Solutions/Personalised video
Personalised agent video at scale. Compliant.
Instant Studio is a managed AI creative production company for global brands and agencies. For insurers, banks and adviser networks we produce personalised video for agents and advisers by the thousand. The compliance rules are built into a Creative Workflow before the first video is made.
Why does personalised video get stuck in legal?
A regulated company reviews every piece of content before release. A personalised video is a new piece of content for every recipient, so five thousand videos mean five thousand reviews. No compliance team has that capacity. The programme usually shrinks to one generic film with a name added at the end.
The programme goes ahead when legal reviews the workflow instead of the videos. Legal approves the workflow once: the rules, the templates and the way the disclosures behave. Every video the workflow then produces can be traced back to that approved set-up.
The hard part of personalised video is approval.
Rendering a video for each name has been possible for years, and several vendors sell it by the minute. The programme still fails at legal, because every rendered video is a regulated communication and no compliance team can review five thousand of them. Approval is work done with counsel, market by market: the claims, the disclosures, the licence details, the likeness consents. We do that work and put it into the workflow, so it is done once.
What is locked, what is controlled, what is generated?
Compliance agrees the three groups, locked, controlled, generative, before anything is made. They are what the reviewers sign. The specification defines them.
Locked
Fixed in every video
- Regulatory disclosures for each market
- Product terms and approved claims
- Brand elements
Controlled
Filled from approved sources only
- Customer name
- Agent name and details
- Product, date, market and language
Generative
Made by the model
- Greeting
- Presentation and delivery
- Background and setting
Controlled fields come from the client's records or from the person taking part, and nowhere else. The workflow never invents a name, a product or a date. It never rewrites a disclosure. Generation only touches the presentation around those fixed and controlled parts.
Where do the personal details come from?
One of two places. The Creative Workflow, the approval and the brand rules are the same either way.
A data export.
The client sends a list: names, the details to mention and the language for each person. The workflow produces a video for every row. An insurer thanking its 500 top clients works this way. No campaign page is needed.
The people themselves.
When each person has to supply their own photo or details, a campaign page collects them, built with Instant Studio Campaigns. Each person takes part on their phone and receives their own video. Sun Life's agents made more than 5,000 this way in about two weeks.
How are faces, voices and data handled?
A real person's face and voice are used with written consent. The consent is recorded against every video that uses them: whose face and voice, under which terms, for which market and period. The Creative Workflow fixes how that likeness may appear, so every video shows the person the way they and the client approved. No video is made without the consent on file.
The data is handled the same way. Personal details come from the client's own system under the client's data agreement, or from the person themselves on a campaign page that states how the data will be used. Client data is hosted on AWS servers in the client's region, and it is never used to train any AI model.
What is checked on every video?
- The locked parts are present and unchanged: disclosures, terms, brand.
- The personal details match their source exactly: the client's records, or what the person entered.
- The language and script are right for the recipient.
- The video meets the channel's requirements: length, shape, file format.
Review effort does not grow with volume. The claims, the disclosures and any real person's likeness are fixed or controlled before anything is generated. Automatic checks run on every output, and people review the exceptions and the final delivery. If the rules change, the workflow goes back to legal before it runs again.
Where does a programme start?
With the lowest-risk use. Service messages, renewal reminders and agent introductions carry fewer claims than a product pitch. The first pilot proves the approval model where the risk is smallest. Legal is briefed in the week the workflow is built, before any video exists.
For New Year, Sun Life's insurance agents in the Philippines, Hong Kong, Singapore and Malaysia each made a greeting video with their own face in it, from one campaign page, in about two weeks. More than 5,000 went out. Sun Life's lawyers reviewed one workflow, not 5,000 videos. The videos used the agents' real faces and voices beside the likeness of Sun Life's brand ambassador in each market.
Read the playbook
Approve the workflow once: where human sign-off belongs in an AI production line.
Let's get started
Start with a pilot.
Scale with confidence.
Bring your compliance team to the first call. We scope a pilot on one low-risk message. We build the disclosures and the data rules into the workflow, and we put legal at the point where their judgement counts.