Video Production Pipeline for IBM's Global Sales Enablement Conference
Replaced a series of manual processes for an eight-day livestream with a unified system for video recording, editing, review, and delivery.
“Paul is an absolute powerhouse who gets the job done. His systems created clear, fast processes for my team, and we would not have been able to meet our deadlines without his work.”
Summary
An internal production team at IBM contacted me about SKO1 2026, their global Q1 sales kickoff conference, featuring hundreds of sessions and 298 speakers.
They had 45 days to film, edit, review, and approve 224 pre-recorded videos, with over 25 crews working across 5 continents. Given the scale of the eight-day livestream conference (viewed by tens of thousands of employees and key business partners), and the compressed timeline, the team needed a solution implemented within ten days of our initial conversation.
I conducted a two-day audit, and delivered a solution that unified the disparate processes into a clear pipeline, fully automating as much as possible.
With my system in place, the project avoided a significant last-minute increase in headcount, saved over 700 hours of manual work, and ensured a timely delivery of all assets.
Problem
When I began, the internal team was already using these tools:
However, their process had many manual steps. Manual folder creation, manual downloads, manual file shares, manual uploads for review, manual “hey where’s that link for the draft again?”
Multiply that by 224 recordings and 542 drafts, and you get a process filled with busy work and room for error.
Since many team members were contractors with little time for onboarding, every manual step became a training burden: more procedures to learn, more opportunities for mistakes.
Solution
I spent two days auditing the existing workflow and proposed a unified system that allowed the different tools to seamlessly share data with one another, and identified the highest-impact places for full automation.
The core concept: everything is triggered by status changes in Airtable, the central database we were already using.
Team members use Airtable to communicate with the various tools, and the tools respond back via Airtable and Slack. The system fills in the rest — Box folder creation, Frame.io uploads and link sharing, metadata like thumbnails and duration, caption file orders and downloads, and, finally, links to the uploaded final videos on the platform. No more copy-pasting links or manually updating important fields.
At every step, the session’s Airtable record automatically reflects its actual status — which sessions have been recorded, what version is in review, and links to all the latest files.
Deployment
The system was rolled out in a few stages, starting with the most time-sensitive features.
Whenever major pieces went live, I gave short tutorials to relevant team members, provided documentation, and gathered feedback. I also delivered detailed weekly progress reports to the team manager, pivoting as needed to accommodate the conference's inevitable changes.
“Paul is our secret weapon. He flagged what needed fixing, involved us in big decisions, and built a system specific to how we already worked, faster than we thought possible. It kept our team focused, reduced costs, and eliminated a ton of busy work. And when a snowstorm doubled our scope, his system handled it.”
Tools Used
The entire system runs on a lightweight cloud server (DigitalOcean droplet), except for the After Effects rendering, which runs on a Mac Mini in the production office.
Every operation is safe to retry, the server auto-restarts on failure, and updates to the codebase take seconds to deploy. A Claude-powered agent reviews logs every 30 minutes, reporting meaningful issues to administrators via Telegram, allowing them to investigate further by chatting with the agent.
Though much of the codebase was built specifically for this project, its modular design means any feature can be repurposed for future projects, with some customization simple enough for non-technical staff.
Reduced Costs
Because the system automated uploads and link sharing, it dramatically reduced the number of paid accounts needed:
Server — a single $5/month DigitalOcean droplet
Box — only key producers and editors needed accounts
Frame.io — one account for the entire project, since uploads from Box to Frame were automated and reviewers used share links
Render Server — a Mac Mini in the production office, securely accessed via Tailscale + Cloudflare Tunnel
The expensive part of this project was never the tools, but the hours that would’ve been lost to manual work, to say nothing of the existential risk of the show not being delivered on time.
I really enjoyed building this system, and, as you can probably guess, learned a ton about each tool’s strengths and weaknesses.
Having begun my career on the other side of the fence as a director, producer, and editor, it’s particularly fulfilling to build these systems because I know all too well the problems they solve.
Please reach out if anything like this might be useful for your team, or check out some of my other projects for more examples.