Reframing AI:
The Tool Was Familiar. The Approach Wasn’t.
📋The Overview
This project started as a tool training. It didn’t stay that way.
Agents were getting an AI assistant embedded in software they already used. The ask was to build training before launch. A clear scope. A solid outline. Work in progress.
Then the sponsor changed. The philosophy changed with it. The first approach asked agents to follow the AI closely. The second positioned agents as the experts. The AI was a suggestion, not a directive.
That shift required more than an update. It required agents to think differently about their own role. The training had to change to match.
Feedback came from every direction. Some of it contradicted. All of it mattered. The solution was to absorb what was useful, act on it quickly, and add video to bridge the gap. Four videos. Built in days. A chance for agents to watch the tool in expert hands before touching it themselves.
The training launched. The relationships built during the process turned out to matter just as much.
⚡Quick Facts
Role: Senior Analyst, Learning & Development
Organization: Walmart
Timeline: Extended rollout with multiple pause points; final video build completed in under a week
Tools Used: Articulate Rise, Camtasia, LMS, ChatGPT
Outcome: Successful launch; strong stakeholder relationships built; reputation established with key sponsors and SMEs
Skills:
🔍The Problem
The original ask was straightforward: agents were getting a new AI assistant embedded in their existing software, and they needed training before launch. How to use it, when to use it, what to expect.
Mid-project, it became clear the ask had changed. A new sponsor came in with a different philosophy about the AI's purpose. Under the first approach, agents were expected to follow the AI’s suggestions closely. Under the second, agents were positioned as the experts. The AI was a tool to support their judgment, not replace it.
That shift changed everything. The training was no longer just about functionality. It was about helping agents understand their own role in a system that had just been redefined.
📊The Research
Before anything was built, the most important step was watching the tool in use. Sitting with an experienced agent and observing how the AI assistant fit into a real customer interaction gave a grounding that no requirements document could have provided. It made clear what agents were being asked to do and where the friction points were likely to land.
The rest of the research happened in motion. This project started and stopped more than once, and feedback came from every direction throughout: sponsors, subject matter experts, and the agents themselves. Those voices didn’t always agree. Sponsors had strong opinions about how agents should relate to the AI, and those opinions shifted when leadership changed. Agent feedback pointed to different concerns entirely.
Holding contradictory input without losing the thread is part of the work. The research here wasn’t a phase that ended before design began. It was ongoing, and the ability to keep listening while still moving forward shaped every decision that followed.
🔀The Options
By the time the second sponsor came in and redefined the approach, the project was already mid-stream. There was a course. There was a deadline. And now there was a fundamentally different message to deliver to agents about their own role.
At that point, the options were real. Stay the course with what had already been built and launch on time. Or find a way to address the mindset shift directly, even with limited time and resources.
The timeline was not comfortable. But the audience was large, the shift was significant, and a click-through module wasn’t going to change how someone thought about their own expertise. The decision was to add something that would reach them.
🗺️The Plan
When the project started, there was a plan. A structured outline mapped out what agents needed to know, in what order, and how the course would flow. That foundation mattered, even when everything around it changed.
When the sponsor shift happened, the outline didn’t get updated. There wasn’t time. But the thinking that went into it didn’t disappear either. It became the scaffolding for a faster, more flexible approach.
The videos weren’t planned from the beginning. They were a response to the moment. The recordings from the observation sessions became raw material. What was available shaped what was possible, and the goal was always the same: give agents something that would change how they approached the tool, not just document how it worked.
That meant letting go of the original format and designing for what the learners needed most under the circumstances. A more polished solution might have looked different. But the best training is the one that reaches people, and this one had to reach thousands of them, fast.
🔨The Build
The build was never a straight line.
It started with a solid outline, a clear scope, and a course taking shape around what agents needed to know. Then feedback came in. From the first sponsor, from SMEs, from stakeholders with different priorities and different ideas about what the training should do. Each round of input shifted something. The course absorbed it and kept moving.
When the sponsor changed and the philosophy changed with it, the build shifted again. The outline didn’t get rewritten from scratch, but the framing did. What agents were being asked to understand about their own role in the process required a different approach than what had been planned.
That’s when the videos entered the picture. One day observing and capturing source material. Two days editing. One day masking client information. A hard deadline waiting at the end. The goal was to show agents what the tool looked like in the hands of someone who already knew how to use it, with emphasis drawn exactly where attention needed to land. The existing course gave agents the steps. The videos gave them something to anchor to.
🎯The Outcome
The training launched. Agents moved through it. But the most meaningful outcomes weren’t in the data.
This project involved navigating competing voices, absorbing a significant scope shift mid-stream, and pushing back when something wasn’t working for learners. That kind of work either builds relationships or burns them. In this case, it built them.
The second sponsor became a strong advocate. The SME who pushed hardest during the process, and who knows they can be demanding, came away with a clear sense of who Hillary is as a designer and what she stands for. The result was something more durable than a completion rate:
“People are using Hillary’s name in rooms she’s not in.”
That’s the outcome that matters.
💡The Reflection
Every project teaches something. This one taught several.
The first sponsor had a meeting style that created unnecessary anxiety. No one knew how long anything would take or what to expect. That stuck. Now every meeting I lead opens with a timeframe and an agenda. It’s a small thing that changes everything about how people show up.
Mid-project, an over-prompting pattern emerged in how agents were interacting with the AI. It wasn’t part of the original scope, but ignoring it would have undermined the training. Catching it and addressing it before launch was the right call, even under time pressure.
The pre-launch feedback session was a revelation. It hadn’t been on my radar as an option, but watching it work changed that. On longer projects with extended rollouts, asking for a structured feedback moment before final launch is now part of how I think about project planning. You don’t know what you can ask for until you see it in action.
