AI Workers / Level AI
I identified the agent-selection problem, prototyped one conversation, and led its product definition and delivery.
Delivered experience
One chatA unified entry point for 10+ specialist agents. Direct access to specialist workflows retained.
Direct access to specialists stays available.
Illustrative decision flow · not a live AI interaction
The problem
Customers had to choose among 10+ specialist agents before asking a question. Requests spanning several agents meant carrying context between tools and assembling the answer themselves.
My part in the change
I led the unified entry point while keeping specialist workflows directly accessible.
I worked with engineering on MCP access: timeouts, progress, result delivery, and comparable answer quality.
I used repeat use, breadth, and concentration across accounts to guide iteration with customers and customer success.
Scope & contribution
AI Workers brings conversational AI agents to customer-experience workflows. Teams can investigate issues, analyze customer conversations, and take requested actions in connected tools. I led discovery, launch, pricing, and adoption, including the move from choosing among 10+ specialist agents to one chat entry point.
Customers had to understand a library of more than ten AI agents before asking a question. Questions needing several kinds of analysis meant switching between agents, carrying context across, and assembling the answer themselves.
I identified the agent-selection problem, prototyped the unified experience, and led product definition and delivery. I worked with engineering on access through Model Context Protocol (MCP), with go-to-market teams on pricing and launch, and with customer success and customers on adoption.
Decision context
One entry point, with direct access to specialists still available.
I chose one entry point while keeping direct access to specialists. Complexity moved into routing: the system needed clear descriptions of each agent’s capabilities, and users needed to see which agents contributed.
Keep progress and results understandable in connected AI applications.
I worked with engineering on bringing AI Workers into other AI applications through the Model Context Protocol (MCP), which connects AI applications to external tools. My focus was timeouts, progress, result delivery, and comparable answer quality.
Learn from repeat use and from accounts where usage was low.
I reviewed breadth of usage, repeat use, and concentration across accounts, working with customer success and customers to investigate low usage and learn from successful accounts.
An illustrative comparison
Compare asking customers to choose a specialist with giving them one place to ask. Direct access to specialist workflows remains available.
AN ILLUSTRATIVE QUESTIONWhy are customers asking for refunds?
Explore a specialist. Direct access is still available.
I prototyped and led delivery of the unified experience.
Interactive illustration of a product decision. Not a live AI demo.
Published product walkthrough
Published product imagery illustrates behavior, not feature-release dates during my tenure. Interactive frameworks are explicitly illustrative.
Experience direction
Customers need an answer about their business. The interface gives them a direct way to ask, then inspect the response.
I led the experience from individual workers toward a unified entry point, alongside product strategy, pricing, and launch.
Natural language makes the business question the entry point.
The response connects the explanation to the conversations it draws on.
Outcomes in context
I prototyped and led delivery of a unified chat entry point, replacing the need to choose among 10+ specialist agents while retaining direct access to specialist workflows. I used account and cohort usage data to guide iteration and adoption.
A simpler entry point still needs to show its work: which agents contributed, where the evidence came from, and how to continue.
Visit the product at Level AIKeep exploring
Model quality + operating costs
GOOD PRODUCTS START WITH A CONVERSATION.