Commonwealth AI Partners
What We Deploy DEP.01

An assistant that handles the busywork

It writes the first draft. Your staff decides what goes out.

The follow-ups, the recaps, the reminders, the summary somebody has to write before Thursday — drafted for you, from your own records, in your own voice. Nothing leaves the building until a person on your staff reads it and hits approve.

Book a Systems Audit 30 min · If we’re not the right fit, we’ll tell you who is.
What we build on Built on Claude, ChatGPT and Gemini — set up in accounts that belong to you, not resold through us.

Diagnosis

You need this if your highest-paid people are doing the dumbest work.

This is not a chatbot bolted onto your website. It’s an assistant that reads what your staff reads, drafts what your staff writes, and hands the draft back to a person to approve. It does not talk to your members, your visitors or your customers on its own, and we don’t build it that way even when asked.

We get hired when the writing is repetitive enough to be a waste of a good person and the stakes are high enough that nothing should go out unread. Common signs:

  • The same four emails get written from scratch every week, by whoever has the least time to write them.
  • Nobody has read the notes from the last meeting, so half of it gets discussed again.
  • Requests sit in an inbox because deciding who handles them is somebody’s job, and that somebody is busy.
  • You tried a chatbot on the website. It made things worse.

What you get

Six concrete deliverables. No vapor.

01 Deliverable

It reads the pile so your staff doesn’t have to.

Replies, sign-ups, forms, meeting notes, the attachment somebody sent at 4:50 on a Friday. It reads all of it and tells you what’s in it in a paragraph, with the original one click away when you want to check.

02 Deliverable

It drafts the messages that keep getting put off.

The renewal note, the sponsor follow-up, the thank-you, the recap of the event. Written in your organization’s voice, using your own records, sitting in a queue ready to edit instead of sitting in somebody’s head.

03 Deliverable

It works from your real records, not from guesses.

It looks up what this person actually paid, which event they actually came to, who spoke to them last. When it doesn’t know something, it says so and leaves a blank instead of filling the gap with something plausible.

04 Deliverable

Nothing sends itself.

Every draft waits for a person. Approve it, edit it, or throw it out. If you later want the lowest-stakes items to go out on their own, we can turn that on once you’ve watched it long enough to trust it — and only if you ask.

05 Deliverable

You can see what it did and why.

Every draft it wrote, what it looked at to write it, who approved it and when. If a board member asks how this works, you can show them rather than describe it.

06 Deliverable

Written instructions, in plain English.

A short guide anyone on your staff can follow, written so the person who takes this over in two years doesn’t have to call us. The handoff is the deliverable, not an afterthought.

Architecture

Drafts. Not sends.

On the left, the things that come in: email, sign-ups, documents, notes from the meeting, what people paid. In the middle, the assistant reads them and checks them against your own records. On the right, what comes out: drafts, sitting in a queue, waiting for a person on your staff to approve, edit or bin them.

Every step keeps a record — what it read, what it wrote, who approved it and when. Being able to show your work is a deliverable, not an afterthought.

DEP.01
Title
How the assistant works
DWG No.
201.01.00
Scale
1 / 100

Process

Six weeks, written down. No surprises.

  1. Week 0

    Systems audit

    Free 30-minute call. We find where the hours are actually going and tell you whether this is worth doing. If it isn’t, we say so.

  2. Week 1

    Watch the work

    We sit with the people doing the job and watch how a real week runs. Then we write down, in your words, what a good draft looks like and what should never be touched.

  3. Week 2-3

    Put it in writing

    You get a plain-English plan: what the assistant handles, what it stays away from, who approves what — and a fixed price for the build. Nothing gets built until you’ve signed off on that page.

  4. Week 3-6

    Build

    We build it inside accounts that belong to you. Short check-ins, something working to look at every week, measured against the standard you set in week one.

  5. Week 6-7

    Turn it on

    First it runs quietly beside your staff, so you can compare its drafts to what your people would have sent. Then it goes live with approvals. The written instructions land here.

  6. Ongoing

    Care

    Optional. We keep an eye on it, fix what drifts and adjust as your work changes. Or we hand it over clean and you take it from there.

FAQ

The questions you’d ask on a call.

Q.01 Is this going to email our members without us?
No. It drafts, a person on your staff reads it, and that person sends it. We don’t build anything that talks to your people without a human in between, and we’ll say no if you ask us to.
Q.02 Are we stuck with one AI company?
No. Nothing is welded to a single one of them. Moving to a different one is an afternoon of work rather than a rebuild, and we’ve done it before when a better or cheaper option showed up.
Q.03 How do we know it isn’t making things up?
You decide in week one what a good draft looks like, and every change we make afterward gets checked against that standard. Each draft shows which of your records it used, so anyone can open the source and check a claim before it goes out. When it doesn’t know, it is built to say so.
Q.04 What happens to our information?
It stays in accounts that belong to you, in your name and on your card. We don’t keep copies and we don’t mix your information in with anyone else’s. If a contract or your insurer needs that in writing, we set it up with the AI company directly.
Q.05 How long until something works?
Two weeks to something you can watch running beside your staff. Six to seven weeks to your team using it daily with approvals. Narrower jobs go faster, and we’ll tell you which one yours is on the audit call.
Q.06 What does this cost?
A typical build lands between $25K and $80K depending on how much it has to handle, at a fixed price set after the first week. The optional monthly arrangement afterward starts at $4K/mo. We quote precisely after the audit.
Q.07 We already bought an AI tool and it isn’t working.
We do two-week look-ins. We go through what you have, write up what’s broken and why, and either fix it or tell you to switch it off. Half of that fee comes off the build if you go forward.

Engagement

What working with us looks like.

Typical scope
$25K – $80K

A fixed price, set after the first week, before anything is built. No surprise invoices.

Typical timeline
6 – 7 weeks

From signed paperwork to your staff using it daily. Faster when the job is narrow.

Optional retainer
From $4K / mo

We watch it, fix what drifts and adjust as your work changes. Cancel anytime.

Not included
What the AI company charges to run it · any new software you decide to buy · staff hired on your behalf

We only spend money you’ve approved.

Next step DEP.01.99

Tell us where AI shouldn’t be.

Half the value of the call is hearing where this doesn’t belong in your operation. Thirty minutes, no deck, no pitch. If we’re not the right fit, we’ll tell you who is.