Don't argue with AI skeptics. Recruit them.

Allison Esposito Medina, Founder, Willa · Published October 21, 2026

Anny Tritchler joined Lytx in June to help shape program management for AI-native delivery. The tools were fun to build. The people taught her more.

Anny Tritchler started as a Senior Technical Program Manager at Lytx in June. Part of the job is figuring out what program management looks like inside an AI-native development lifecycle. She has been building tooling for it ever since. But the story she tells about her first months isn't about the tools. It's about one skeptic.

When she arrived, a senior TPM peer was openly doubtful of AI. Program management runs on human judgment, he said, and Anny agrees he was right. But the rote work around that judgment had become background noise, and he'd underestimated how much time and energy it was costing him.

So she didn't argue. She brought him into her AI-native redesign sessions as the process expert. He knew how Lytx actually runs, and she needed that to get her tooling right. Then she kept the ask small: just try it. They started with the work that ate his week, tracking progress and risk across a large portfolio, and she gave him starter prompts to adapt to how he works.

Within a month he'd built his first agentic workflow. By month two he was expanding it. His first results came back faster and about as good as his manual work. Once he pointed the agent at more of his source material, it outperformed what he'd been doing by hand. It now adds to his risk log from transcripts and chats, drafts update requests to engineering managers, and writes their answers back into his executive update. Nothing goes out until he approves it. It all started in Claude, where he was most comfortable. When he hit context and cost limits, they migrated the full process to Copilot, so his agents live where the data already is.

By his own estimate, he gets 8 to 10 hours a week back. Now he demos it to peers and managers.

People push back on AI when it's done to them. Start from their pain, put them in charge of the experiment, and they'll take it further than you would.

What she'd do differently

Go department-wide sooner. She coached him one-on-one because he was vocal about his doubts and wanted her to challenge them, but that left others out of the loop. Their 22-person program management office got blanket training instead. The hands-on coaching toward a mindset shift is what got real results, faster. That's the part she'd scale.

Meanwhile, the tooling continues. She's building delivery-intelligence systems in Copilot Studio, Power Automate and Power Platform, with connectors, MCP servers and APIs pulling from Lytx's systems of record into a Power BI dashboard that's in active use. Her rule: if a step is deterministic, it gets plain automation. AI only gets the steps that need judgment. Long term, she wants a digital twin for delivery, like the one she helped build the business case for back in manufacturing. Model the delivery history, then ask what happens to a milestone when a dependency slips, before it slips.

Join Willa, the community where women learn AI together.

All member stories