Sierra AI hit a $15B valuation — and its job board shows the agent era has already split engineering into 20 new specialties

A $15B Bet on Agents as a Permanent Category

Sierra AI raised $950 million on May 4 at a post-money valuation above $15 billion, eight months after a $350 million round set its valuation at $10 billion. Tiger Global and GV led the Series E, with Benchmark, Sequoia, and Greenoaks participating. The company reports $150 million in annual recurring revenue, up from $100 million in late November 2025, a $50 million ARR gain in roughly 10 weeks.

The velocity matters more than the number. Sierra went from zero to that ARR figure in eight quarters, a ramp the company calls unprecedented in traditional software. More than 40% of the Fortune 50 are customers. Those are the metrics investors are underwriting at 100x ARR — a multiple that only makes sense if you believe enterprise AI agents are becoming a permanent software category, not a feature that gets absorbed into existing CRM platforms within two years.

Enterprise spending patterns back up the bet. Uber CTO Praveen Neppalli Naga disclosed at a TechCrunch StrictlyVC event that the company "blew through our budget" after opening the door to agentic AI tools late last year. Across roughly 8,000 engineers, about 10% of all code is now autonomously generated. A hotel-booking integration that would normally take a year was done in six months using agentic workflows. That is the kind of proof point that turns pilot projects into procurement commitments.

Sierra's hiring reflects the same urgency. Zero G Talent's board shows the company added 5 roles in the past 7 days alone, including a Forward Deployed Infrastructure Engineer in London and a People Partner for Europe, signals that the new capital is going toward international expansion and customer-facing technical talent, not just model research. The roles span San Francisco, London, and Portuguese-speaking markets.

Bret Taylor, who co-founded Sierra with Clay Bavor and serves as OpenAI's board chair, framed the round as a land grab in a market he expects to consolidate. "There's just a lot of competition. We are multiples larger than the next biggest and are trying to invest aggressively so that we can continue to expand our lead," he said. He also predicted a market correction within two years, a culling that will wash out companies without real revenue or retention. Sierra is positioning itself to be the platform enterprises lock in before that shakeout arrives.

When a company valued at $15 billion starts deploying capital into forward-deployed infrastructure and regional people operations, it is building the team to support production-scale enterprise contracts, not demos. For engineers watching the agentic AI job market, Sierra's trajectory is a leading indicator of where the money and the roles are concentrating.

Ghostwriter Turns Customer Experience Into Agent Code

Sierra built its business the expensive way: forward-deployed engineers working side by side with each new customer to hand-build AI agents for customer service. Ghostwriter, launched in March 2026, is the company's bet that it no longer has to.

The platform lets non-technical staff describe what they want an agent to do in plain English, then produces a production-ready agent across voice, chat, email, and more than 30 languages. Upload SOPs, support-call transcripts, audio recordings of subject-matter experts, or whiteboard photos. Ghostwriter parses the material, identifies key behaviors and edge cases, and outputs an agent with guardrails already built in. Sierra co-founder Bret Taylor demonstrated the workflow in a public demo, building an agent from scratch using only natural language prompts.

What makes Ghostwriter different from the dozens of agent-builder tools already on the market is what happens after the initial build. Sierra's "Agents as a Service" blog post, published alongside the launch, makes the distinction explicit: the hard part of agent development isn't going from zero to one, it's improving over time. Ghostwriter auto-generates test simulations with every change, diagnoses failures, and implements fixes without manual debugging. A companion tool called Explorer analyzes live customer conversations, surfaces improvement recommendations, and feeds them back into Ghostwriter. Sierra describes the loop as an "agent assembly line": analyze, improve, test, validate, ship, repeat.

Under the hood, the platform required Sierra to rearchitect itself as headless infrastructure. Ghostwriter needs direct access to the full workspace, a sandboxed testing environment, and a coherent action space, which Sierra calls the "agent harness." That rearchitecture is what lets the system build and modify agents autonomously rather than relying on a human to click through menus and form fields.

The business case is straightforward. Sierra, valued at $10 billion after its November 2025 funding round, had relied on engineering-heavy onboarding for every customer. The Information noted the launch lands as large companies shift from testing agentic AI to embedding it in core workflows.

Ghostwriter doesn't eliminate the need for engineers. It shifts where they're needed, from building individual agents to building and maintaining the platform that builds agents. Sierra's own job board reflects that shift: the company is hiring product managers for agent development and forward-deployed infrastructure engineers, roles that sit between the platform and the customer rather than inside each deployment.

What Sierra's Open Roles Reveal About Enterprise-Agent Engineering

Sierra's careers page lists over 120 open roles. That number alone signals scale, but the composition of those roles tells you what "enterprise-agent engineering" actually means, and it's not what most job boards would lead you to expect.

