Run your CRM on autopilot with an AGNT agent

In the AGNT dashboard, navigate to Integrations and select HubSpot or Salesforce. Authorize via OAuth 2.0 — the agent gets read/write access to Contacts, Deals, Companies, and Activities. On first sync, the agent pulls your pipeline stages, custom properties, existing deals, and the last 90 days of activity history.

The historical data gives the agent context: which deal stages have the highest drop-off, what follow-up cadence your top reps use, which lead sources convert best. The agent uses this to calibrate its own behavior from day one.

# Verify the connection
curl -H "Authorization: Bearer $AGNT_TOKEN" \
  https://api.agntdot.com/v1/integrations/crm/status
# → { "provider": "hubspot", "connected": true, "contacts_synced": 12480, "deals_synced": 342 }

Define your qualification criteria in the AGNT rules editor. The agent scores every inbound lead against these criteria using data from the CRM record, enrichment sources (company size, industry, location), and behavioral signals (pages visited, emails opened, form fields filled).

Leads that meet your threshold are automatically moved to 'Qualified' and assigned to the appropriate rep based on your routing rules (round-robin, territory, deal size). Leads below threshold go to a nurture sequence. The agent writes its reasoning into a CRM note on the contact record so reps see why a lead was qualified or deferred.

This runs in real time — a lead fills out a form, and within seconds the agent has scored it, enriched it, assigned it, and sent the assigned rep a notification with context.

Configure follow-up sequences by deal stage. For each stage, define: the follow-up interval (e.g., 3 days after last activity), the channel (email, WhatsApp, phone task), the message template, and the escalation rule if there is no response after N attempts.

The agent executes these sequences autonomously. It sends the follow-up, logs the activity in the CRM, and watches for a response. If the prospect replies, the agent updates the deal stage, adjusts the next follow-up, and notifies the assigned rep. If the sequence exhausts without a response, the agent moves the deal to 'Stalled' and optionally archives it.

Unlike static email sequences, the agent adapts: if a prospect opens an email but does not reply, it might switch to WhatsApp. If a prospect replies with a question, the agent answers it (using the knowledge pack) and resets the sequence timer.

Beyond the initial sale, the agent manages post-sale touchpoints: onboarding check-ins, renewal reminders, upsell opportunities based on usage patterns, and churn risk detection. Configure lifecycle stages and the agent moves customers through them based on activity data.

For subscription businesses, the agent monitors usage metrics (pulled via API or webhook) and flags accounts showing churn signals: declining usage, support ticket spikes, payment failures. These accounts get proactive outreach before the renewal date.

Every lifecycle action is logged in the CRM with full context, so any human who picks up the account has complete history.

The agent generates pipeline reports on schedule or on demand. Configure a daily pipeline snapshot for the sales manager, a weekly forecast for leadership, or ask ad hoc: 'What deals are closing this month and what is the weighted forecast?'

Reports include: deals by stage with aging, conversion rates stage-to-stage, rep performance comparisons, forecast accuracy (predicted vs actual), and pipeline velocity trends. The agent highlights anomalies — deals stuck longer than average, stages with declining conversion rates, reps with unusual patterns.

Reports are delivered via your preferred channel. The agent does not just dump numbers — it provides a one-paragraph executive summary with the three things that need attention this week.

Not everything should be automated. Configure escalation triggers: deal value above a threshold, enterprise-tier prospects, complaints, legal mentions, competitor mentions, or any custom keyword. When an escalation triggers, the agent pauses its autonomous actions on that record and routes it to a human with full context.

The escalation notification includes: contact details, deal history, conversation history, the agent's recommended action, and the reason for escalation. The human resolves the situation, marks the escalation closed, and the agent resumes normal operations.

Track escalation volume and resolution time in the AGNT dashboard. A decreasing escalation rate over time means the agent is learning your edge cases; an increasing rate means your rules need refinement.