Builders are Evolving: Lessons from the AI Frontlines

Jessica Hall
Oct 2
3 min read

AI urgency, and economic uncertainty are making teams and leaders adapt fast. This talk for Agile New England explores how AI is reshaping not just our tools, but our teams, workflows, and what it means to lead. Thank you for having me.

What’s Changing

The macro environment is shifting. High interest rates, layoffs, and restructuring are putting pressure on teams to do more with less.Inside organizations, AI is becoming a teammate, not just a tool.

Roles are evolving, and what’s valuable is changing. It’s not just about building, but about defining, validating, and collaborating with machines.

The “ICEO” (AI-powered CEO) is emerging. Leaders are becoming more hands-on, often diving into the technical details to make sure AI work gets done right.

What Makes a Good AI Use Case

A strong use case delivers real value to the business and customer.

It uses unique data, starts with a measurable baseline, and solves problems that can only be addressed with AI.

If traditional software can do the job, or if the task requires absolute precision, it’s probably not a good fit.

How We Build at OpsCanvas

We use AI to eliminate cloud waste, improve visibility, and orchestrate smarter infrastructure decisions. Here’s what we’ve learned:

AI agents are powerful but need scoped, precise prompts
Productivity can jump 2–3x, but validation still takes time
Small, clear requests work better than broad prompts
Human-in-the-loop review is essential

The New Must Have for a PM: Evals

Product managers now need to be skilled in evaluating AI performance.

This means measuring outputs for bias, accuracy, and value. Monitoring and feedback loops are essential to ensure AI remains consistent, safe, and useful over time.

The AI Builder Playbook

Start with strategy. Define value clearly for both the business and the customer.
Validate the use case.
Make sure you have the data, a measurable baseline, and a clear picture of what success looks like.
Prototype and test. Use scrappy methods like mockups or light models to test quickly and learn fast.
Always keep humans in the loop and design for learning, not perfection.