01AI guidance for growing businesses

Put AI to work—with clear ground rules.

Give your team the confidence to use AI without guessing what information is appropriate, which actions require approval, or who remains responsible for the result.

How a request flows through your ground rulesYour team’s request enters the pipe and passes through the approved tools filter into the client data check. A branch to off-limits systems is capped shut. A person approves at the valve before the work is sent to the client.Your team’s requestApproved toolsClient data checkOff limitsPerson approvesSent to clientHow a request flows through your ground rulesYour team’s request enters the pipe and passes through the approved tools filter into the client data check. A branch to off-limits systems is capped shut. A person approves at the valve before the work is sent to the client.Your team’s requestApproved toolsOff limitsClient data checkPerson approvesSent to client
  • Access

    Decide what AI can see.

  • Actions

    Decide what it is allowed to do.

  • Approval

    Decide when a person needs to step in.

02The questions

Questions your team is already asking

Without clear rules, employees and AI agents are left to make important decisions about data, permissions, and approvals on their own.

  1. Can I paste this client’s email into ChatGPT to tidy up my reply?

    Answered by Check 02Sensitive information

  2. Which AI tools are we actually allowed to use for client work?

    Answered by Check 01Approved tools

  3. Does someone need to check this AI-drafted report before it goes to the client?

    Answered by Check 03Human review

  4. Should an AI agent be allowed to send that message or change that record without asking first?

    Answered by Check 07Agent permissions

03How it works

Three steps to clear ground rules

  1. Choose

    Choose the right rules for your business.

    Decide which tools staff can use, what information they can share, and which actions need a person’s approval.

  2. Customize

    Customize the ready-to-use templates.

    Adapt the policy, register and checklists to your own tools, your work and your client commitments.

  3. Share

    Give your team clear boundaries they can follow.

    Share the approved rules, name who reviews the work, and give people a simple way to ask or report a problem.

Use AI to move faster while protecting your clients, your information, and the judgment your business depends on.

04Free resource

Start with the free AI Use Checklist for Businesses

Work through seven short checks to see where your team’s AI rules are clear and where people are left to guess. Then use the worksheet to pick up to three priorities, give each one an owner, and set a due date.

  • Three-page PDF
  • Takes 10–15 minutes
  • No sign-up needed
Get the Free AI Checklist

The seven checks

Example answers

  1. Approved toolsYes (example answer)PartlyNo
  2. Sensitive informationYesPartly (example answer)No
  3. Human reviewYes (example answer)PartlyNo
  4. DisclosureYesPartlyNo (example answer)
  5. OwnershipYesPartly (example answer)No
  6. Incident reportingYesPartlyNo (example answer)
  7. Agent permissionsYesPartlyNo (example answer)

05Toolkit

Coming soon

AI Ground Rules Starter Kit

Set workplace AI rules, document the decisions, and roll them out to staff. The kit gives you editable Word and Excel files you can adapt and share with your team.

4 files

Set the rules

Employee AI Use Policy, Approved Tools Register, Data Use Guide, and Roles and Responsibilities.

4 files

Review work and handle exceptions

Human Review Checklist, Incident Report, Vendor Intake, and Agent Permissions Addendum.

4 files

Put the rules into use

A Start Here guide, an Internal Announcement for staff, fictional Worked Examples, and a reply for when clients ask about AI.

The Starter Kit is in preview and not on sale yet. You can read a sample policy section and see the full contents now.

See what’s in the Starter Kit

06AI agents

What counts as an AI agent?

Most AI tools suggest. An agent can act: it can send, book, update or delete something on your team’s behalf. The difference is easiest to see with an ordinary message.

Plate 01One client reply, start to finish
The route
  1. AI drafts the reply
  2. A person reviews it
  3. Approve, or send back to fix
  4. Only then is it sent
A plan drawing of one reply: the AI drafts, a person reviews, then decides whether to approve. If not, it goes back to be fixed. Only approved messages are sent. Hatched zones for client records and off-limits systems sit outside the path.Client recordsOff limitsFixAI draftsPerson reviewsApprove?SendWAYPOINT ADECISIONA plan drawing of one reply: the AI drafts, a person reviews, then decides whether to approve. If not, it goes back to be fixed. Only approved messages are sent.Client recordsOff limitsFixAI draftsPerson reviewsApprove?Send
Legend
  • Approved path
  • A person decides
  • Sent back to fix
  • AI may not enter

A person stays in charge of what goes out.

Suggests

The AI drafts the message

You ask an AI tool to draft a reply to a client. It writes the text. You read it, fix what’s wrong, and decide whether to press send. A person stays in charge of what goes out.

Acts

The agent sends the message

An agent connected to your inbox writes the reply and sends it itself. Nobody sees it first unless you have decided which messages need a person’s approval before they go.

Ground rules for agents answer a few plain questions: what the agent can see, what it is allowed to do, and when it has to stop and ask a person.

07Our method

Written to be adapted to your business

Plain-language policies, checklists, skills, and buyer's guides—built for businesses without a dedicated AI-governance team.

A starting point

Every business uses different tools, does different work and makes different promises to its clients. Our materials give you a clear starting point.

Your decisions

You adapt them to your own tools, your work and your client commitments, and you decide what your rules should be.

The limits

The templates support your business decisions. They are not legal advice, and someone on your team still needs to configure each tool’s settings and permissions.

Find out where your team is left to guess.