Citi needed to train a global, distributed workforce on inclusive design and accessibility. The subject carried real weight, ethically and in how the bank showed up for its customers. The ask was a video training series. The problem was that the production infrastructure to make one didn't exist: no central studio, no shared equipment, presenters scattered across countries and time zones with whatever gear was on hand. The traditional route was a multi-month, multi-vendor build that would have delayed training the workforce needed then, not in the fall.
I went a different way. Instead of lowering the bar or stretching the timeline, I built a capture workflow that let presenters record on whatever they had, Zoom, a webcam, a phone, and used AI tools to do the unifying in post.
Adobe Podcast cleaned the audio. Cutout Pro standardized backgrounds across recordings shot in dozens of different rooms. I corrected lighting and color across every clip. ElevenLabs cloned voices to replace VO lines that weren't usable in the original take. Synthesia and Descript produced avatars for segments where the on-camera presenter had to be swapped entirely. The finished series looked like it came out of one studio, which is the trick, because no studio existed.
This was the first time AI creative tools went into production at Citi. At a bank, new tooling moves carefully through legal, compliance and information security, and getting it through those gates was as much of the job as the edit.
The AI never did the creative work. The curriculum, the presenters, the subject-matter expertise were all human. AI handled the unification layer, the part that would have eaten months of editing and a six-figure production budget under the old model. That's what made the program shippable on the schedule the business actually had.
What I took from it: new tools don't replace creative authorship. They shorten the distance between an idea and the thing itself. The work is still the work, you just get there faster, with less space between what's possible and what's practical.
Citi needed to train a global, distributed workforce on inclusive design and accessibility. The subject carried real weight, ethically and in how the bank showed up for its customers. The ask was a video training series. The problem was that the production infrastructure to make one didn't exist: no central studio, no shared equipment, presenters scattered across countries and time zones with whatever gear was on hand. The traditional route was a multi-month, multi-vendor build that would have delayed training the workforce needed then, not in the fall.
I went a different way. Instead of lowering the bar or stretching the timeline, I built a capture workflow that let presenters record on whatever they had, Zoom, a webcam, a phone, and used AI tools to do the unifying in post.
Adobe Podcast cleaned the audio. Cutout Pro standardized backgrounds across recordings shot in dozens of different rooms. I corrected lighting and color across every clip. ElevenLabs cloned voices to replace VO lines that weren't usable in the original take. Synthesia and Descript produced avatars for segments where the on-camera presenter had to be swapped entirely. The finished series looked like it came out of one studio, which is the trick, because no studio existed.
This was the first time AI creative tools went into production at Citi. At a bank, new tooling moves carefully through legal, compliance and information security, and getting it through those gates was as much of the job as the edit.
The AI never did the creative work. The curriculum, the presenters, the subject-matter expertise were all human. AI handled the unification layer, the part that would have eaten months of editing and a six-figure production budget under the old model. That's what made the program shippable on the schedule the business actually had.
What I took from it: new tools don't replace creative authorship. They shorten the distance between an idea and the thing itself. The work is still the work, you just get there faster, with less space between what's possible and what's practical.
