4 months of Work in 6 Weeks: Spec-Driven Development on an Agile Team
Software Development Practice Lead
Atomic Object
Kealy Williams is a Practice Lead and Software Consultant at Atomic Object, where she helps shape a culture of collaboration, craftsmanship, and sustainable agile practices. Over 20+ years in software, she's coached developers and championed pair programming, test-driven development, and mentoring as the foundations for building both products and teams. Her work has spanned biosciences, finance, container trucking, survey management, and the court system, each one reinforcing her belief that collaboration is the constant that drives quality. Lately she's been wrestling with how teams adopt AI-assisted practices without lowering the quality bar that earns client trust. Outside of work, she's happiest grilling in the backyard with her husband and their Shih Tzu, Trooper, and she'll never turn down a Star Wars reference.
When the deadline says six weeks and the work says four months, something has to give. For most teams in regulated or enterprise environments, the answer can't be "cut quality." The cost of AI slop, hallucinations, or sloppy code shipping to production is too high.
This session is a case study of an agile team that bet on spec-driven development with frontier coding agents to close that gap on a real client engagement in a security-conscious enterprise environment. They hit code-complete in five weeks, used the sixth for polish, absorbed unplanned scope mid-flight, and shipped with zero showstopper bugs.
The time savings were real, and the data shows where they came from, and where they didn't. Writing code dropped to a median of one day per story card; code review and QA each took three. The bottleneck moved, and that changed what the team did all day, how they collaborated, and what "doing the work" even meant. We'll dig into the cycle-time data, the tests that passed without proving anything, the developer pushback over ownership, and the small process changes that paid off across every plan that followed.
The takeaway isn't that AI writes your code for you. It's that the shape of the work changes, and so does where your team should spend its attention. If you're an engineering leader, tech lead, or practitioner weighing agentic development for real projects, this talk gives you the grounded, real-numbers version.