Using GenAI to increase sales without losing trust

We have rich customer data available before a field visit, gathered through call center interactions and device context. That made it possible to tailor sales guidance to individual customers in a way static scripts never could.

However, that data was fragmented across different systems and databases, which made putting it together into something meaningful a challenge in and of itself. Throughout this project, we partnered closely with the data analytics team to improve how and where our customer data is stored and accessed, so that we could show useful insights within Field app.

The goal was never for experts to read directly from the app. It was to help them internalize guidance and bring it into conversation authentically. We would know if we've done our job well if experts actually spend less time viewing Field app while with customers.

Early explorations made one thing clear very quickly: this could go wrong.

Some early model outputs referenced deeply personal details about customers — how many kids they had, their ages, or specific devices throughout the home. Even if accurate, surfacing that information risked making conversations uncomfortable or unsettling.

A feature designed to help experts could just as easily damage trust with customers.

This became the central design challenge of the project.

I worked as Senior Manager, Product Design, overseeing one senior designer. My focus was on product strategy, placement within the Field app, and the detailed UI decisions that shaped how SmartPitch appeared and behaved.

I partnered closely with Product, Data Science, Legal, and Engineering to define what insights were appropriate to surface, shape guardrails around tone and specificity, and ensure the experience felt supportive, not prescriptive. Design's role here wasn't just execution — it was judgment.

We were intentional about what SmartPitch did not do. Instead of full scripts or overly specific recommendations, we focused on:

We iterated through pilots, refining prompts so outputs felt distinct enough to be useful without becoming repetitive — and broad enough to stay comfortable in conversation.

We also made the AI visible but transparent. Experts were told the content was GenAI-generated and given guidance on how to use it, helping them trust the information without over-relying on it.

Legal review played a significant role in shaping SmartPitch. There were strict requirements around what experts could say, how products could be positioned, and what claims were allowed. These constraints influenced the language used in prompts and outputs, how insights were framed, and where SmartPitch appeared in the flow.

Rather than fighting those constraints, we designed within them — using structure and clarity to keep the experience safe, compliant, and useful.

SmartPitch lives directly within the Field app, embedded in the customer's job page.

The feature is optional, easy to scan, and designed to support — not interrupt — the flow of the job.

Sales impact was significant.

Sales per 100 jobs (SP100) — historically a difficult metric to move — increased meaningfully across roles and markets. In pilots where SmartPitch was used, SP100 jumped dramatically compared to jobs without it.

Even more telling: only about 9% of eligible jobs had a SmartPitch viewed, suggesting significant upside as usage grows.

The success of SmartPitch shifted internal confidence in AI-assisted tools. What started as an experiment is now planned for expansion to additional platforms across the company, supporting other expert roles beyond Field, and is a key driver for achieving the company's revenue goals for 2026.

Perhaps more importantly, it reframed how teams think about AI: Not as automation, not as replacement, and not as a chatbot slapped on top of the product, but as thoughtful augmentation when designed with care.

This project reinforced that designing with AI is less about capability and more about restraint.

Concise content, clear guardrails, and transparency mattered more than sophistication. Designing for trust — with both experts and customers — was the real challenge.

It was a strong first step in bringing GenAI into Field in a prominent way, and it shaped how I now guide teams exploring emerging technology: start with the human moment, then decide what technology earns a place there.