Forward This Intake to Your Agent

Why this page exists

The HowDoWe.AI intake email is designed to be forwarded to an agent. That is intentional.

Most intake forms create a lead notification. Useful, but passive. A better intake creates an implementation brief. It should help a human or agent answer:

1. Is this company a good fit?
2. What workflow should we connect first?
3. What systems are involved?
4. What is missing?
5. What can we prove in the first week?
6. What should Rob say next?

This page gives you a reusable prompt for turning that intake into action.

The copyable prompt

Paste this into Codex, Claude Code, ChatGPT, Cursor, Gemini CLI, or another implementation agent. Then paste the full intake email below it.

You are helping evaluate an AI implementation intake.

Read the forwarded email below. Do not start building yet.

Return:
1. The first workflow to connect.
2. The systems and tools involved.
3. The company knowledge that needs to be gathered.
4. The likely source of truth for each part of the workflow.
5. The human review owner and approval points.
6. The safest first proof of value.
7. Missing access, missing examples, or risk flags.
8. A 7 day implementation plan.
9. A short reply Rob could send to the prospect.

Rules:
- Keep it practical.
- Prefer one small working loop over a broad transformation plan.
- Do not invent integrations that are not mentioned.
- If the intake is too vague, ask for the minimum missing context.
- Separate "can build now" from "needs discovery first."

Forwarded intake:
<paste the email here>

When to use this prompt

Use it after someone submits an intake form, sends a long email, records a workflow walkthrough, or shares a screen recording. It is especially useful when the prospect already has tools and documentation but cannot explain the first AI implementation clearly.

Good signals:

1. They mention tools like Notion, Slack, Google Drive, ClickUp, Asana, HubSpot, QuickBooks, Fireflies, or Airtable.
2. They describe a repeated copy-paste workflow.
3. They have examples of inputs and outputs.
4. They know who reviews the work.
5. They have a timeline and budget.
6. They can name the first workflow they want connected.

Weak signals:

1. They only say "we need AI."
2. They have no documented workflows.
3. They cannot name a painful handoff.
4. They want full automation with no human review.
5. They expect AI to fix unclear ownership.

The prompt helps separate fit from fantasy.

The agent should produce a triage brief

The output should be short enough for a human to read and specific enough to become a next step.

Ask for this structure:

Fit:
High / Medium / Low

Why:
<short explanation>

First workflow:
<one workflow>

Systems involved:
<tools and knowledge sources>

Missing context:
<questions>

Risks:
<security, access, adoption, data quality, review gaps>

7 day proof:
<small implementation plan>

Draft reply:
<email Rob can edit>

That structure makes the agent useful without letting it wander.

How to include recordings

If the prospect records audio or a screen walkthrough, summarize it before asking for implementation. A transcript is even better.

Use this add-on prompt:

The intake includes a recording or transcript.

First summarize the workflow in plain English:
1. Trigger
2. Input
3. Tools opened
4. Decisions made by the human
5. Output
6. Final destination
7. Review or approval step

Then complete the implementation triage brief.

Recordings are powerful because they show the real workflow. People often forget to mention the browser tab, spreadsheet, Slack thread, or judgment call that makes automation hard. A short walkthrough reveals those details.

How to use it with Codex

Use Codex when the intake points toward a file-based implementation: website updates, scripts, internal tools, data normalization, documentation, or a product feature.

Prompt Codex like this:

Use the forwarded intake to create an implementation plan only. Inspect the repo if needed, but do not edit files. Return the likely files involved, the smallest safe first change, tests to run, and questions for Rob.

Then approve a second task only after the plan makes sense. Codex is excellent when the outcome is a pull request or reviewable diff.

How to use it with Claude Code

Use Claude Code when the work happens in a local project, docs folder, notes vault, or repo where a human wants to steer closely.

Prompt Claude Code like this:

Read this intake and map it to the current workspace. Do not edit. Identify files, notes, docs, or scripts that could support the first workflow. Propose a smallest reversible edit.

Claude Code is especially good when the prospect's context needs to be organized before implementation.

How to use it with ChatGPT

Use ChatGPT when the intake requires synthesis across business context, market language, workflow design, or internal knowledge. ChatGPT is the right place to turn a messy intake into a clean operating map.

Prompt ChatGPT like this:

Turn this intake into an AI Operating Map. Identify the workflow, source systems, decision points, human review steps, risks, and first proof of value. Write it for an operator, not an engineer.

If connected apps are enabled, ChatGPT can also help reason over relevant internal docs, project files, or meeting notes, depending on workspace permissions.

What not to do

Do not ask the agent to close the sale.

Do not ask the agent to promise ROI.

Do not let the agent email the prospect without review.

Do not ask for a 90 day roadmap before identifying the first workflow.

Do not let the agent invent access to systems it does not have.

The first job is clarity. Implementation comes after.

A strong follow-up email shape

The agent can draft something like this:

Thanks for sending this over. The strongest first workflow looks like <workflow>, because it already has source context in <tools> and a clear review step with <owner>.

The first useful proof would be <small loop>. Before we build, I would want to see 2 or 3 real examples of the input, the current output, and the place where the finished work lands.

If that matches what you had in mind, the next step is a short implementation call where we map the handoff, confirm access, and decide what should be drafted, automated, or kept under human review.

That is specific enough to earn trust and flexible enough to keep the human in control.