The work it replaces
Before a good first call, someone usually opens the company website, a few news results, a professional network and the CRM history, then pastes what they found into a notes field. Done properly it is valuable. Done for every new lead, it quietly eats the hours a team meant to spend selling.
A research agent does that groundwork on the record and leaves the judgement to you.
The six steps of a run
In FlowCRM a research run follows the same visible steps every time:
- Research the company: read public company context for the record.
- Find decision makers: propose the people most likely to be involved, added as contacts.
- Analyse context: weigh the deal stage, history and recent activity.
- Summarise the opportunity: write a short, reviewable summary onto the record.
- Update the CRM: fill in only the fields the workflow allows, with every change logged.
- Recommend the next action: suggest a task. A person decides whether to act.
Where a person stays in control
The agent drafts, a person decides. Recommendations arrive as tasks you approve or edit, nothing is sent to a prospect without someone choosing to send it, and the run log shows each step so you can check the work before relying on it.
How runs are counted
One completed task is one agent run. Plans include 100 runs per seat on Starter, 300 runs per seat on Team, 600 runs per seat on Business, 1,000 runs per seat on Enterprise, pooled across the company, so a busy rep can draw on runs a quieter colleague has not used. How pooled runs work covers what happens at the limit.
Beyond the built-in agent
Every plan includes the research agent on every record and a no-code agent builder, so you can turn your own qualification checklist into steps that run the same way on every account. Agents can also run inside an automation, for example whenever a deal reaches Qualified.