Watch it first (2:48): Two Teams, One Table
The story
Two fictional companies are combining. Meridian Trust has a structured culture where people stay 22 years in the job. Fernway is kinetic and fast.
Denise, from Meridian, and Marcus, from Fernway, are handed the same real piece of work: a loan file with a flagged discrepancy. That's the whole setup. No town hall, no CEO, no slide about synergies.
Telling it through distance
The film tells its story through proximity. Denise and Marcus start in entirely separate frames. Then they meet at a shared table, a chair-length apart. By the end they're sitting close enough that the distance itself has become the plot.
There's almost no dialogue until the back half. Nobody announces that the merger is working. The audience does the math.
That restraint was a rule from the start. The visual brief explicitly rejected anything that looks like an AI avatar reading a script to camera, because that's the fastest way for a piece like this to fail.
How it was built
This was not a single prompt typed into a video model. It was a pipeline of stages, and a person approved each one before the next could begin.
- Script. A full manuscript, scene by scene, with confirmed voiceover copy and original dialogue. If the words alone don't work, nothing downstream will save them.
- Characters. Several visual directions were generated and reviewed, and one was locked per character. A facial-uniqueness step then wrote each face's specific details into a reusable spec: a freckle cluster and two-texture hair for Denise, a fade and a brow asymmetry for Marcus. Every later image reads that spec, so nothing downstream can quietly change what they look like.
- Voices. Auditions were generated against written casting briefs, reviewed, and locked per character. The one exchange in the film was written from the same lines used to cast the voices, so the casting and the performance came from the same material.
- Locations. Reference stills were generated, reviewed and locked. The losing candidates were deleted so they couldn't drift back in.
- Production. Stills were generated scene by scene, then animated clip by clip. The one face-to-camera moment was lip-synced to the locked voice. Score, effects and ambient sound were built and mixed underneath, not dropped on top.
What was automated, honestly
It would be easy to say "AI directed this film." It didn't.
The shot-by-shot motion sequencing ran on a hand-planned script. No AI agent made its own camera or pacing decisions. The agentic work happened earlier in the chain: design, casting and continuity. Several narrow systems each own one decision (what a character looks like, how they sound, what a room looks like), and each decision is written to a locked file the moment a person approves it.
That's what keeps dozens of separate generation steps consistent. It works much like a real crew, with a casting department, an art department and someone watching continuity.
Four days instead of four weeks
The film took about four days to make, against roughly four weeks for traditional production. No cameras, no actors, no location shoot.
Speed matters here for a simple reason. The moments inside a merger don't wait. A film that arrives six weeks after the announcement is answering a question people have already stopped asking out loud.
The reaction
A fractional HR professional watched it with no introduction and no explanation. She was visibly moved.
One reaction is not a study. But it was the test I cared about: does it feel like what that moment actually felt like?
A demonstration, not a client project
Meridian Trust, Fernway, Denise and Marcus are fictional. No real employee's face or voice was used. I made this film to show what's possible before asking anyone to trust me with their own story.
If your organization is going through a moment that a memo can't carry, I'd like to hear about it: gerard@architectsofthecloud.com.
Gerard Beaubrun is a software and AI architect who has built systems in healthcare, education and finance, most recently AI for M&A due diligence at PwC. He leads Corporate Storytelling at Architects of the Cloud.
