Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Glossary term

AI Use Policy

A team's written rules for AI: approved tools, what each use may do, what data may go in, what is reviewed, when customers are told and who owns problems.
StrategyconceptUpdated

Definition

An AI use policy is a team's agreed rules for using artificial intelligence in its work. In marketing that covers standalone assistants, the generative features built into tools the team already pays for, and any automated system that talks to customers on the team's behalf. The policy exists because each person using these tools otherwise makes their own call, under deadline, about questions the team never discussed: whether a customer list may go into a chat window, whether a drafted email needs a second reader, whether a generated image needs a label.

The tutorial on writing a marketing AI use policy frames the document as decisions written down so that nobody has to make them again. The policy states the decisions; the guides it links to hold the method. That division keeps the policy stable when a checklist changes.

What a policy settles

A policy that does its job answers a short list of questions, each matched to a failure most teams have already seen in some form:

  • Scope and tools. Which tools and features count, which are approved and on which plan, since the consumer and business plans of one product can carry different data terms.
  • Uses by tier. How much control each kind of use needs, judged by what happens if the output is wrong and nobody notices, not by how impressive the tool is.
  • Data. What may go into a prompt or upload, with personal data limited to tools whose terms allow it and only when the task needs it.
  • Review. Which outputs a person checks, how, and who signs off.
  • Disclosure. When customers and platforms are told that AI was involved.
  • Ownership and incidents. A named owner for each use, what counts as an incident, who hears first, and a review date that keeps the policy current.

In practice: the register, agencies and new tool requests

The policy is drafted from what the team already does rather than what it should do. The AI use register template records each use as a row, with its tool, team, data and output, and the policy points to it; a register usually turns up more built-in features than anyone expected. A reviewer from legal, compliance or data protection reads the draft before it is published, particularly the data and disclosure clauses.

Two boundaries decide whether the policy actually reaches the work. The first is suppliers: the guide to setting AI rules for agencies extends the policy to agencies, freelancers and contractors, who often make most of the customer-facing work in their own tools. The second is new tools. The guide to handling requests for new marketing tools routes AI requests to the policy and triages them against the register, instead of restating the rules on the request form. Launches lean on it too: the AI customer service launch checklist takes its scope and data lines from whoever owns the policy, and records the new use as the policy requires.

Illustrative example — not a real result

A housing association writes its rules after an inbox discovery

Sneldeld Housing Association's communications team of 6 found, during a routine data protection check, that 3 officers had been pasting tenant emails into a free chatbot to draft replies. Nobody had told them not to. The head of communications listed every AI use the team had, which came to 11 rows, then wrote a policy of a single page: one approved assistant on the association's business plan, no tenant names, addresses or case details in any prompt, every reply or public post drafted with AI read by a named officer before sending, and incidents reported to her within 1 working day. The data protection officer reviewed it, the team walked through it in a 45-minute session, and a review date was set 6 months out.

Most of a policy's data rule is about what goes into a Prompt, and the strictest line usually concerns Personal Data. The policy's review rules reuse the team's existing Approval Matrix rather than inventing a parallel sign-off route. Many of the uses it governs run on a Language Model, but the policy is about the uses and their risks, not the technology.

Common confusions

  • An AI use policy is a legal document. The marketing team drafts it because only the team knows its tools and work; legal and data protection review it. The obligations themselves are for the organisation's advisers.
  • Built-in features are outside the policy. Subject-line suggestions, asset generation in ad platforms and reply drafters in the service inbox are uses like any other, with data and review implications.
  • The longer the policy, the safer. A policy people cannot remember is consulted once and then ignored; the detail belongs in the guides it points to.
  • The policy covers whoever reads it. Agencies and contractors working in their own accounts are outside it until the relationship extends it to them.
  • Approving a tool settles the matter. Approval is per use and per plan, with an owner for each use.

More terms to look up

Where this term is used

Related