
An AI-ready project brief: the template that stops endless prompting
An AI project brief connects outcomes, scope, sources, constraints, and validation. Use this template before assigning work to an agent.
A worked example and a reusable prompt for turning client emails into tasks, keeping source links, spotting missing details and checking deadlines.
Last reviewed on August 30, 2026

One client email can contain two tasks, a restriction and a deadline that still needs agreement. AI can help sort them out — if you ask for more than a summary.
Gather
The email and the right attachments.
Separate
Actions, constraints and open questions.
Review
Agree on scope before dates and owners.
Here is a fictional example: an agency is preparing an event registration page. The client writes:
Could you replace the programme with the attached PDF and add an optional Company field to the form? We’d like to see the updated page on Thursday.
Keep the current price. Please wait for my approval before changing the partner logo: the attached version is a draft.
Let me know whether that timing works.
Creating a task called “Update the page” would lose the important details. Optional changes how the form works. The logo must not be published. And Thursday is a request, not a promise.
Ask AI to propose one task per verifiable outcome. This example needs two actions. The other information should stay visible without becoming work to start immediately.
To do · task 1
Replace the programme
The page link must open the supplied PDF.
Check: download the file through the preview page and compare it with the attachment.
To do · task 2
Add the Company field
The field appears and remains optional.
Check: submit the form with a company name, then without one, in the test environment.
Keep unchanged
The current price and logo
No price change. The draft logo is a reference, not an asset ready to publish.
To confirm
A Thursday preview
Check the exact date and team availability before committing. Nobody has been assigned the work yet.
Keep the email link and PDF with the tasks: they remain the source of truth if the request changes.
Three checks prevent a lot of rework:
If an attachment is missing, AI should flag it. It should not invent the contents or fill gaps with a plausible date or name.
Selecting relevant information is central to Anthropic’s work on agent context. For sensitive documents, also decide whether to process them locally or in the cloud.
This prepares a proposal while keeping the decision with you.
From email to tasks
Read the email and supplied documents. Propose one task per expected outcome.
For each task: state the action, source passage and a way to check the result.
Separate actions, things to preserve and decisions awaiting confirmation. Do not invent dates, owners or file contents.
Flag missing attachments and duplicates against existing tasks, if I have supplied them.
Do not create anything or send messages before my approval.
Then review the email sentence by sentence. Is every request covered? Have the restrictions survived? Has the AI added anything the client never asked for?
After approval, create the cards and their checklists, then link useful documents from the knowledge base. Check for existing tasks first to avoid duplicates.
You can start with a manually copied email. This method does not require a connected inbox or permission for the agent to send messages.
Keep the client reply specific: “We’re preparing the two changes and will confirm the preview slot after checking the schedule.” That acknowledges the request without committing to an unconfirmed deadline.
Next: how to check the agent’s work when it says “done”.
No. Start with a copied message and the relevant attachments in a tool your organisation permits.
The team. AI can carry over an agreed decision or make a suggestion, but it must distinguish between them.
Keep the request and flag the missing file. Any conclusions that depend on its contents should wait until it arrives.

An AI project brief connects outcomes, scope, sources, constraints, and validation. Use this template before assigning work to an agent.

Vibe coding accelerates prototyping. Use this practical method to secure architecture, data, testing, and deployment before going to production.

Building a video game with AI takes more than code. Follow a practical method to organize mechanics, assets, tests, decisions, and agents in one project.

AI backlog grooming keeps cards fresh by detecting duplicates, stale work, weak descriptions, missing context, and risk before planning starts.
Stellary brings together your board, docs, and AI agents in one command center.