Keller Williams Realty International (KWRI) · Case study
Momentum Guide: AI planning and coaching
I created Momentum Guide to help agents decide what to do next and turn their situation into a practical plan. The prototype evolved into KW Test Pilot, with a separate coaching flow added after Gary Keller’s input.
Corporate headquarters · Industry & Learning. Research, prototypes, and enablement for the Keller Williams network of agents, Market Centers, and regional teams.
The starting point
Agents were asking for direction: a next step, a plan, and guidance for the situation they were facing. The opportunity was to make that support accessible through a structured AI application.
My responsibility
I originated the application, designed its intake and workflow logic, built the prototype, diagnosed routing and answer-correlation issues, tested it with agents, and prepared it for leadership and technical-team review.
Tools & methods
AppSheet · Google Sheets · Apps Script · Claude Code · Claude API · GitHub Pages
Building the initial planning prototype
I began with the Momentum Guide concept: collect the agent’s situation and return an actionable plan. The structured version used AppSheet, Sheets, and Apps Script, with Workflows, Questions, Runs, Answers, and Outputs connected in its data model.
I worked through workflow misrouting and answer/data-correlation issues, then refined the outputs toward situational analysis and milestone-based plans. The structured prototype covered approximately 10 workflows.
Adding a distinct coaching path
After a brainstorm with Gary Keller about a coaching-style application, I added Coaching as a separate experience within the same app. Users selected their context, such as Agent or Market Center, and then chose Momentum or Coaching.
The coaching path used prompt selection and text input. I developed candidate prompts with AI and gathered additional ideas during a Market Center visit. Planning and coaching were separate flows, rather than names for separate product launches.
Evolving the build for broader testing
I developed the proof of concept through AppSheet and Claude Code work and a Claude API demonstration on GitHub Pages. The API-connected demo expanded to 12 real estate scenarios. These were stages in one evolving prototype: establish the idea, adapt it to broader use, and make it easier to distribute for testing.
Real estate peers and invited agents tested the experience. Their feedback identified robotic language and a preference for a voice workflow, which led me to explore ElevenLabs. Voice remained an exploration; the tested coaching flow used text.
Presenting the prototype and handing it over
I presented the work to leadership and received approval to advance the demonstration to CEO-level testing. I then supplied it to LABS for further exploration and development.
The delivered work was the functioning prototype, testing, demonstration, and handoff. Broader production development did not follow during the period I could confirm.
Outcome
A tested prototype with a clear next iteration
An agent-tested planning and coaching prototype, a 12-scenario API-connected demonstration, and a technical-team handoff. User feedback established a clear next development direction around language quality and voice interaction; production rollout remained pending.
Let’s connect
What could work better?
I’m interested in AI innovation and solutions roles where research becomes useful systems.