The Challenge
The January executive workshop set the lender's sequence, leadership fluency first, then the workforce, and the roadmap named the hard part plainly: a first bootcamp for product development, with learner cohorts to follow. That middle tier is where most enterprise AI enablement quietly fails. A day of demos produces enthusiasm; enthusiasm dies on contact with real work, because real work demands things demo-driven training never covers: evaluating whether outputs are actually good, building to standards, working inside policy, and knowing when a prototype is ready to become something more.
The lender set the bar higher than tool training. The bootcamp's stated goal was to create versatile, tool-agnostic AI practitioners, people who could "go anywhere and lead any team through any aspect of AI adoption." Not employees who can use a copilot, but employees who can select techniques independently, build and evaluate solutions, and carry adoption into whatever corner of a large mortgage operation they work in. That's a different kind of program: five consecutive days, homework included.
The Approach
PhoenixTeam delivered the bootcamp as an immersive, team-based course built on its seven-pillar curriculum (Understand, Discover, Reimagine, Pilot, Prove, Scale, Adapt), the same workforce-centered adoption framework introduced to the executive team in January, now taught to the people who would live it.
Day one started with teams, not tools: cohort members formed named teams, generated their own logos with AI, and set expectations that fingers stay on keyboards. The Understand pillar moved from foundations and prompt craft into territory corporate training almost never reaches: prompt evaluation, complete with grader types and evaluation approaches, so participants learned not just to generate outputs but to judge them systematically. Discover added retrieval-augmented generation and turned the cohort's attention to their own work: building ragbots, then prioritizing, analyzing, and roadmapping their real use-case backlogs.
The middle of the week shifted from using AI to changing work with it. In Reimagine, participants learned to restate problems as opportunities, practiced the facilitation craft of guiding groups through divergent and convergent thinking, ran experiments end to end, and built a reusable company brand skill they could invoke to regenerate any artifact in house style, a small piece of infrastructure that outlives the course. Pilot put it all together through vibe coding: teams built a working loan-officer assistant website, built agents, and then built their own prioritized use cases into functioning applications.
The final stretch made them stewards, not just builders. Prove introduced project standards, application cleanup, and the definition and implementation of real guardrails. Scale had participants engage conversationally with the lender's actual AI policy, learning precisely what they can and can't do, then articulate what they'd need from the enterprise to fully realize AI's value: policy evolution, data access, further enablement, captured as documented questions for the leadership team. Adapt closed the arc with continuous improvement and strategic renewal, the habits that keep a practitioner current after the instructors leave.
The Outcome
The cohort graduated with proof of capability rather than a certificate of attendance: working applications they vibe-coded themselves, agents and ragbots grounded in real content, evaluated prompts they can defend, a reusable brand skill now available to every future build at the company, project guardrails they defined and implemented, and prioritized roadmaps for their own use-case backlogs.
Just as valuable is what flowed upward. Because the bootcamp had participants engage the company's real AI policy and articulate what they need to create value (policy changes, data access, enablement), leadership received something rare from a training program: a documented, specific agenda from a newly fluent workforce. That's the adoption flywheel working in both directions.
And the bootcamp didn't end when it ended. Within ten days, the cohort was back for the first of three advanced sessions (agentic coding environments, then connecting to enterprise data through Model Context Protocol, then LLM-as-judge evaluation), supported by weekly office hours, a certification exam, and a standing working group. Graduates arrived already building: first applications underway, one participant running a self-built testing agent complete with the guardrails they'd learned to write. Their bootcamp use cases, meanwhile, entered the enterprise pipeline, with top candidates from each builder moving into value-proposition review ahead of prioritization with the company's newly stood-up AI governance team.
And the program-level story completed itself. The sequence designed in January (executives first, workforce next, cohorts by spring) executed on schedule, with the first five-day practitioner cohort delivered roughly a quarter after the executive kickoff. For an organization of the lender's scale, that's the claim most AI enablement programs can't make: not that training happened, but that a dated plan became a pipeline of builders.
