'Most companies don't want to make software. They want solutions to their problems.' — That's what Bret Taylor, former co-CEO of Salesforce, told the audience at HumanX conference in April 2026. And his company Sierra is turning those words into a product.
What Is This?
Sierra is a startup that builds AI agents for enterprise customer service. Founded in early 2024 by Bret Taylor (former Salesforce co-CEO) and Clay Bavor (former Google executive), the company hit $100M ARR in just 21 months. With 40% of the Fortune 50 already on board as clients, their growth has been nothing short of explosive.
But what Sierra released in March 2026 isn't just another AI chatbot. It's called Ghostwriter — literally 'an agent that builds other agents'. Upload your SOPs, call transcripts, whiteboard photos, or process docs. Or just describe what you want in plain English. Ghostwriter takes it all and autonomously creates a production-ready agent that handles voice, chat, and email across 30+ languages.
Taylor calls this 'Agent as a Service.' The core idea is simple — prompts instead of clicks, conversations instead of menus. Think about how often you actually log into Workday. Most enterprise software is really just a 'complex system you rarely use'. So why should anyone have to learn how to navigate it?
Code to no-code to no clicks. As clicks turn to prompts, building agents will become as simple as describing what you want.
— Sierra Blog, Agents as a Service
What Actually Changes?
The gap between traditional enterprise software and Sierra's approach is fundamental.
| Traditional Approach | Sierra Ghostwriter | |
|---|---|---|
| Agent Building | Developers write code + design flows | Describe in natural language, auto-generated |
| Deployment Speed | Months | 4 weeks (Nordstrom case) |
| Optimization | Manual log analysis + rule tweaking | Explorer analyzes conversations & auto-improves |
| Channels | Chat-only mostly | Voice, chat, email, 30+ languages |
| Pricing Model | Per-seat | Outcome-based pricing |
What's especially worth noting is Explorer. While ChatGPT's Deep Research scours the internet, Explorer analyzes your customer conversation data to find improvement opportunities. The agent learns from real support patterns and proactively suggests upgrades.
The bottom line: Sierra's model is closer to selling outcomes, not software. They charge per resolved outcome rather than per seat, which means customers can measure ROI immediately.
Reality Check
Full autonomy isn't here yet. According to TechCrunch, many AI agent companies including Sierra still rely on 'forward-deployed' engineers who continuously tune agents at customer sites. The fully automated future Ghostwriter promises still has some distance to cover.
Getting Started: The Essentials
If you want to bring AI agents into your organization, here are the key principles from Sierra's playbook.
- Define the problem first
Not 'let's adopt AI agents' but 'let's cut repeat inquiry handling time by 50%.' Sierra's entire philosophy is solutions, not software. - Gather your existing data
The first thing Ghostwriter does is learn from SOPs, call recordings, and FAQs. No data, no agent. - Start narrow
Sierra focused on customer service first. Prove value in one use case, then expand. That's how you earn enterprise trust. - Set outcome-based metrics
Measure resolution rate, response time, and CSAT — not seat count. Sierra's pricing model embodies this philosophy. - Redeploy people, don't remove them
Even Sierra needs forward-deployed engineers. AI handles 80%, but the remaining 20% of complex cases and quality management still requires humans.
