STATION
Starting with ElevenLabs, Python, and AI Agents
ALBUM
How audio-ad production went from $500 a spot to about a nickel.
Stephen Laphen ·2025 ·5 songs
When Stephen took over the channel, an audio ad cost roughly $500 and moved at the speed of traditional production: external studios, cost-engineering requests, and long turnaround times for even simple variations. Creative went stale because refreshing it was expensive and slow.
He rebuilt the production pipeline around AI. Using ElevenLabs for voice and lightweight Python to assemble, version, and package the assets, the cost per ad dropped to approximately $0.05 — a ten-thousand-fold change in production economics.
The point was never just cost. Cheap, fast production means creative can be produced quickly, refreshed before it goes stale, tested readily, and scaled without proportional increases in budget or headcount. The team built more creative in a single quarter than it had produced in the entire prior year.
ALBUM
LLMs and lightweight code, pointed at every repetitive workflow.
Stephen Laphen ·2025 ·5 songs
Automation has become the defining thread of Stephen’s Amazon Music work. Using LLMs and lightweight "vibe coding," he builds working tools himself — no dedicated engineering support, no ticket queue — and points them at whatever repetitive workflow is slowing the team down.
Each automation takes something people were recreating by hand — trafficking steps, reporting, formatting, copying between systems — and turns a one-off process into a reusable system. Together they have eliminated more than 500 hours of manual employee work.
The effect compounds: shorter production cycles, avoided engineering costs, and more output without equivalent headcount growth. Stephen calls automation and technology his greatest professional super skill, and this album is why.
ALBUM
Five years of experiments, centralized and conversational.
Stephen Laphen ·2024 ·5 songs
The problem was one Stephen had first identified at Meta: experiments were scattered, institutional memory was fragile, and teams risked duplicating work because nobody could easily find what had already been tried or what it taught.
At Amazon Music he built the fix — a centralized repository holding roughly five years of experiments, campaigns, results, and learnings, stored in SharePoint. Then he connected an AI agent on top of it, so instead of digging through folders, anyone can simply ask: What campaigns have launched? Has anything like this been tested before? What did we learn?
The repository closes a loop. Experiments are documented, learnings enter the archive, the agent retrieves them on demand, and new experiments get designed on historical evidence instead of guesswork. Knowledge stops living in individual memories and starts living in a system.
ALBUM
Owning the promotional surfaces inside Amazon Music itself.
Stephen Laphen ·2024 ·6 songs
Stephen joined Amazon Music in September 2024 on the Global Promotions team, owning the Family Plan through the holiday promotional cycle. He moved quickly from promotions into ownership of first-party audio advertising — the ads Amazon Music runs on its own surfaces.
As the program expanded from audio into video, the channel was restructured as first-party media ads, with Stephen as its owner across fifteen marketplaces. He introduced entirely new formats along the way, including playlist ads — not a refresh of existing creative, but a new way for Amazon Music to communicate with customers through its own media.
He also rebuilt the evergreen advertising system. The old production process created delays and let creative sit stale; the new refresh cycle means evergreen ads are produced faster, refreshed more frequently, tested more readily, and scaled without proportional cost. This channel is what every other album on this page feeds.
ALBUM
Live from the cross-org AI Martech Summit — organizer and MC.
Stephen Laphen ·2026 ·5 songs
Stephen was one of the organizers of the cross-org AI Martech Summit — and the MC, hosting across all three days. Part programming, part performance: keeping sessions moving, introducing speakers, and connecting the threads between talks for an audience drawn from across organizations.
The summit is the visible end of a longer campaign. Stephen has been a champion for GenAI adoption across the marketing org — demonstrating what the tools can do, sharing the systems he builds, and making the case that AI-enabled workflows are how marketing teams scale from here.
ALBUM
From doing the work to building the systems that do the work.
Stephen Laphen ·2026 ·3 songs
Amazon Music brings together nearly every prior stage of the career: the zero-to-one mindset from Produce Mate, the channel-building from Tony’s, the tracking discipline from Paysafe, the performance transformation from Stillwater, the enterprise coordination from Meta, and the department-building speed from Castle.
The distinction now is leverage. Earlier roles meant building websites, brands, content calendars, and pipelines directly. This role means building automated production systems, AI-enabled workflows, knowledge repositories, and queryable institutional memory — tools that let many people produce better work, faster, with better access to what the organization already knows.