Resource

AI Governance Policy Template

6 min read29-point list

Most small and mid-sized businesses already have employees using AI tools at work, usually without a written policy to guide them. This template gives you a plain-language starting point you can adapt to your own business: which tools are approved, what data can and cannot go into them, who reviews AI output before it is used, and who owns the decisions. It is a starting point to adapt with your own legal and compliance counsel, not legal advice and not a guarantee of compliance or certification.

Why a written policy matters, and what this template is

The risk with workplace AI is rarely a single dramatic incident. It is the steady accumulation of small, undocumented decisions: an employee pastes a customer list into a free chatbot to draft an email, a contractor uploads source code to summarize it, a manager ships AI-written copy without checking the facts. None of these people are acting in bad faith. They simply have no guidance, so they improvise. A short written policy replaces improvisation with a shared expectation everyone can point to.

This document is a template, not a finished policy. It is organized into sections you can copy, edit, and drop into your own employee handbook or standalone AI policy. Each clause is written as a plain statement an SMB can adapt to its actual tools, data, and risk tolerance. Treat the wording as a draft to review with your own legal and compliance counsel before you adopt it, especially if you operate in a regulated industry or handle personal, health, or financial data. The goal is a policy people will actually read and follow, so keep it short, specific, and free of jargon.

Scope and purpose: define what the policy covers

Start by stating plainly what the policy is for and who it applies to. Scope should cover employees, contractors, and anyone acting on the company's behalf, and it should name the kinds of AI tools in play: general assistants like chatbots, AI features built into software you already use, coding assistants, and any AI agents or automations that act on their own. Being explicit here prevents the common excuse that 'I didn't know this counted as an AI tool.'

The purpose statement should set the tone: the company supports using AI to work better, and this policy exists to make that use safe, approved, and accountable, not to ban it. People follow rules they understand the reason for. A policy framed as 'here is how to use AI well' earns more compliance than one framed as a list of prohibitions, while still drawing clear lines around the things that genuinely create risk.

Approved tools and data handling: the two clauses that prevent most problems

The single most effective control for an SMB is a short list of approved tools. When the company names which AI services are allowed, and which account tiers or settings to use, employees stop quietly defaulting to whatever free tool they found. Pair the list with a simple request path so that adding a new tool is easy and sanctioned rather than something people do in the shadows. The point is not to slow people down; it is to make the safe option the easy one.

Data handling is the other half. The hard truth is that anything typed into an external AI tool may leave your control and be processed by a third party, so the policy must classify what is allowed in and what is forbidden. Be concrete: name the categories that must never be pasted into a general AI tool, such as customer personal data, credentials and secrets, regulated records, and confidential business information. Then point people to the approved, contracted tools for any work that genuinely needs sensitive data. Read the data and retention terms of the specific tool tier you approve, because consumer and business plans often handle your inputs very differently.

Human review and accountability: AI drafts, people decide

AI output is fluent and confident, which is exactly why it needs review. The policy should establish a simple principle: AI produces drafts, and a responsible person remains accountable for anything used, published, or acted on. This closes the gap where fluent-sounding output gets a free pass that no junior employee's first draft would ever get. The person who uses the output owns it, the same as if they had written it themselves.

Make the level of review proportional to the stakes. Low-risk internal drafts may need only a sanity check, while anything customer-facing, legal, financial, or safety-related needs a knowledgeable human to verify facts, figures, and claims before it goes out. Spell out that AI must not be the sole basis for decisions that significantly affect a person, such as hiring, credit, or termination. Accountability that lives with a named role, not with 'the AI,' is what keeps a confident-but-wrong output from becoming a confident-but-wrong action.

Roles, training, and enforcement: making the policy stick

A policy with no owner becomes stale the week after it is written. Assign clear roles: someone accountable for the policy overall, someone who maintains the approved-tools list and reviews new requests, and a clear expectation that every employee follows the rules and asks when unsure. In a small company these can be part-time responsibilities held by existing staff, but they should be named so there is always a person to go to.

Finally, the policy only works if people know it exists and understand it. Plan a short onboarding briefing, a periodic refresher, and a named contact for questions, so the policy is a living reference rather than a forgotten document. State the consequences of violations in the same measured way you would for any other workplace policy, and review the whole thing on a set cadence because tools, vendors, and risks change quickly. A policy you revisit twice a year stays useful; one you write once and file away does not.

Key takeaway

A short, specific AI policy that names approved tools, forbidden data, human review, and a real owner prevents far more harm than a long one nobody reads, but adapt it with your own legal and compliance counsel before you adopt it.

Practical

Put it into practice.

A copy-ready list to apply to your own workflows, tools, and AI usage.

Scope & purpose

  • This policy applies to all employees, contractors, and anyone using AI tools on the company's behalf.
  • It covers general AI assistants, AI features inside our existing software, coding assistants, and any AI agents or automations that act on their own.
  • The company supports using approved AI tools to work more effectively; this policy exists to keep that use safe, approved, and accountable.
  • When this policy and a more specific contract, regulation, or client requirement conflict, the stricter rule applies.
  • Questions about whether a tool or use is allowed should be raised with the policy owner before proceeding, not after.

Approved tools & access

  • Only AI tools on the company's approved list may be used for work, using the account tier and settings the company specifies.
  • Use company-provided or business-tier accounts for work; do not use personal AI accounts for company data or tasks.
  • To request a new tool, submit it to the policy owner for review of its data handling and terms before first use.
  • Do not connect AI tools, plugins, or agents to company systems, email, or files without approval.
  • Treat AI-generated code, content, and recommendations as drafts to be reviewed, not as finished, trusted output.
  • Access to higher-risk tools and integrations is granted on a least-privilege, need-to-use basis.

Data handling rules

  • Assume anything entered into an external AI tool may leave the company's control and be processed by a third party.
  • Never paste customer or employee personal data, passwords, API keys, or other secrets into a general AI tool.
  • Never enter regulated data (such as health, financial, or other legally protected records) or confidential business information into an unapproved tool.
  • Use only approved, contracted tools with appropriate data terms for any task that genuinely requires sensitive data.
  • Minimize and redact data before using AI: share only what the task needs, and remove identifying details where possible.
  • Verify the data-retention and training settings of each approved tool match the company's privacy commitments.

Human review & accountability

  • AI produces drafts; a named person remains responsible for anything that is used, published, sent, or acted on.
  • Match the review effort to the stakes: heavier scrutiny for customer-facing, legal, financial, or safety-related output.
  • Fact-check AI claims, figures, names, quotes, and citations before relying on them; AI can state false information confidently.
  • Do not use AI as the sole basis for decisions that significantly affect a person, such as hiring, credit, or termination.
  • Disclose AI involvement where a client, contract, regulation, or honesty toward the audience requires it.
  • Do not present AI-generated work as solely human-authored when accuracy or disclosure obligations apply.

Roles & responsibilities

  • A designated policy owner (for example, a manager or operations lead) is accountable for this policy and its updates.
  • A tool administrator maintains the approved-tools list, reviews new tool requests, and manages access.
  • Every employee and contractor is responsible for following this policy and asking the policy owner when unsure.
  • All new staff receive a short AI policy briefing during onboarding, with a periodic refresher for everyone.
  • A named contact answers day-to-day AI questions so people have somewhere to go before they improvise.
  • Policy violations are handled under the company's normal disciplinary process, and the policy is reviewed on a set schedule (for example, twice a year) as tools and risks change.

This is general guidance, not a guarantee of any outcome. Book a call if you would like help applying it to your own business.

Want help putting this into practice?

Book a call to find where AI can save your team time, reduce manual effort, and reduce risk.

Book a Call