The single largest category is Agent Engineering. Sierra has posted roughly 20 variations of "Software Engineer, Agent" and "Agent Engineer, TLM" across its offices in San Francisco, New York, London, Singapore, Tokyo, Paris, Madrid, Munich, Toronto, and Sydney. Many of these are language-specific: Arabic, Cantonese, Dutch, French, German, Italian, Korean, Spanish, and Thai. Sierra isn't building one agent for one market. It's building agent systems that operate natively in dozens of languages, and it needs engineers who can work in those languages to train, evaluate, and deploy them.

Then there's the platform layer. Under Platform Engineering, Sierra is hiring for Agent Architecture, Agent Builder, Agent Data Platform, Agent Intelligence, and Agent SDK, all San Francisco-based. These are the infrastructure roles: the people building the tooling that lets the agent engineers ship. A Forward Deployed Infrastructure Engineer role spans San Francisco, New York, and London, suggesting Sierra embeds platform staff alongside customer-facing teams rather than centralizing them.

The Product department mirrors this split. Product Managers for Agent Development sit in San Francisco, New York, London, and Tokyo, again with language-specific variants. But Sierra also lists dedicated product roles for Ghostwriter, Agent Studio, Agent SDK, Agent Data Platform, and Voice, each a distinct surface within the platform. Someone owns the roadmap for each one.

What's missing is just as telling. There are no "prompt engineer" titles. No "AI whisperer" roles. The job architecture treats agent development as a software engineering discipline with specializations, not a prompt-crafting exercise. The Agent Engineer, TLM role (TLM meaning "tuning and language model") comes closest to the research side, and it's posted in New York and London, not San Francisco, which suggests Sierra colocates its model-tuning work with specific talent pools.

The Early Career Program, called APX, feeds directly into this structure. Sierra's careers page says APX members "work directly with customers to build AI agents for leading global brands." That's not a rotation program in the traditional sense. It's a compressed path into agent engineering.

The same tracker shows 18 roles added in the past week, with posted salaries ranging from $150,000 to $390,000 for product roles in San Francisco. The Forward Deployed Infrastructure Engineer role in London lists £170,000–£290,000. These are not entry-level numbers, but they reflect a company hiring across seniority levels simultaneously, another sign that the org is building out faster than it can promote from within.

Strip away the department labels and a profile emerges: Sierra wants engineers who can build and ship agent systems in production, product managers who understand both the model layer and the enterprise customer, and infrastructure people who can deploy alongside clients. The agent stack has already split into distinct engineering specialties, and the companies building it need all of them at once.

San Francisco's Concentration of Agent Talent

The agentic-AI hiring surge isn't evenly distributed. It's clustering, and the cluster that matters most right now sits in San Francisco.

Of the five roles Sierra AI added in the past week, three are based in San Francisco: two Product Manager positions for Agent Development (one for flagship deployment, one requiring Brazilian Portuguese) and an RFP Strategy & Operations Analyst. The other two are in London. That's a 60% San Francisco weighting from a company that, at a $15B valuation, is one of the most watched names in enterprise agents.

Anthropic's recent postings tilt the same way. The same tracker shows 18 roles added in the past 7 days, with four listing San Francisco as a primary location, including two Staff Software Engineer positions on the Claude Code developer productivity team and a Research Engineer role in domain scaling. New York and Seattle appear as secondary options on several of these, but San Francisco is consistently the first city named.

This isn't new, but the concentration is deepening. San Francisco has been the default launchpad for AI startups since at least 2023, when OpenAI, Anthropic, and a wave of smaller labs set up headquarters there. What's different now is the type of role. Early AI hiring in the Bay Area skewed toward research scientists and ML infrastructure engineers. The current wave is product-facing: product managers who understand agent orchestration, forward-deployed engineers who embed with enterprise customers, operations analysts who structure the RFP process for agent deployments. These are the roles that turn a model into a product a Fortune 500 company will actually buy.

Other hubs are competing, but none match the density. New York has a growing presence — Anthropic lists it alongside San Francisco on multiple roles, and the city's financial-services sector is a natural buyer of agentic tools. Seattle benefits from its legacy cloud-infrastructure talent pool. London is picking up European-facing roles, as Sierra's London-based postings for a People Partner and Forward Deployed Infrastructure Engineer suggest. But these cities are adding agent roles on the margin. San Francisco is where the core teams are being built.

The salary data reinforces the pull:

Those numbers aren't just competitive; they're a signal that companies are willing to pay a premium to keep agent talent in one metro area, where hiring managers, product leads, and the engineers who build the underlying tooling are all within a short distance of each other.

The risk of this concentration is obvious: a single geography holding too much of a critical talent pipeline creates fragility. But for now, the gravitational pull is winning. If you're an engineer or product manager who wants to work on enterprise agents in 2026, the job board and the money are pointing to San Francisco.

