AI for Product Management & Roadmapping Playbook
- Practitioner
- Intermediate
- Template Included
- Workshop Ready
A hands-on playbook for product managers deciding which AI features actually belong on the roadmap, using a four-lens evaluation framework, prioritization templates, and a worked SaaS example.
How is this different from a standard RICE or ICE prioritization framework?
RICE and ICE assume the main uncertainty is impact and effort. AI features carry two additional, often-larger uncertainties — whether the right data exists and what happens when the feature is wrong — that standard frameworks don't surface until you're already mid-build. The 4-Lens framework adds those as first-class scoring criteria.
What if we don't have an ML engineer to help score Feasibility and Data Readiness?
You can run a lightweight version with any technical stakeholder who can answer "does this kind of data exist, and can a reasonably capable model do this task today." The goal at this stage is a directional score, not a technical spec — bring in deeper ML expertise once an idea clears the initial gate.
Should every AI feature start in the assistive tier?
Not always, but it should be the default unless the task is low-stakes and well-understood (e.g., auto-formatting) where automated tier is safe from day one. The assistive-first approach mainly protects you against features with weak Responsibility scores, where a wrong automated action has real cost.
How often should we re-score the backlog?
Quarterly is usually the right cadence — frequent enough to catch improving data readiness or new evidence of desirability, infrequent enough that the scoring stays credible instead of becoming a constant re-litigation of the roadmap.
What do we do with ideas that score well on Desirability but poorly on Data Readiness?
Don't kill them — convert them into a data-prep task. A short, scoped effort to label or collect the missing data can turn a 2 into a 4 within a quarter, and those converted ideas often become the strongest releases because the underlying user problem was already validated.
Subscriber access
Unlock this playbook
This playbook — including every framework, template, and step-by-step section — is available free to Think Insights subscribers. Enter your email to unlock it instantly and get our weekly insights newsletter. No account needed, and access is remembered on this device.
