Unofficial. A job application for Cresta’s Forward Deployed Product Manager, AI Agent role, built from public pages. Not affiliated with Cresta.

Why me.

I have never held the title Product Manager, and I have not worked in a contact center. I have spent twenty years taking client work from scope to launch, and this year I have built and run AI agents of my own.

What the posting asks, and what I have done

The posting asks forWhat I have done
Five or more years in consulting, implementation, product or customer successEight years as an agency producer owning client accounts from pitch to launch, for American Express, Citi, Southeast Toyota, Unilever and PGi. Then delivery roles at A+E Networks, Meta and Eptura. Today I run the release train for ten vendor teams on a tolling modernization program.
Leading complex deployments, with senior customer stakeholdersAt a publicly traded retailer I ran readiness checkpoints and go or no-go decisions through build, test, acceptance and cutover. The go-live landed on time. On my current program I closed the first increment with every committed requirement accepted, and ran the two-day close-out for the client. Retailer, planning board
Workshops and design sessions that align people on use cases and success criteriaI interviewed every workstream lead before planning and turned the answers into one script that ten vendor teams now plan from. I run a retrospective with all leads at the end of each increment. For Southeast Toyota I planned, ran and presented a user-testing study of the dealer site with a colleague.
Exposure to building agents. Personal projects count.A manager-and-worker loop: one agent writes the tasks, and local models work them in four stages against hidden tests and planted bugs, with a scoreboard that shows whether they improve. Also one written workflow that three agents from three vendors all run, with a ledger and an audit that stop them inventing skills. The loop, the workflow
Design, build, test and iterate on AI agentsFour music products shipped with AI agents behind more than 9,000 automated tests. Tests are the referee, not the model’s word. I check the agents’ numbers: a reported 682 tests turned out to be 384. How I work
Own outcomes and define the metricsMy planning board never shows a number it cannot source, and a gate with no data shows a dash, never green. The prototype on this site carries that rule into a launch decision. Prototype
Improve the playbooksThe interview script every lead plans from. Six design documents that agents and people both build from. A rules file for agent sessions that records two times an agent overreached, and the rule each one produced. Case study
Work closely with AI or LLM-first productsIn 2024 I was delivery lead for teams fine-tuning Gemini, ChatGPT and Claude models on customer data. At Meta I produced the voice recordings that trained its AI voices, heard across Facebook, Instagram, Oculus and Portal.

The gap, plainly

I have not worked in a contact center.

I have learned a new client’s operation many times, and I would expect to learn this one the same way: in the room with the people who run it. This site is the reading half of that, done in a day.

My titles have been Producer, Scrum Master, Agile Coach and Release Train Engineer. I am the product owner and builder of the tools named above.

I do know voice. At Meta I supported the embedded voice in the Ray-Ban Stories glasses and ran final QA. It shipped on the public launch date.