an extra export step. iPhone users can use iMazing ($34.99 one-time as of March 2026) to export iMessage threads as plain text. For Google Messages on Android, SMS Backup & Restore (free on Google Play) creates an exportable file. Once you have the text, the prompt and AI workflow are identical — the model doesn't care which app the messages came from.

Does the free version of ChatGPT handle long transcripts? GPT-4o on the free tier handles 100,000+ tokens of context as of early 2026, which covers even very long WhatsApp exports without truncation. The free tier may limit how many GPT-4o messages you can send per hour before falling back to an older model — if you're processing multiple briefs in one session, you may hit that limit. ChatGPT Plus ($20/month) removes the cap. Claude's free tier on claude.ai offers similar context capacity with less aggressive rate limiting, making it a practical free alternative.

What's the actual time saving here, and is it worth building into a routine? McKinsey research on workplace communication{:target="_blank"} puts message reading and response at 28% of the average workday. For a small business owner managing three to five active client threads, re-reading old WhatsApp conversations to recall decisions is a repeatable drain — conservatively 15–20 minutes per project per week. At five projects running concurrently, that's over an hour a week of re-reading that a saved brief eliminates. The five-minute setup cost pays back on the first use.

Is an AI-generated brief legally binding? No, and you shouldn't treat it as a contract. What it is: a clear written summary of what your client communicated, which you've asked them to confirm. That confirmation — a "yes looks right" reply — creates a reference point for scope disputes that a WhatsApp thread alone does not. For formal contracts, use a proper agreement. The brief is operational documentation, not legal documentation.

What if my client uses voice notes instead of text messages? WhatsApp voice notes are not included in a text export. If your client communicates primarily through voice, you'll need a transcription step first. Whisper{:target="_blank"} (OpenAI's open-source transcription model) is free to run via several tools. Otter.ai{:target="_blank"} offers a free tier covering 300 minutes of transcription per month — enough for most client threads. Once transcribed, paste the transcript into the same brief-generation prompt; it works the same way.

Using AI to turn a string of customer WhatsApp or text messages into a clean project brief you can actually work from