How to Write an AI Use Policy That Matches Real Work

The owner or CEO names one accountable person, and that person gives employees an approved way to finish a real task.

ES
EdgePoint Strategy
AI Strategy
November 19, 2026
ShareLinkedInTwitter

Most mid-market manufacturers and distributors already have employees using AI at work. A policy that only lists banned tools will not change that, because people still need a practical way to finish quoting, engineering, and customer service work.

Why Ban-Only AI Policies Fail

People use AI when it helps them get their work done. If your policy tells them what they can’t use, but doesn’t give them an approved path for the same task, they will find their own tools. That pattern shows up in the research and on the shop floor.

Gallup’s August 2026 survey of 21,648 U.S. employees found that 53% use AI in their role at least a few times a year, and 32% use it a few times a week or more. A PagerDuty survey of 1,250 office professionals at companies with $500 million or more in revenue found that 66% of those who have used AI at work did so even though they believed company policy didn’t allow it. In the same survey, 88% have shared work information with public AI tools, and 34% have entered customer data.

Most policies haven’t kept up with that. ISACA’s 2026 AI Pulse Poll of more than 3,400 digital trust professionals found that 90% believe employees in their organization use AI, while only 38% of organizations have a formal, comprehensive AI policy.

Those numbers matter less than the question employees actually ask. They want to know whether they can use this tool with this information for this task. Somebody inside the company has to own that answer.

What Ownership Looks Like in Practice

The owner or CEO should sponsor the AI use policy and name one internal person who is accountable for it. That person answers day-to-day questions, keeps the one-page guide current, and makes sure approved tools stay aligned with real work.

Internal IT should run the program with the department leads who own the work. Quoting, engineering, and customer service each know which information is sensitive and which steps already have a human review. IT configures the tools and controls, and the department lead decides whether the output is good enough for customers.

If your company has no IT leader, the CFO should own the policy. The managed service provider can help configure the tools and controls, but do not hand ownership to the MSP. An outside vendor can support the setup, yet they can’t answer an employee’s judgment call about customer data or quoting files.

NIST’s Generative AI Profile (NIST AI 600-1) is a useful public reference for risk categories. Your policy still has to answer local questions about your files, your customers, and your review steps.

How to Learn What People Already Do

This month, the accountable person should spend a week with a few people in quoting, engineering, and customer service. That person should ask what they are already using AI for, what information goes into those tools, and what they would use if the company offered an approved option.

That person should listen for repeated tasks, not one-off experiments. A shared need across a few people is a better place to start than a wish list from one power user. That person should write down the information types that show up again and again: customer drawings, pricing sheets, open order notes, supplier emails, and personal data.

The accountable person should not treat the interviews as a hunt for policy violations. People will tell that person more if they believe the goal is a workable path, not a list of violations. The goal is to replace shadow use with one approved way to do a job they already do.

That person should ask each person to walk through one recent example, and should note where the file came from, who else touched it, and whether a customer would recognize their own information in the prompt. Those details tell the accountable person what belongs in an approved tool and what has to stay out.

Start With One Common Task

After the interviews, the accountable person and the department lead should pick one common task and give people an approved way to do it. Keep the first release small enough that the accountable person can support it and the department lead can review the results.

A useful pattern shows up with customer-submitted artwork. A team wants AI to help check that artwork and often starts by asking for a custom agent. The accountable person and the department lead should ask first whether a shared prompt in an approved company tool, used at the existing review step, would do the job. If a shared prompt would cover the need, the accountable person and the department lead should start there instead of building an agent.

The accountable person and the department lead should be explicit about which files the team can use in that prompt. Keep the person who already owns the artwork review responsible for the final call. The tool helps them check the work, and it does not replace their judgment. That approach gives employees something they can do this week, instead of waiting for a larger AI project.

Once that first task works, the accountable person and the department lead should put the same pattern on the next common job. A clear AI decision map before buying tools helps you avoid buying software before you know which task and which data are in scope. When a process is ready for production use, name an owner for that AI workflow so questions do not bounce among IT, the MSP, and the department.

What Belongs in the One-Page Guide

The accountable person should publish a one-page guide with real examples from the task just approved. Employees should be able to scan it in a minute and know what to do next.

First, show the approved way to do the task, with a short example of a good prompt and a good output check. Second, state what stays out of personal AI accounts, including customer files, pricing, drawings, and any data you would not email to a stranger. Third, name who to ask when they are unsure. The accountable internal person handles policy questions, the department lead handles work quality, and internal IT or the CFO handles tool access.

Keep the banned list short and specific. A long list of forbidden products does not help someone who needs to finish a quote, while an approved path for that quote does.

The accountable person should review the one-page guide whenever a new task is approved or a tool changes. The owner or CEO should expect a short update from the accountable person, not a binder of policy language nobody reads.

How to Announce and Launch the Policy

The owner or CEO announces the policy, names the accountable person, and sets the expectation that approved use is welcome. The accountable person publishes the one-page guide and holds a short walkthrough with quoting, engineering, and customer service. Internal IT, or the CFO with MSP help, turns on the approved tool and the related access controls before that walkthrough.

The accountable person should give people a place to ask questions during the first two weeks. Most uncertainty shows up early, when someone has a real file and a real deadline. The accountable person should answer those questions in the guide so the next person does not have to ask the same thing.

The accountable person should measure follow-through in plain terms. That person should check whether people use the approved path for the chosen task, whether personal AI accounts still receive customer files, and whether department leads are comfortable with the review step. That person should adjust the guide based on those answers.

An AI use policy that employees follow starts with a clear owner and an approved way to do real work, not a longer list of bans. The owner or CEO should have the accountable person begin with one task people already need to finish.

If you want help sorting ownership and the first approved workflow, talk with our team.

Sources

Frequently Asked Questions

Who should own an AI use policy?

The owner or CEO should sponsor the AI use policy and name one internal person who is accountable for it. That person answers day-to-day questions, keeps the one-page guide current, and makes sure approved tools stay aligned with real work. Internal IT should run the program with the department leads who own the work. If the company has no IT leader, the CFO should own the policy. The managed service provider can help configure the tools and controls, and the owner or CEO should not hand ownership to the MSP.

What should the accountable person learn before approving a task?

This month, the accountable person should spend a week with a few people in quoting, engineering, and customer service. That person should ask what they are already using AI for, what information goes into those tools, and what they would use if the company offered an approved option. That person should write down the information types that show up again and again: customer drawings, pricing sheets, open order notes, supplier emails, and personal data.

What should the one-page guide include?

The accountable person should publish a one-page guide that shows the approved way to do the task, states what stays out of personal AI accounts, and names who to ask. Customer files, pricing, drawings, and any data the company would not email to a stranger stay out of personal AI accounts. The accountable person handles policy questions, the department lead handles work quality, and internal IT or the CFO handles tool access.

How should the owner or CEO launch the policy?

The owner or CEO announces the policy, names the accountable person, and sets the expectation that approved use is welcome. The accountable person publishes the one-page guide and holds a short walkthrough with quoting, engineering, and customer service. Internal IT, or the CFO with MSP help, turns on the approved tool and the related access controls before that walkthrough.

Tags

AI governanceAI use policyacceptable usemanufacturingdistribution

Work Through the Next Decision.

Start with a free initial conversation. Assessments, advice, and delivery are separate paid engagements with an agreed written scope.

Talk through your situation