HeiGin

AI agents

Don't just track the work. Move it forward.

HeiGin's agents read customer context, notice what needs attention, and help your team act, inside limits you control.

The operating loop

How an agent works, step by step.

The loop runs continuously. Where a step requires human approval, the agent waits.

Learning from updated context is part of the design; how much adaptation is enabled is configurable and confirmed with each pilot team.

The agents

Specific jobs, not vague AI.

Each agent has a defined scope and is available today. Approvals keep you in control of what it sends or changes.

Follow-up agent

Available

Follow-ups get forgotten when the team is busy.

Scenario

After a renewal call, a customer asks for pricing. A day later nothing has been sent. The agent raises it and drafts the reply.

  1. Context it uses

    Meetings, conversations and open questions on each record.

  2. What it does

    Finds open loops and drafts the next message or task.

  3. Human oversight

    Outbound messages always wait for approval by default.

  4. Result in the CRM

    The follow-up is logged on the timeline with who approved it.

Lead qualification agent

Available

New leads come in faster than anyone can sort them.

Scenario

A web enquiry mentions a 40-person team and a contract ending next quarter. The agent summarises it and suggests an owner.

  1. Context it uses

    Form submissions, first messages and company details.

  2. What it does

    Organises incoming information and highlights qualification signals.

  3. Human oversight

    Suggests ownership and priority; reps decide.

  4. Result in the CRM

    Leads arrive with a summary and suggested next step.

Deal assistant

Available

Deals stall quietly before anyone notices.

Scenario

A deal has sat in Proposal for 12 days with no reply. The assistant flags it and suggests a check-in referencing the last call.

  1. Context it uses

    Stage history, last activity and conversation sentiment.

  2. What it does

    Flags stalled opportunities and recommends a next step.

  3. Human oversight

    Recommendations only. No changes to deals without a person.

  4. Result in the CRM

    Risk flags appear on the pipeline with the reason attached.

Customer relationship agent

Available

Existing customers drift without anyone noticing.

Scenario

A long-standing customer hasn't been in touch for two months, unusual for them. The agent puts them on this week's list.

  1. Context it uses

    Engagement patterns across the full customer history.

  2. What it does

    Surfaces customers who may need attention.

  3. Human oversight

    Surfaces insights; outreach stays with your team.

  4. Result in the CRM

    A clear list of accounts to look at this week, with reasons.

Workflow agent

Available

Repeatable processes still need someone to drive every step.

Scenario

When a deal is won, the agent creates onboarding tasks, notifies the account team and schedules the kickoff, pausing before any customer email.

  1. Context it uses

    The workflow you configure and the permissions you grant.

  2. What it does

    Carries out multi-step processes within its permitted scope.

  3. Human oversight

    Steps you mark as sensitive pause for approval.

  4. Result in the CRM

    Each step and outcome is recorded in the activity log.

Workflow builder: in design

A visual builder for multi-step agent workflows is being designed with pilot teams: pick a trigger, define the steps, and mark which ones pause for approval. We'll publish an interactive preview here once it matches what the product actually does.

Help shape it as a pilot team