Who Else Is Hiring — and How Sierra Competes

Sierra isn't hiring in a vacuum. The company's push into enterprise-agent engineering runs headlong into a market where Salesforce, Anthropic, and a pack of well-funded startups are all chasing the same narrow band of engineers who can build, deploy, and maintain production-grade AI agents.

Salesforce made its intentions clear at Dreamforce 2025. The company unveiled Agentforce 360 as the backbone of what CEO Marc Benioff calls the "Agentic Enterprise", a model where AI agents work alongside human employees across every business function. More than 12,000 customers are already on the platform, including Reddit, OpenTable, Adecco, and Williams-Sonoma. The pitch is trust and integration depth: Salesforce is pairing its 26-year CRM infrastructure with frontier models from both OpenAI (GPT-5) and Anthropic (Claude), the latter targeted specifically at regulated industries like finance and healthcare.

That dual-partnership strategy matters for the talent war. Salesforce isn't just building agent tooling — it's positioning Agentforce 360 as the connective tissue between enterprise data and multiple model providers. Engineers who understand how to orchestrate agents across that kind of multi-model, compliance-heavy environment are exactly the ones Sierra is also trying to recruit.

Anthropic, meanwhile, is hiring aggressively on its own. Zero G Talent's board shows 18 roles added in the past week, including a Research Engineer for Domain Scaling with a listed range of $350,000–$850,000 per year, and multiple Staff Software Engineer positions in the $405,000–$485,000 band. The company's partnership with Salesforce gives Claude a direct pipeline into enterprise workflows, but Anthropic is also building its own agent-facing tooling, and Claude Code positions suggest the company wants to own the developer-experience layer, not just supply models.

Big Tech still writes the biggest checks, but leaner startups are competing on mission and equity upside. The dynamic is playing out across San Francisco, where Sierra, Anthropic, and Salesforce all list roles, and where the concentration of agent-specific engineering talent is densest. For engineers deciding between a $400,000 staff role at Anthropic and a position at a pre-IPO agent startup, the calculus isn't just compensation. It's whether the smaller company can ship a product before the platform players absorb the entire market.

Sierra's bet is that enterprises will want a dedicated agent layer rather than a CRM add-on. Salesforce's bet is that the CRM is already that layer. The engineers being hired right now will determine which vision wins.

What Engineers Should Learn to Get Hired in Agentic AI

The job postings tell you what companies want. The gap between what's listed and what most engineers actually know is where the opportunity sits.

Sierra's own careers page describes agent engineers as people who "tackle problems at the intersection of software engineering, frontier AI deployment, and the most pressing challenges of our customers." That's not a vague mission statement; it's a skills map. Break it apart and three distinct competency clusters emerge.

Software engineering that survives contact with production. Sierra's job board listing for Software Engineer, Agent says you'll "build and ship AI agents that handle thousands of customer conversations a day." That's not a research prototype. It means writing code that scales, fails gracefully, and integrates with real enterprise systems — CRM platforms, ticketing tools, APIs that were never designed for an LLM to touch them. If your experience stops at Jupyter notebooks, you're not ready for this layer.

LLM orchestration and tooling. The open-source agentic-ai-engineering curriculum on GitHub (agenticloops-ai/agentic-ai-engineering) structures its tutorials around LLM APIs, prompt engineering, tool calling, and the agent loop, the cycle where a model decides to use a tool, executes it, reads the result, and decides what to do next. Coursera's agentic AI engineering course adds workflow orchestration, memory and state management, and multi-agent coordination. These aren't electives. They're the core of the job. A paper on production-quality agentic systems published on arXiv puts it in lifecycle terms: workflow decomposition, multi-agent design patterns, Model Context Protocol (MCP), deterministic orchestration, and environment-aware deployment. Engineers who can move from "the model outputs text" to "the agent reliably completes a multi-step workflow using five tools and hands off to a human when confidence drops" are the ones getting interviews.

Customer-facing deployment skills. This is the part most AI engineering guides leave out. Sierra's early-career page says you'll do exactly that alongside leading global brands. The Exponent interview guide for Sierra's Agent Engineer role confirms the process tests not just technical depth but how you reason about customer problems and translate them into agent behavior. That means understanding a business workflow well enough to decompose it into agent steps and explaining your design choices to a non-technical stakeholder.

The salary data backs up the demand. Sierra's Product Manager, Agent Development roles in San Francisco are listed at $180,000–$390,000 per year. Anthropic's Research Engineer, Domain Scaling roles (adjacent agentic work) range from $350,000 to $850,000. These aren't premiums for PhDs doing novel research. They're premiums for engineers who can ship.

Zero G Talent's board shows Sierra AI added 5 roles in the past week alone, and Anthropic added 18. The hiring velocity is the signal. The skills gap is the opening.

Working in AI? Zero G Talent tracks the openings: browse AI jobs, openings at Anthropic, Sierra Space and Sierra AI, and the people building the field.