See how the agents actually work
Not a demo and not a prompt. Explore what each agent does, the execution logic behind it, the outcomes it drives, and how real revenue teams use it — then step through a live run.
How an agent actually works
Every agent follows the same anatomy. It is not a single prompt — it is a sequence of reasoning, real tool calls, grounded retrieval, decisions, and human checkpoints. Here is the logic behind every run you can step through below.
- 1
Trigger
A webhook, schedule, inbound email, or Slack command kicks off the run with an input.
- 2
Reason
The model plans what to do — which tools to call and in what order — before acting.
- 3
Tool call
The agent takes real actions: querying the CRM, enriching data, drafting, building.
- 4
Retrieval
It grounds itself in your data — knowledge base, docs, warehouse — to avoid guessing.
- 5
Decision
Confidence and policy gates decide whether to proceed, skip, or hand off to a human.
- 6
Approval
Sensitive actions pause for a person to approve, edit, or reject before anything ships.
- 7
Output
It writes results back into HubSpot and your stack, then logs the full run for evals.
6 agents
SDR Agent
Sales / SDR
search_crmenrich_companyweb_researchdraft_emailrequest_approvalcreate_taskHubSpot Forward Sync
Creates the contact, sets lifecycle stage, logs the enrichment + research as notes, drafts the email as a logged engagement, and creates the follow-up task — all in HubSpot.
Scenario: A new demo request lands from dana@northwind.co via the website form.
SDR teams point their "Request a demo" and high-intent forms at the agent. It does the first-touch research and drafting the moment a lead arrives, so reps spend their time on live conversations instead of tab-hopping. Reps approve or tweak each draft from Slack, keeping their voice on every send.
How a run executes
Ready to put an agent to work?
Book a build call and we will scope the highest-ROI agent for your revenue team — and show you exactly how we would ship it.