Enterprise AI Product Strategist /// Consulting + Full-Time Opportunities /// Enterprise AI Product Strategist /// Consulting + Full-Time Opportunities /// Enterprise AI Product Strategist /// Consulting + Full-Time Opportunities /// Enterprise AI Product Strategist /// Consulting + Full-Time Opportunities ///
AI Enablement /// Meeting Prep Agent

From Prototype to Platform: AI Meeting Prep That Sales Teams Trust

I moved a GenAI use case from idea to validated evidence in 3–4 weeks, securing additional funding and engineering capacity to integrate an agenda-first meeting prep workflow into an existing sales platform. The capability later expanded from a 12-person pilot to ~90 users, and ultimately scaled to 300+ active sales associates within a year.

01 /// The Challenge

The Weekly Prep Bottleneck and the Last-Minute Pivot Problem

Sales associates typically prepared at the start of the week, spending ~2–3 hours to get ready for ~6–8 client meetings. They also faced a second constraint: meetings could appear mid-week, or plans could change last minute, leaving minimal time to prepare.

A data science team supporting the client group had access to rich internal context across client databases, prior meeting notes, sales activity, and internal research commentary. But that context lived in different places and did not reliably support fast, meeting-specific preparation. The result was inconsistent preparation, uneven meeting quality, and missed opportunities to follow up on the right topics.

Project Visual
02 /// The Strategy

Validating the Workflow, Not Just the Tech.

Instead of starting with features, we ran an innovation sprint to validate where AI could remove the most friction in meeting prep. I led sessions with sales associates, partners, and stakeholders to map their prep workflow, identify the highest-impact breakdowns, and generate multiple solution directions.

To ground decisions in real behavior, we had associates sketch what their ideal prep experience would look like. Those sketches, combined with workflow mapping and targeted 1:1 conversations with associates and managers, helped define must-haves, nice-to-haves, and what not to build.

  • Key learnings that shaped the experience:
    • Sales associates did not want to “chat.” They wanted an agenda-first output they could send to clients, backed by specifics when they needed to go deeper.
    • Trust is fragile, one incorrect or low-quality output can stall adoption.
    • The experience needed to avoid information overload, especially for the 5-minute prep scenario.
  • We shifted from a prescriptive meeting plan to an agenda-first workflow:
    • Provide a high-level agenda outline that was usually ready to send.
    • Support it with relevant details inside the platform so associates could quickly verify what to discuss.
    • Require editability to maintain a human-in-the-loop and to handle last-minute changes.
  • We also made deliberate scoping decisions:
    • We did not build a social profile enrichment feature due to privacy and PII concerns. Instead, we surfaced personal context already captured in prior meeting notes.
    • We avoided throwing too much information at associates at once, using progressive disclosure.
  • Fast view hierarchy (designed for minimal prep time):
    • Agenda draft (short description plus 3–5 meeting topics)
    • Open follow-ups from prior notes
    • Suggested talking points informed by internal research and current market dynamics

To validate direction and trust cues, we tested both a high-fidelity Figma prototype and a functional Streamlit prototype with real users. We structured the output into separate sections (follow-ups vs suggested talking points), added a visible generated-on timestamp, and ensured associates always reviewed and edited the agenda before sending.

Workflow Design
03 /// The Outcome

From Validated Evidence to Funded Integration

By validating the workflow, output structure, and trust cues before full build, we accelerated the path from idea to evidence. In 3–4 weeks, we produced validated learnings that stakeholders used to commit additional funding and engineering capacity, and we translated those learnings into clear build priorities.

The sprint concluded with a stakeholder readout and handoff package (workflow map and friction points, plus a Figma prototype) so product and engineering teams could continue validation and implementation without restarting discovery.

That foundation enabled downstream teams to integrate the capability into the existing platform and expand adoption from ~12 pilot users to ~90 users, later scaling to 300+ active sales associates within a year. The result was reduced assumption-heavy development, faster decision-making, and a repeatable pattern for validating GenAI use cases inside an existing workflow.