Triggered sends
Fires the right email on the right event: sign-up, purchase, booking, stage change: with live data merged in.
Task tier · Inbox
An AI agent that handles the email your workflows generate.
Order confirmations, onboarding sequences, follow-ups, reminders, it drafts each one in context, personalises it, sends on the right trigger, and chases when there's no reply. The repetitive sending layer, automated and on-brand.
ModelFast Workers AI models in your own Cloudflare account, stronger model for hard cases
Cost per taskModel usage billed at cost, measured per task in your pilot
Connects toGmailOutlookHubSpotKlaviyoSendGrid
Autonomy
- Suggests
- Acts with approval
- Acts on its own
Sends approved sequences on its own and stops on replies. New templates go through approval first. How autonomy works
What does an AI email automation agent do?
An email automation agent owns the outbound and transactional email a business sends on autopilot. It drafts context-aware messages from templates and live data, personalises them, sends on triggers (a sign-up, a stage change, a schedule), and runs follow-up sequences until a reply or a stop condition. It's the sending layer. Task tier, running on Workers AI in your own Cloudflare account. This is distinct from inbox triage, for sorting and replying to incoming mail, see Lyra, the email triage agent.
theagency47 · Updated June 2026 The spec sheet
How the Email Automation Agent actually works.
| Job-to-be-done | Draft, personalise, send, and follow up on outbound/transactional email from triggers and templates |
| Tier | Task (bounded, high volume) |
| Underlying model | Fast Workers AI models in your own Cloudflare account, stronger model for harder cases |
| Trigger | Event-driven (sign-up, purchase, stage change) + scheduled sequences |
| Inputs | Templates, brand voice, trigger/event source, contact data, send rules |
| Tools available | Email send (ESP / Gmail), CRM read, event webhooks, suppression-list check |
| Autonomous decisions | Personalisation, send timing, sequence branching, stop conditions |
| Escalation rules | Reply needing judgement → route to human / triage agent · Bounce/complaint spike → pause & notify · Off-policy content → block + flag |
| KPIs measured | Emails sent, open/reply/click rates, follow-up coverage, opt-out rate, deliverability |
| Eval suite | 18 test cases (personalisation correctness, trigger logic, suppression compliance). |
What it does
The work it takes off your team.
Context-aware drafts
Personalises beyond {{first_name}}: it writes from the actual context of the contact and the moment.
Follow-up sequences
Runs multi-step follow-ups that stop the instant someone replies or converts: no awkward "just checking in" after a yes.
On-brand
Every message matches your voice and templates, so automated never reads as robotic.
Deliverability-safe
Respects suppression lists, rate limits, and opt-outs, and pauses on bounce or complaint spikes.
Hands off replies
When a reply needs a human, it routes it: to you or to an inbox triage agent: instead of auto-answering blind.
What it looks like in practice
A typical run.
- On every event, Sign-ups, purchases, and bookings each fire their contextual email instantly.
- Continuously, Active sequences advance or stop based on replies and conversions.
- On reply, Anything needing a human is routed; the sequence pauses for that contact.
- Daily, Digest: volume, engagement, opt-outs, deliverability health.
FAQ
Questions about the Email Automation Agent.
How is this different from Mailchimp or Klaviyo?
Those platforms send the templates you build and personalise manually. This agent writes context-aware content per message and decides follow-up logic. It can run on top of your ESP (it drafts, the platform sends) or standalone.
Is it the same as the email triage agent?
No, opposite direction. This handles outbound and transactional sending. Lyra handles your incoming inbox: sorting, drafting replies, surfacing what needs you. They pair well and don't compete for the same job.
Will it hurt my deliverability?
Configured correctly, no. It honours suppression lists, rate limits, warm-up, and opt-outs, and pauses on bounce/complaint spikes. Good content plus good hygiene is the whole deliverability game.
How is it delivered?
As a Spark build wired to your ESP, triggers, and templates.
Related
Agents and engagements that pair with this.
Related: Meet Lyra (triage) →
