Managing tasks between developers and AI agents in small teams

Small development teams are increasingly using AI coding tools, but introducing agents creates a new project-management question: how do you track work that is being performed by both humans and AI?

A useful workflow should cover:

1. A task with clear requirements and acceptance criteria.

2. Assignment to either a developer or an AI agent.

3. Visible execution progress and blockers.

4. A record of the output.

5. Human review before the work is accepted.

This is the approach behind tools such as Sharkly.ai. It provides a shared work system where people and AI agents can be assigned to tasks, with progress and results attached to the same work item.

Its free plan currently covers organizations of up to five people. Model execution is connected to the team's existing coding plan or API account, so the platform itself does not provide free model tokens.

For teams already using Jira, Sharkly.ai also offers Jira import and two-way sync.

The interesting question is whether teams should keep their existing project-management system and add an agent coordination layer, or move toward a workspace designed around both human and agent assignments.

How are you managing this in your own development workflow?