BUILDING
BESTFIT'S
AI STRATEGY

Rather than simply asking, "Where can we use AI?" we focused on identifying moments where generative AI could create meaningful value. Every opportunity was evaluated through three lenses: improving the user experience, supporting BestFit's product strategy, and remaining technically feasible within the scope of the GitLab cohort.

This included mapping AI touchpoints across the user journey and prioritized concepts based on user impact, implementation effort, and long-term business value. While ideas ranged from AI-powered eligibility explanations to document analysis and case manager automation, we ultimately focused on two experiences where AI could immediately reduce friction without introducing unnecessary risk.

This framework helped us narrow dozens of ideas into two high-impact initiatives that balanced innovation with user trust: AI-powered Support Plans and a conversational Benefit Assistant.

Once we identified Support Plans as our first AI experience, we needed to determine which model produced the most useful outputs. I evaluated multiple LLMs against product-specific criteria including:

• readability
• response structure
• knowledge
• tone
• token usage

This approach ensured recommendations remained accurate, transparent, and grounded in trusted eligibility rules.

Using what we learned from model testing, I redesigned Support Plans as a more adaptive, personalized experience. Instead of giving every user the same static checklist, the system could use their screening results, household details, expenses, and eligible programs to generate guidance tailored to their situation.

We selected the o3 model because it produced the strongest balance of reasoning, structure, readability, and token efficiency. Its reasoning capabilities helped the product go beyond repeating program instructions—it could organize related tasks, remove unnecessary duplication, account for user context, and present next steps in a more useful order.

Next, we explored how conversational AI could support clients and case managers throughout their journey by providing answers in a more natural, accessible way.

Instead of searching through FAQs or policy documents, users could ask questions conversationally and receive clear, contextual guidance. The experience was designed to feel less like a search tool and more like a knowledgeable guide—helping users better understand their benefits, application requirements, and next steps without overwhelming them with technical language.

The GitLab Demo Day was an opportunity to showcase how AI could make benefit navigation more accessible and actionable. Beyond designing the product experience, I helped shape how we communicated its value to partners, funders, and other cohort participants.

I collaborated on the demo narrative, organized the presentation flow, and designed marketing collateral that highlighted our AI vision. During the event, I presented the product to attendees, walked them through the AI-powered experiences, and answered questions about our design decisions and implementation strategy.