How Flight Story built 48 AI tools and saved 62,000 hours

Executive overview

Most media companies fear AI will replace their teams. Flight Story ran a company-wide competition to build with it instead. The result: 48 tools, 62,000 hours saved, and a structural shift in how they make editorial decisions.

Two tools stand out: Unicorn Hunter, which monitors live cultural trends to identify the right guest at the right moment, and pre-watch, which tests long-form episodes with thousands of viewers before release to remove human bias from editing.

Stop guessing — use data to make confident, high-impact decisions on guests, content, and strategy.

The $20,000 AI agent competition

• Stephen Bartlett framed AI as a signal to lean in, not avoid — cognitive dissonance around new technology is a prompt to engage
• The competition asked every team: what could you automate, redesign, or reimagine with AI as a co-partner?
• Teams worked cross-functionally with people they hadn't collaborated with before
• Fear of being replaced gave way quickly to recognition that directing AI tools is the most valuable skill
• The experience of breakthrough — making visible progress together — was a powerful motivator
• Independent judges scored entries on efficiency impact, social impact, and usability
• Result: 48 tools across all departments, 62,000 hours saved, ~$1M cost impact

Unicorn Hunter and the guest radar system

• Guest selection is the highest human-effort, highest-payoff decision in the business
• Guest Radar identifies when a person appears on a channel and the content overperforms — surfacing who drives outsized results
• Example: Dr. Eric Weinstein delivered 11x standard channel performance; the team booked him and replicated that result on Diary of a CEO
• Unicorn Hunter adds a real-time cultural trends layer — monitoring what topics and voices are surging across the internet
• This lets the booking team act immediately: "We can see you're talking about this right now — can you be in LA in two days?"
• The tool supports fast-turnaround debate formats on topics like geopolitics, feminism, and young men's societal struggles
• Guest relationships often take years to build; some episodes require two to three years of sustained outreach before recording

The pre-watch testing system

• YouTube data arrives after publishing — all editorial decisions have already been made by then
• Pre-watch lets 5,000–7,000 people watch raw episodes before release, using eye-tracking to detect where attention drops
• Viewers can also comment in real time, generating an interest curve across the full episode length
• High-interest moments at the middle or end of an episode inform trailer selection and episode packaging
• Demographic breakdowns reveal which audience segments care about which segments — removing the editor's personal bias
• The tool has expanded into brand partnerships: A/B testing creative executions to give clients data on which version performs

Building audience beyond the show

• Data review frequency matters — knowing your numbers weekly, not just monthly, sharpens editorial instincts
• Community is the multiplier on audience reach: attention that is nurtured becomes loyalty
• The three questions your audience repeatedly asks you are your three product ideas
• Showing up on social, engaging directly, and facilitating audience-to-audience connection compounds the value of every episode