Solo product team with AI agents — one-developer case study

Cassidy Hilton’s article is one of the more compelling pieces of evidence for how AI agents are changing the economics of product development — not by making large teams faster, but by making solo practitioners viable.

The setup

Hilton ran 43 sprints as a solo developer with two AI agents filling roles that traditionally require separate team members: one agent handled planning and prioritization (acting as a PM/scrum master), while the other assisted with code review and testing. The sprints followed standard agile ceremonies — planning, daily standups (with agents), reviews, and retrospectives.

What the data shows

Over 43 sprints, Hilton tracked velocity, defect rates, and time spent on different activities. The results suggest that a single developer augmented by AI agents can maintain throughput comparable to a small team for certain types of work — particularly well-defined features with clear acceptance criteria.

The retrospective data is especially interesting: the AI agents identified recurring patterns in sprint outcomes that Hilton had missed, leading to process improvements in later sprints.

Why PMs should care

The implications extend beyond solo developers. If AI agents can effectively fill supporting roles in product development, the optimal team size for certain products may shrink. PMs should consider what this means for team composition, hiring plans, and the economics of building new products.

The article also demonstrates that AI agents work best within structured frameworks (like Scrum). The ceremonies provided the rhythm and checkpoints that kept the agents’ contributions useful rather than divergent.

Limitations

Solo-developer projects tend to be smaller in scope than team-built products. The approach may not scale to complex systems with cross-team dependencies, regulatory requirements, or enterprise-grade reliability expectations.

Who should read this

PMs at startups or in innovation labs evaluating lean team structures. Also relevant for PM leaders thinking about how AI agents might complement (not replace) existing team members.