The sales rep hadn't even opened the stalled-deal alert yet. The AI had already written the follow-up email. Nobody asked it to.
That's what Writer's new event-driven agents do. No prompt required — they detect business events on their own and execute the next step automatically.
What does "no prompt" actually mean?
Until now, AI was reactive. You had to ask first, then it answered. You had to initiate, then it acted. ChatGPT, Copilot, Claude — all the same. The foundational assumption of AI assistants was: humans go first.
Writer broke that assumption. Its new agents, launched in late April 2026, directly detect business events and autonomously run pre-defined workflows called Playbooks. An email arrives, a meeting ends, a Slack message gets posted, a file lands in Drive — each of these events becomes a trigger.
Writer was founded in 2020 and just closed a $200M Series C. Uber, L'Oréal, Accenture, and Mars are already customers. The kicker: they're not running on OpenAI or Anthropic — they use Writer's in-house Palmyra LLM, tuned specifically for enterprise work, ranging from 128M to 43B parameters.
What's a Playbook Builder?
A no-code builder that lets non-technical employees create agent workflows. Describe what you want in plain language — "If a sales deal stalls for 7+ days, draft a follow-up email and notify the sales team" — and the AI converts it into an executable Playbook automatically.
What actually changes, though?
The gap is clearest in a side-by-side. Every AI tool you've used is reactive. Writer's agents are proactive. The core shift: they don't wait for human intent.
| Traditional AI assistant | Writer event-driven agent | |
|---|---|---|
| How it works | Human types prompt → AI responds | Event detected → agent acts autonomously |
| Human involvement | Required every time | Zero after trigger fires |
| Timing | When someone remembers to ask | Immediately when event occurs |
| Multi-step | Single response (manual chaining) | Automated chain via Playbook |
| Governance | No logs / no audit trail | Datadog observability plugin |
Triggerable events currently include: Gmail email received, meeting start/end (Google Calendar), Gong sales call completed, Slack message posted, Google Drive file uploaded, and Microsoft SharePoint changes.
The competitive angle is worth watching. AWS, Microsoft, and Salesforce are all expanding their own agentic platforms. Writer's play is to be the vendor-neutral, full-stack alternative — an enterprise AI layer that sits on top of whatever stack you already have.
Forrester pegged the average ROI for Writer customers at 333%, with a 6-month payback period. The enterprise AI agent software market is forecast to grow from $1.5B in 2025 to $41.8B by 2030.
How to actually get started
- Get on Writer Enterprise
Request a demo at writer.com. SOC 2, HIPAA, and PCI compliance are all covered — check which certifications your industry requires before the sales call. - Connect your highest-volume tools first
Pick the one or two connectors your team uses most — Gmail, Slack, or Gong are good starting points. Each connector monitors specific event types (received, ended, posted, etc.). - Write your first Playbook
Be specific: "When a Gong sales call ends → summarize the call → update the customer profile → post a briefing to the sales Slack channel." More specificity = more accurate execution. - Use the Skills library
Playbooks are built from Skills — reusable AI task blocks. Start with Writer's built-in Skills (doc summarization, email drafting, report generation) before building custom ones. - Set up governance via Datadog
Every autonomous action should be traceable. The Datadog plugin shows what event fired, what Playbook ran, and what was produced. For high-stakes actions (external emails, data changes), add a human approval step to the Playbook first.
