Agentic AI has spent much of its early life in the back office, where the work is repetitive and the risk of an agent speaking to a customer is low. Revenue teams have been more cautious, and for understandable reasons: a demo that misrepresents the product or a deal room that contradicts what the seller said in a meeting does real damage. Walnut‘s new platform is a notable case of agents moving into that customer-facing space, and the way the company has approached it says a good deal about what enterprise adoption in go-to-market functions is going to require.
The Release
Walnut, used by go-to-market teams at hundreds of B2B software companies to build interactive product demos, deal rooms, and buyer-facing content, has announced a family of AI agents that covers both sides of the buying experience for the first time. Two agents are new. AI Mode for Playlists and Deal Rooms assembles a complete, personalized buyer-facing experience from a single prompt. Walnut Xpert is an agent that sits on top of a demo and lets buyers ask questions and find their own way through the product. Both join InsightsAI and AI Mode for demos, already live.
The Accuracy Problem, Solved Structurally
The reason customer-facing agents have been slow to arrive is the accuracy problem, and Walnut’s answer to it is structural rather than procedural. All four agents are built on the same foundation. Rather than generating content from a prompt and a guess at what the product looks like, they work from Walnut’s own capture technology, which clones the actual product, its real screens, real flows, and real data, and builds every asset on top of that captured context.
When AI Mode assembles a demo or Walnut Xpert answers a buyer’s question, the output is anchored to what the product actually does. Walnut describes this as the architectural bet behind the announcement: a buyer’s journey holds together, from first click to signed deal, because every agent along the way is drawing on the same grounded context rather than starting from a blank page.
The Oversight Problem, Solved First
The second precondition for customer-facing agents is oversight, and here Walnut’s sequencing is the interesting part. Ahead of shipping any agent with autonomy, the company released the AI Center, a governance layer that gives administrators visibility, permissions, and guardrails over how AI is used across their workspace. The order was deliberate. In Walnut’s words, agents that act on an organization’s behalf are only adoptable once someone inside that organization can see and control what they are doing, across every agent, in one place.
That layer is being extended with bring-your-own-key (BYOK) support, giving customers direct control over security, privacy, and spend through their own model key. Walnut’s premise is that its verticalized buyer-enablement harness, the captured context, the workflows, and the governance, is the durable value, and it works with any underlying model. As frontier models keep improving, the quality and value of Walnut’s agents improves with them, regardless of which model runs underneath. A companion MCP will let customers run Walnut from the AI provider of their choice.
What the Agents Do Once Those Conditions Are Met
With accuracy and oversight handled, the agents themselves can be judged on what they change.
On the seller side, AI Mode for Playlists and Deal Rooms takes a description of what a specific buyer needs and assembles a complete, personalized playlist or interactive deal room, drawing on the company’s captured product content, the buyer’s CRM record, and the seller’s stated intent. Curation that used to take an account executive or customer success manager roughly three hours per deal now takes a sentence. Because the agent works from the same grounded foundation as the rest of the platform, the deal room reinforces the story the demo already told rather than starting over.
Demo creation, handled by AI Mode for demos, has been through the same compression. Over the past year of Walnut’s demo creation and personalization agents, a process that took days, then hours, has come down to minutes, with greater fidelity and interactivity rather than less.
On the buyer side, Walnut Xpert replaces the fixed, click-through path with an open one. A buyer can ask questions, go deeper on what matters to them, and find their own way through the product, guided by an agent grounded in the seller’s actual, captured product. Walnut reports that sessions with Walnut Xpert alongside a demo more than double the time a buyer spends. The questions asked generate intent signal a static demo never could. The company likens the result to an AI SDR, qualifying interest and surfacing what buyers care about at a scale no team could staff for, and turning a click or a page view into a record of what buyers actually asked, in their own words.
The Scale Effect
The cumulative outcome, in Walnut’s framing, is that personalization stops being something a team builds for its five biggest deals and becomes something it builds for all of them. That is the shift agentic AI has promised revenue teams for some time. Walnut’s release is one of the more complete attempts to deliver it across the whole journey rather than at a single step.
The Strategic Read
Oren Blank, CEO of Walnut, placed the release in the context of a market where the underlying models are widely available:
“Every company has access to the same frontier models now, so the ability to build features isn’t a competitive advantage anymore. What is: your ability to go to market effectively, tell a coherent story, and create a buyer experience that actually helps your buyer make the informed decision.”
Where People Remain
Walnut is explicit that autonomy has limits. In every case, the person on the GTM team remains the one who decided what story should be told and to whom. The agents execute, working from one shared, grounded context. The judgment stays with the humans who understand the account.
Availability
The AI Center, AI Mode for demos, AI Mode for Playlists and Deal Rooms, and Walnut Xpert are available to Walnut customers today. BYOK support and the companion MCP will roll out in the coming weeks.
Walnut describes itself as a vertical AI platform for product-led buyer enablement, built on a single agentic architecture that lets GTM teams direct one cohesive buying story instead of stitching together disconnected AI tools. For anyone watching how agents make their way into customer-facing work, the combination of grounded context and governance-first sequencing is the pattern to note.
