AI governance

The four AI governance foundations marketing and communications teams need before AI becomes part of everyday business.

An AI-generated product claim reaches your website. A confidential client brief goes into an unapproved tool. A synthetic image looks like a photograph of work your company actually delivered.

Each starts as a production decision. Each can become a question about your organisation’s credibility.

For marketing and communications leaders, AI governance now reaches into familiar territory: who can speak for the business, what evidence supports a claim, how customer information is handled, and who takes responsibility when something goes wrong.

A useful way to assess your organisation’s readiness is through four practical foundations of AI governance: accountability, clear rules, working oversight, and people who know what to do. These are an editorial framework for assessing communications readiness, rather than a formal certification or universal standard.

1. Accountability: someone owns the decisions

When AI contributes to a campaign, who is responsible for the result?

“The person using the tool” is an incomplete answer. That person may have no authority to approve the tool, assess its handling of confidential information, or decide whether a synthetic spokesperson is appropriate.

Effective AI governance requires clear ownership at two levels: responsibility for how AI is adopted, and responsibility for each piece of work it produces.

For marketing communications, this means identifying who approves tools, who checks sensitive claims, who authorises publication, and who can stop a campaign.

An agency also needs to understand its client’s expectations. Permission to produce a campaign should not be assumed to cover every possible use of AI, particularly where confidential material, someone’s likeness, or generated evidence is involved.

The test: If an AI-assisted campaign attracted a serious complaint tomorrow, would everyone know who leads the response?

2. Clear rules: people know what is acceptable

“Use AI responsibly” gives a team very little help when a deadline is approaching.

Useful rules address actual decisions. Can staff upload an unpublished financial announcement? Can they generate a customer testimonial? Can an agency use a senior executive’s voice to create new audio? What checks are required before publishing an AI-assisted translation?

At a minimum, an AI governance policy for marketing teams should cover:

  • Approved tools and the information permitted in them.
  • Uses that require additional review or are prohibited.
  • Verification of facts, quotations, statistics and product claims.
  • Permission to use people’s voices, likenesses and creative work.
  • When and how audiences should be told about AI use.

The distinction between drafting and evidence deserves particular attention. AI can help express a genuine customer story. An invented customer story presented as real creates a different problem.

NIST’s Generative AI Profile identifies risks including confidently expressed false information, privacy, intellectual property and information integrity. These provide useful prompts for reviewing communications policies. NIST Generative AI Profile

The test: Could a new team member use your guidance to make a sensible decision without interpreting a broad ethical statement?

3. Working oversight: the checks happen in practice

AI governance becomes useful when it changes how work is commissioned, reviewed and approved.

The level of review should reflect the consequences of an error. Brainstorming internal headlines and publishing a medical product claim need different treatment. A realistic generated image of a factory also needs scrutiny if viewers could reasonably believe it depicts your facilities.

Build the checks into the existing workflow. A brief can record planned AI use. A reviewer can require the original source for a statistic. Final approval can confirm that images, claims and quotations accurately represent the business.

“Human reviewed” should mean that someone checked the evidence and context, with enough time and authority to challenge the output. Reading a draft for fluency is insufficient.

For higher-consequence work, retain enough documentation to explain where important claims and assets came from, who reviewed them, and what was approved.

The test: Can you show how your checks operated on a recent campaign?

4. People and readiness: the team can recognise and respond to problems

AI governance depends on practical training that gives teams the confidence to question convincing output.

A writer should know how to trace a claim to its original source. A designer should recognise when an image could misrepresent a product. An account manager should know when a proposed use needs client approval.

Organisations also need a response process. If an AI-assisted asset is inaccurate or misleading, who pauses distribution? Who assesses its reach? Who corrects it? When should clients, customers or other stakeholders be informed?

Practise with a realistic scenario. Imagine an AI-assisted press release contains a fabricated executive quotation and has already reached journalists. Work through the decisions before a real incident forces them.

The test: Would your team recognise the problem, raise it promptly, and know what happens next?

What AI governance changes about your marketing communications

These foundations affect both how you produce communications and what you can credibly say about your organisation.

If you describe your business as a responsible adopter of AI, that statement should have operational support. If you promise confidentiality, your production practices need to support that promise. If you publish an image as proof of your capabilities, its origin and accuracy matter.

Start by reviewing one campaign from brief to publication. Identify where AI enters the process, what information it receives, what it produces, and who checks the result. Include work done by agencies and freelancers.

That exercise will often reveal more than a general discussion about whether the organisation “has an AI policy”.

Established frameworks can support the wider programme. NIST’s AI Risk Management Framework, for example, organises risk management around Govern, Map, Measure and Manage, with governance running across the other functions. It treats risk management as an ongoing activity throughout the life of an AI system. NIST AI RMF Core

For communications leaders, the immediate question is practical: can your organisation explain and stand behind the work it publishes?

If ownership is unclear, rules are vague, checks are inconsistent, or people are unprepared, you have identified where to begin.

To explore what AI governance means for your brand, agency relationships and communications workflow, start a conversation with Foundeast.