Implement
Ship it, instrument it, keep learning.
Implementation is where products become real, earn trust, and start teaching the team. The discipline is to build with the same care you used to prototype, to instrument so you can see how the thing lives in the wild, and to stay accountable after launch.
Product teams do not hand off. They own. The next iteration starts from what you learned in production, not from where the design left off. The “handoff” boundary that older delivery models taught is often where quality drops, so we close that gap on purpose.
What you do
Build for production with prototype-grade care
The polish that makes a prototype believable is the same polish a shipped product owes its users. Carry it through.
Instrument the success metrics from Define
Without telemetry tied to the outcomes you said you would move, you have shipped a feature, not a product.
Plan for the support cost
Who handles the inbound, who patches the regression, who writes the docs. If those names are blank at launch, you are not done.
Close the loop with a post-launch read
Two weeks in, six weeks in. What did the data say? What surprised the team? Write it down. That is where the next iteration begins.
What you produce
A shipped, accessible, supported product
Working in production, with documentation and the support model named.
Telemetry tied to outcomes
Dashboards that show whether the success metrics from Define are actually moving.
A post-launch read
A short doc summarizing what worked, what surprised, and what the next loop should consider.
AI changes what is buildable. Ownership stays human.
Code generation, deployment automation, documentation drafting, even live-ops agents: much of the production work is now augmented. The team's leverage is no longer about who can write the most code. It is about who can frame the work, review the output, and own the outcome.
What does not change is accountability. When the thing breaks in production, when the user complains, when the metric tanks, those moments still belong to the humans on the team. AI is a force multiplier on capable people. It does not replace ownership.
Watch for
Patterns that can make the work look farther along than it is.
The handoff gap
Treating "design hands to engineering" as a real boundary. In product teams, the work belongs to the team end-to-end.
Launch-and-leave
Shipping the feature and never reading the telemetry. The first six weeks teach more than the previous six months.
Polish debt
Skipping the small things ("we will come back to it") that turn out to be the user’s whole experience.
Missing the support model
Launching without a clear answer to "who fixes this when it breaks." That gap turns into avoidable support pressure.
Ownership. The product is live, the metrics are reading, and the team owns whatever comes next, including the loop back to Empathy when the data tells you the user has changed.