Agentman is a platform that turns expert workflows into AI products. Its core idea is: AI agents do repeatable, high-stakes work, executing like the best member of the team, while keeping every step visible and every decision still made by a human.
Main capabilities and how to use it:
- Agentic Harness: provides AI agents with the context, tools, state, and checks needed to complete tasks. The Harness can support one or more agents and is the operating environment built around agents.
- Agent Skills: capture the team's procedures, decision rules, and exceptions as maintainable, versionable skills, built and shared in the Agentman skills platform and library. Skills are the layer of domain expertise within the platform.
- MCP connectors: native support for 600+ MCP connectors, able to connect to SaaS applications as well as the documents and business systems needed for workflows.
- State retention: preserves case context across steps and sessions, and supports resuming from where it stopped after an interruption.
- Evaluations and checks: before a workflow is updated, evaluate agents with test cases and review the results.
- Human review: set review points and route exceptions to the people accountable for the outcome.
- Permissions and evidence: manage access to workspaces, skills, and connected systems, inspect the information behind agent activity and outputs, track skill versions, and review changes before deployment.
Example usage flow (using medical prior authorization as an example):
- A fax is received; the pharmacy says the medication requires prior authorization
- Build the case from the medical record: diagnosis, prior treatments, and lab results
- Save a draft in CoverMyMeds or Availity, with every answer linked to the medical record
- Staff submit from the clinic's own account The case retains its context, and the team is responsible for the final decision.
Applicable scenarios:
- Operations work at independent medical clinics, such as eligibility verification, inbox triage, and prior authorization drafts (the Medman product).
- Select, client-funded enterprise workflows built on the same harness, which can add skills, integrations, and evaluations.
- Debt restructuring work as well as life-critical workflows.
The platform also provides Agent Skills and Agent Builder for teams to build on their own.
Limitations and notes:
- Deployment requirements, data handling, and security documentation need to be discussed with the team.
- Customer-reported results (such as reducing monthly verification calls from 50 hours to 7 hours) come from founding design partners, and results vary by clinic.Clinic may vary.
- Information such as fees and pricing is not provided in the materials.
