About the role
Enterprise agentic AI begins with ambiguity: a consequential problem, unfamiliar systems and no ready-made blueprint. You will turn that uncertainty into working intelligence. This is a chance to define how adaptive agents enter real organisations, earn trust through evidence and grow from a promising pilot into an operational capability.
What you would do
- Lead discovery that reveals the real decision, workflow and constraint behind a customer's stated request.
- Design and build agentic systems that connect advanced platform capabilities to measurable operational outcomes.
- Create evaluations with customer evidence so progress is visible and trust is earned rather than assumed.
- Investigate difficult failures across agents, tools and services, then explain the path forward in clear language.
- Bring what you learn in the field back into the product and help shape how enterprises adopt autonomous systems responsibly.
What we look for
- Questions before answers, and measurable acceptance criteria.
- Clear separation of what the agent did, what the tools did, and what the service did.
- Honesty with the customer about limits.
Levels
| Level | What we expect |
|---|---|
| Junior | Runs and explains a demo; follows a spec to a working project. |
| Middle | Turns a brief into a specification with acceptance criteria. |
| Senior | Builds, evaluates and defends a project; handles an escalation end to end. |
| Staff+ | Plans pilots and rollouts; decides what the product should and should not do. |
Example challenges
These are starting points, not a strict assignment list. Choose one that suits your interests and experience, adapt its scope, or propose a focused challenge of your own.
SOL-1 A demo script
Easy Run npm run demo -- simple, or scaffold a project with zen init. Write a walkthrough you
could give a customer: what to run, what they will see, what to say, and what can go wrong.
Deliver the script with the real output at each step, and three questions a customer might ask with your answers.
SOL-2 From a brief to a specification
Medium The brief: "Our support team triages a shared mailbox and books follow-up calls. We want help with that." Write the specification using the project template: agents, tools and boundaries, environment, and measurable acceptance criteria.
Deliver the specification and a list of the questions you would ask the customer first, with what you assumed meanwhile.
SOL-3 Build the project
Hard Build the project from your SOL-2 specification against the mock: configuration, prompts,
skills, sandbox. zen check must be clean, and the acceptance criteria must be demonstrable.
Deliver the project, a run for each criterion, and any criterion you could not meet and why.
SOL-4 "It said it booked the meeting"
Hard A customer writes: "The agent told me it booked the call, but the meeting is not in the calendar." Reproduce the situation with the mock and an agent. Remember that the mock keeps no state.
Deliver an incident report: how you reproduced it, how you told what the agent did from what the service did, the most likely causes ranked with evidence, what you would say to the customer, and what you would change so the next report can be answered in minutes.
SOL-5 A pilot plan
Expert A customer has three internal APIs, no documentation for two of them, and wants an agent on top. Plan the pilot.
Deliver a plan: access and data handling, how the APIs get described, the evaluation set and who writes it, the rollout stages, the success criteria, the main risks, and what you would refuse to do in a pilot.