AI Map
You know which jobs at your company AI can do, which one to start with and how many hours come back.
Tech Lab to hand the work that repeats over to AI
The work that repeats moves into the tools your team already uses: Microsoft 365 Copilot, Claude or Gemini. You start with a map of what AI can do at your company, write one page on what can be uploaded and what can’t, and test it on real documents. People keep the work where a decision is needed.
What changes at your company
The work you’ll see happening, one piece at a time.
You know which work to give to AI
A map of the jobs that get done the same way every time, with the hours they cost each week and the order in which to start.
Everyone knows what can be uploaded
One page that says what goes into the AI tools and what stays out, with examples from your own work. People stop asking for permission and start trying.
Good prompts belong to everyone
What works lives in a shared playbook, written on your documents, instead of in the head of whoever learned first.
The draft arrives ready to check
A reply to a client, meeting minutes, a quote: the assistant prepares it from your documents, and whoever signs it reads it through.
Your team keeps the tools running
One person updates the playbook and the assistant when something changes, without calling anyone.
Start from the step you need
Every step leaves you with something, even if you stop there. Most companies start with the AI Map if they have no licences or don’t know what to use them for, with AI Adoption if the licences are there and few people use them, and with the Assistant if there’s one specific job to take off the desk.
AI Adoption
Your team uses Copilot, Claude or Gemini on real work, with written rules and a prompt playbook built on your documents.
Custom AI Assistant
One repetitive job comes off the desk: an assistant built on your documents prepares it, and your team knows how to update it.
If you’ve tried before
You bought the licences, and two people use them. You tried an automation that nobody opened after three weeks.
It happens when the tool arrives before the rules, and nobody at the company is in charge of keeping it running. You start again from there: where AI actually helps, what can be uploaded, who looks after it after the first month.
The 12 solutions, three per Lab
If what you need sits in another Lab, start here.
Frequently asked questions
Copilot, Claude or Gemini?
It depends on where your team already works. With Microsoft 365, the first one to look at is Copilot, which sits inside Outlook, Word and Excel; with Google Workspace, Gemini. Claude is often chosen for long documents and for assistants built on your own material. They can work side by side, and the AI Map tells you who needs what.
What shouldn’t go into AI tools?
Anything you wouldn’t send to an outside service: personal data about clients and staff, contracts, documents under a confidentiality agreement. The business versions, which don’t use your data to train the models, change the rules. That’s why the rules are written on one page, tool by tool, with examples from your work.
What is an AI agent, and do I need a developer?
It’s an assistant that takes several steps on its own: it reads an email, looks up a figure, drafts the reply. Simple cases are set up inside Copilot, Claude or Gemini without writing code. Connecting it to your management software takes extra technical work: we do it in the Custom AI Assistant, or you ask a supplier, using the specification from the AI Map.
How do you choose who helps you bring AI into the company?
With four questions. Which job at your company would they start from, and why. What they wouldn’t let you upload. How they measure the hours that come back. What your team is left with when the work is done, and who keeps it running.
Before you buy licences or call a supplier: the one-page policy and ten checks. Email it to me
AI in the company, before you start: the one-page policy and ten checks_
Ten checks to run before buying licences or calling a supplier. Tick the ones that are true for you today.
Where should we send the checklist?
It appears below, and we email it to you too, with the outline of the one-page policy.
Here it is
We’ve emailed it to you as well.
The email didn’t go out: the checklist is below, save it or print it.
Your details didn’t reach us, and we won’t pretend they did: if you’d like us to get back to you, write to info@updatelabs.it.
- 01
You know which repetitive jobs your team does every week, and how many hours they cost.
Without this list, AI gets bought because everyone’s talking about it and used out of curiosity.
- 02
There’s a one-page policy: what goes in, what doesn’t, in which tool.
The first thing that ends up in a public chatbot is a client’s contract.
- 03
You know how many licences you have, who uses them and who doesn’t.
Licences bought and never opened are the most common AI expense in small companies.
- 04
You use the business versions of the tools, which don’t train the models on your data.
That’s the whole difference between the free version and the business one.
- 05
You chose the tool based on where your team works: Microsoft 365 or Google Workspace.
Useful AI is the kind that sits inside the windows already open.
- 06
There’s one pilot process to start from, and only one.
Start in every department at once and you finish in none.
- 07
Good prompts are saved in a shared place, with your own documents as examples.
Otherwise the one person who learned leaves with it all in their head.
- 08
You know how to measure the hours that disappear, before and after.
Without the before, the after is an opinion.
- 09
Someone in-house looks after the tools once they’re running.
The tool that works for three weeks and then dies had nobody looking after it.
- 10
You’ve told the team what changes for them, and what doesn’t.
Fear of being replaced holds adoption back more than any technical limit.
If more than three boxes are empty, start with the AI Map, or take the Update Check: it tells you whether AI is the first thing to do.