Outlook
Part of How video marketing agencies can plan ahead
What UK buyers should check when AI changes video marketing agency briefs
How AI changes the brief, data duties and the review cycle for video marketing agencies, and what UK buyers should check before signing off.
What to take away
- Decide first whether AI sits in the creative work, the editing work or the distribution work, because each carries a different approval route.
- Ask every supplier which model or tool touches client footage, and get that written into the statement of work.
- Treat training data, consent and retention as procurement questions, not technical ones.
- Expect the review cycle to shorten, so budget for more frequent sign-off rather than one large approval.
- Check the legal duties that apply to the platform you publish on before the first cut is delivered.
Where AI sits in agency work
Most video marketing agencies AI applications fall into three areas: script and concept drafts, post-production tasks such as rotoscoping and upscaling, and distribution work like captioning, tagging and audience matching. Each area has a different risk profile.
AI Areas and Risk Profiles
Script and concept drafts
- Example tasks
- Drafting concepts
- Risk level
- Low
- Main risk
- Little if rewritten
Post-production tasks
- Example tasks
- Rotoscoping, upscaling
- Risk level
- Higher
- Main risk
- Consent reuse
Distribution work
- Example tasks
- Captioning, tagging
- Risk level
- Medium
- Main risk
- Audience mismatch
A concept draft carries little risk if a human rewrites it. An automated edit of client footage carries more, because the output may be reused in ways the original consent did not cover.
For a wider view of how these shifts sit alongside demand and pricing, see video marketing agencies trends and outlook, which maps the commercial pressures behind the technology change.
Data protection duties
If a tool processes personal data, the usual UK GDPR principles apply. The Information Commissioner's Office publishes guidance on artificial intelligence and data protection covering lawful basis, transparency and automated decision-making.
Data Protection Questions for Suppliers
- Is footage of identifiable people used to train a model?
- How long is the footage retained?
- Where is the footage stored?
- Is there consent for voice cloning and likeness tools?
- Does consent cover synthetic replicas in later work?
In practice, that means asking whether footage of identifiable people is used to train a model, how long it is retained, and where it is stored. A supplier that cannot answer those three questions is not ready for a client campaign.
The same logic applies to voice cloning and likeness tools. Consent for a one-off shoot is not consent for a synthetic replica of the same person used in later work.
Platform and legal rules
The Online Safety Act 2023 sets duties for many online services that host user-generated video, and the full text of the Online Safety Act 2023 is the starting point for checking which duties apply.
Separately, the government's digital strategy executive summary sets out the direction of travel for digital adoption across the economy, which helps explain why clients now ask about AI in pitch documents.
Advertising rules still apply to AI-generated claims and imagery. A synthetic scene that implies a real result can breach the CAP Code even if no human was misled intentionally.
Scoring rubric for supplier answers
Use a simple scoring table when you compare agency responses. Score each line from one to five, then weight the lines that matter to your sector.
| Check | What a strong answer looks like | Weight |
|---|---|---|
| Tool disclosure | Names the models and vendors used on client work | High |
| Training data | States whether client footage trains any model | High |
| Retention | Gives a retention period and deletion route | Medium |
| Human review | Names who approves each AI-assisted output | High |
| Sub-processors | Lists any third party touching the footage | Medium |
| Audit trail | Keeps a record of prompts and versions | Low |
A supplier scoring well on disclosure and human review is usually safe to brief even if the audit trail is thin. A supplier that scores badly on disclosure is a risk regardless of the rest.
Review cycle and triggers
AI-assisted production shortens the gap between concept and cut. That changes the approval rhythm. Instead of one sign-off at the end, expect two or three checkpoints during the edit.
Set review triggers in the contract. Examples include a new model being introduced, a change in sub-processor, or a shift in the platform hosting the final video. For a structured way to plan around those events, read video marketing agencies risk scenarios in England, which sets out the situations that most often force a re-brief.
Budget implications are modest at first. For example, a team paying £400 a month for editing tools may find the licence cost rises to £700 once AI features are added, and that increase should sit in the production line rather than a hidden overhead.
Common questions
Do AI tools change who owns the finished video?
Ownership depends on the contract, not the tool. Confirm in writing that the agency assigns rights in the final output to the client, and that any third-party model terms do not restrict commercial use.
Can an agency use client footage to train a model?
Only with a lawful basis and clear notice. Ask for the specific purpose, the retention period and the opt-out route before any footage is uploaded.
How often should the brief be reviewed?
Review at least once a year, and immediately whenever a new model, sub-processor or publishing platform enters the workflow.
Does the Online Safety Act apply to brand video?
It applies to many services that host user-generated content. Brand video published on a regulated platform falls under the platform's duties, so check the host's terms as well as your own.
