Medium: Integrating AI into your workflow as a product manager
What the article is about
Seyifunmi Olafioye published this piece in February 2026 in Medium’s Bootcamp publication, a product and design community. The article lays out a staged model for PM AI adoption — not a list of tools, but a framework for thinking about where a practitioner sits in their AI usage and what the next step looks like. It covers seven specific PM use cases and names five common mistakes that slow or block effective implementation.
Context
The piece is written for PMs who have some AI familiarity but have not yet developed a systematic approach. The author makes a clear distinction between experimenting with AI in isolation and embedding it into repeatable work processes. The framing throughout is practical: which problems AI is suited to, which it is not, and how to avoid the most common failure modes.
The four adoption stages
Olafioye structures AI adoption for PMs across four stages:
Experimenter — Using AI ad hoc, mostly for writing tasks, without consistent application or defined workflows.
Optimizer — Applying AI regularly to established tasks, measuring impact, and starting to replace manual steps with AI-assisted ones.
Integrator — Embedding AI into team processes so that colleagues can benefit, not just the individual PM.
Strategist — Shaping how the organization thinks about AI in product work, selecting tools systematically, and building institutional capability.
The model is useful because it reframes AI adoption as a skill that develops over time rather than a binary switch.
The seven use cases
The article maps AI application across seven PM domains: customer feedback analysis, competitive intelligence, opportunity identification, stakeholder communication, product documentation, data analysis, and prototyping. Each is described at multiple levels of automation, from manual AI-assisted tasks to autonomous agent-driven workflows.
Two points stand out. First, the author notes that AI “operates on patterns, not principles” — it cannot replicate strategic thinking, user empathy, or the judgment required to work through organizational politics. Second, the strongest advice in the article is to start with a defined business problem, not a tool: “AI becomes valuable when applied to a defined need.”
The five mistakes
The article names five implementation errors worth avoiding: handling sensitive customer data carelessly through AI tools, treating AI as a search engine, automating workflows built on poor data, automating before validating that the manual workflow is actually effective, and building AI solutions before confirming that a genuine problem exists.
Who it is useful for
Product managers who have tried AI tools but feel their usage is inconsistent or unstructured. The four-stage model gives a clear self-assessment framework. The seven use cases provide concrete starting points for each stage. The mistakes section is particularly useful for PMs who have encountered resistance from colleagues when proposing AI-assisted workflows, as it addresses the root causes of failed attempts.