Product and implementation management

Every project decision, with who approved it, when and why.

Paste the meeting transcript into the AI you already use. The agent picks out what was decided, by whom and why, and drafts the proposals. The client confirms through a link. Months later, when someone asks who asked for what, the answer is there.

Private beta, by invitation only. Join the waitlist to get your spot.

Your AI assistant · Checkout project
  1. Here’s the transcript from today’s call with Carla.

    call-mar-12.txt · 42 min
  2. Hands-off · reading the transcript

  3. I found 3 decisions and 2 next steps, and drafted the decisions as proposals:

    • D-014Checkout button: red to greenCarla · 2:32 pm
    • D-015Free shipping above $150Carla · 2:40 pm
    • D-016Guest checkout stays onBruno · 2:51 pm
  4. Send them to Carla to confirm.

  5. Hands-off · confirmation link sent to Carla

Hands-off · Checkout project

Mar 12 call3 decisions

  • D-014Checkout button: red to greenWaitingApproved
  • D-015Free shipping above $150WaitingApproved
  • D-016Guest checkout stays onWaitingApproved

v1.4 3 decisions added · approved by Carla at 3:40 pm

Ana · via Hands-off

Confirm 3 decisions from the Mar 12 call

  • Checkout button: red to green
  • Free shipping above $150
  • Guest checkout stays on
A made-up project, to show the flow.

The problem

Project documents die once they're done.

PRDs, ADRs, business rules, decisions, personas and notes are written with care and then go quietly stale, because nothing makes anyone keep them up to date. The same problem comes back on every project.

Scattered decisions

They live in meetings, chats and files, with no record of who decided and why.

Loose sign-off

Client and consultant comment in different places, and nobody is sure what's approved.

No way to check

Months later, it's hard to say whether the system still follows the rules everyone agreed on.

No map of the system

It's unclear which function does what, where it lives, or which rule it answers to.

How it works

From meeting to delivery, and nobody has to take minutes.

The unit is the decision, not the text edit. Every meeting and every work session turns into proposals, and only a person approves.

  1. After the meeting

    Paste the Zoom, Meet, Teams or Fathom transcript into the AI you use. The agent picks out what was decided, by whom, when and why, plus the next steps.

  2. Decision makers confirm

    The client and the other stakeholders get a personal link, read each decision and approve or comment, with no new tool to learn.

  3. During development

    At the end of a session, the developer asks to update the project. Technical decisions keep their reason; anything that goes against an approved one goes back to the decision makers.

  4. When someone asks

    "Who asked for the green button?" It's there: the Mar 12 meeting, approved by Carla at 3:40 pm, shipped in version 1.4.

A new decision replaces the old one, with the reason. Nothing gets deleted.

The bet

Sign-off across every party, and a check afterwards.

A wiki stores text. Hands-off holds your project documents too, in an editor like Confluence’s, plus who decided, why, and whether the delivery followed the decision.

Documents you edit by hand

PRDs, ADRs and BDRs in an editor like Confluence’s: headings, sections, tables and images. People edit them, not just the agent.

Everything on the record

Every save becomes a page revision with who, when and what changed. Edits to approved content go back for review, and nothing is ever deleted.

Multi-party sign-off

Consultant, client and stakeholders comment right on the text and approve what falls in their area, all in one place.

Version check

Compares what was decided with what was delivered, so you can answer without digging through old threads.

Versions made of decisions

An edit becomes a page revision; an app version comes only from approved decisions. You see the current version and how the product got there.

The agent does almost everything

It drafts, links the pieces, summarizes, flags contradictions and opens proposals, using the same actions people use on screen.

Rules that don't change

  1. The agent does almost everything, except approve.
  2. Every change shows who made it, person or agent, and at whose request, and stays in the audit log.
  3. A decision is never deleted: a new one replaces it, with the reason.
  4. The agent has no back door: it uses the same actions as the screen.
  5. Anything that goes against an approved decision goes back to the decision makers.

Designed to work with any AI, without depending on a specific platform.

Who it's for

For anyone who needs the client to decide with them.

Consulting, implementation and development projects, where more than one party has to agree before something counts.

  • Consultancies
  • Independent consultants and freelancers
  • Product managers
  • Project managers
  • Developers
RoleWhat they doApproves?
Consultant or implementerRuns the project, proposes decisions and maps the systemYes
ClientReads, comments on and approves rules and decisions, including through a link, without signing in to the toolYes
Other stakeholdersComment on and approve what falls in their areaYes
AI agentDrafts, links the pieces, summarizes, flags contradictions and opens proposalsNo

Private beta

Get your spot in the private beta.

Hands-off opens by invitation. All it takes is your email; on the next screen, two optional questions about how your team decides today. We invite people from the list first, a few teams at a time.

The form opens soon.

Come back in a few days to join the list.

We only use your contact details to talk about Hands-off. Privacy policy.

FAQ

Frequently asked questions

Can I use it yet?

By invitation. Hands-off is in a private beta that opens to a few teams at a time; join the waitlist to get your spot. The first version proves one loop: from proposal to approved decision, and from there to a version.

Does it replace Notion, Confluence and ProdPad?

That’s where we’re headed. Project documents live in Hands-off, in an editor like Confluence’s, and every edit is logged and auditable. On top of that you get what those tools don’t give you on their own: sign-off across several parties, and checking what was decided against what was delivered. The first version starts with documents and decisions.

Does my client need to learn the tool?

No. They get a link, read each decision and approve or comment. On screen, things are called proposals, decisions and versions; developer terms like merge and branch stay out.

Do I have to paste the transcript by hand?

In the first version, yes: you paste it into the AI you already use and the agent does the rest. Integrations that bring the transcript in on their own when the meeting ends are in the plans.

What about meeting privacy?

A transcript is your client's material, so privacy is part of the first version. The details of what is kept, and for how long, are still being defined.

Do I need a specific AI or platform?

No. The idea is to work with any AI and not depend on a specific platform.

How much will it cost?

We haven't set a price yet. People on the list will hear first.