Sanjeev Kumar.
← AI in Practice
Anonymised health and wellness work / AI-assisted film

When one person talking stopped working.

We had made plenty of testimonial and single-person UGC films. Then the format began to lose its pull. So we tried a story with two people and a conversation instead.

My part: I led the strategy and creative direction, then worked with my team to make the new format repeatable.

2 peopleA harder AI-film format
1.5×Average ROAS versus normal single-person films
2–3 weeksPeriod of repeated success reported
The work

Give people a scene to stay with.

These three supplied films show different ways into the wider story experiment: a familiar family moment, a pharmacy exchange, and a health-focused film. The everyday situation gives the message somewhere to live.

01 / Family scene

Sarcastic Bahu for Mummyji

A family-table moment built around character and dialogue.

02 / Conversation

At the pharmacy

Put a health question between two people, not into a monologue.

03 / Health story

From concern to action

A more direct health-focused treatment within the same body of work.

These films sit within the story-led group that averaged about 1.5× the ROAS of normal single-person films. That is a group result, not a claim about any one clip.

The experiment

We didn't start with a perfect film.

The problem was creative first: the familiar UGC format was losing its pull. Making two AI characters interact was the craft problem we had to solve next.

  1. 01See the patternSingle-person testimonial films were no longer doing enough for these accounts.
  2. 02Try a conversationMy team researched public techniques, tested dialogue and staging, and got two people into the same story.
  3. 03Release the imperfect first skitThe delivery-person film still had flaws, but it was good enough to test with an audience.
  4. 04Follow the responseAcross the story-led group, ROAS averaged about 1.5× normal single-person films for a short window. We made more skits and also tested filmed versions.

The useful question became: is the story convincing enough for someone to stay?

What we learned

People stayed for the story.

Across these story-led films, ROAS averaged about 1.5 times that of normal single-person films. The format continued to work over the next two to three weeks, so we made more.

The AI-versus-shoot comparison suggested the production method was not the only thing driving response. In this test, a convincing situation and dialogue mattered a great deal.