SaaStr AI 2026: how five companies ship agents that carry quota

Key Takeaways

• Use LLMs to generate plans but keep execution deterministic—Rubrik's pattern for production-safe agents
• 50% of website visitors now come from answer engines, making agent-led discovery critical for B2B visibility
• Vendors are accountable for what customers do with agents, not just whether the core product works

At SaaStr AI 2026 in San Mateo, the main stage ran end to end on agents. Not chatbots. Not copilots. Agents in production, carrying quota, writing to systems of record. Salesforce, Snowflake, Databricks, Harvey, and Lovable shared playbooks that separated their shipping products from the demo-ware flooding the market. The through-line: agents that touch production systems need deterministic execution, and the vendor stays accountable even when customers run workflows nobody anticipated.

Why CPOs went from the cushiest job to the most stressful

Jason Lemkin opened by arguing that chief product officer went from the easiest seat in B2B to the hardest in about 18 months. For a decade, the job was deciding what to push to next quarter. Now every CPO is under pressure to ship agentic features that customers will actually pay for.

The panel brought together Emrecan Dogan from Glean, Anneka Gupta from Rubrik, Rachel Wolan from Webflow, and Anique Drumright from Harvey. Each company is shipping agents, but their approaches diverge based on what failure looks like in their domain.

Rubrik's split: probabilistic planning, deterministic execution

Rubrik's agent, Ruby, started as a simple RAG application over documentation. It is now agentic, handling tasks like forward-looking capacity planning that used to take customers a full day. The constraint in cyber recovery: nothing can take the core service down.

Rubrik's solution is to use LLMs to generate plans but keep recovery execution deterministic and explainable. The LLM proposes; the system validates and executes with guardrails. This pattern, probabilistic planning with deterministic action, showed up repeatedly across sessions. It is the emerging standard for any agent touching production systems where a wrong move loses the customer.

Answer engines now drive 50% of visitors

Rachel Wolan described Webflow as an agentic web marketing platform that started as a website builder. The new launch is an answer engine optimization agent that detects when you need technical AEO upgrades, analyzes competitor content, drafts changes, and runs autonomously with a human approving each step.

Her framing: answer engines now drive about 50% of website visitors, up from 10% a year ago. A brand that does not appear in answer engine results is invisible to buyers. Webflow holds roughly 18% of the top 2,000 websites, so they have the data to back this claim. For teams still focused purely on traditional SEO, tools like Semrush or Ahrefs remain useful for keyword research, but the optimization target is shifting toward structured content that agents can parse.

Glean's MCP server strategy

Glean's Emrecan Dogan made the case that retrieving information was the original job, but the world is moving from getting informed to getting AI to perform. Glean now operates two ways: as the AI assistant front door, and as a single MCP server feeding superior company context into tools like Claude Code, Cursor, and Codex.

The interesting wrinkle: the more usage happens in external coding harnesses, the more it pulls users back into Glean's own surfaces. External adoption creates a spillover effect rather than cannibalizing Glean's core product.

Harvey's legal agents and the forward-deployed model

Harvey's Anique Drumright described an entire platform that is agentic, not a single agent feature bolted on. Litigators now ask how to coach associates on using AI safely and how to verify an agent's plan, not just its output.

Harvey deploys legal engineers, most of whom practiced law for eight to ten years, alongside forward-deployed engineers. Lemkin noted that selling AI into law firms is harder than selling security software because partners often pay out of their own pocket. That makes Harvey's scale even more striking.

Agent-led growth is the new race

The recurring point across four leading CPOs was responsibility. As agents take more core actions, the vendor stays accountable even when customers do things with the agents nobody anticipated. Build the guardrails before you ship the autonomy.

Lemkin and Wolan both landed on the same conclusion: the new race is agent-led growth and agent experience. The default choice of an agent matters more than ranking in a chatbot answer. If your product is not what an agent recommends, you are invisible regardless of your SEO or your sales team.

Tired vs wired: Lemkin's state of the union

In his solo keynote, Lemkin delivered a state of the union. Anthropic was reportedly projecting a $50 billion revenue run rate, while public software leaders were down as much as 70% on the year. He framed the gap as tired versus wired.

The fear that you would vibe code your own CRM? Dead. You can vibe code a lot, including most of the apps running the event, but you cannot vibe code something good enough to replace a real platform. Salesforce, HubSpot, and Pipedrive are not going anywhere because the moat is not code. It is integrations, data gravity, and workflow lock-in.

Frequently Asked Questions

What is agent-led growth?

Agent-led growth is a go-to-market motion where AI agents, not humans, discover and recommend products. If an agent defaults to your competitor, your SEO and sales team cannot compensate.

How do enterprise agents differ from chatbots?

Chatbots answer questions. Agents take actions, write to systems of record, and execute multi-step workflows autonomously with human approval at key checkpoints.

What is the Rubrik pattern for production agents?

Use LLMs to generate plans probabilistically, then validate and execute those plans with deterministic, explainable logic. This prevents hallucinated actions from reaching production systems.

Why is answer engine optimization replacing SEO?

Answer engines now drive roughly 50% of website visitors, up from 10% a year ago. Brands optimizing only for traditional search are missing half their audience.

Who is accountable when an enterprise agent fails?

The vendor. SaaStr panelists agreed that vendors remain accountable for what customers do with agents, not just whether the core product works as designed.

Source: SaaStrAI