Overview
Problem: How do you get people to trust an AI they can't see inside?
Salesforce wanted to fix its loyalty program. Marketers were spending hours by hand on work that should've taken minutes, and customers still got offers that felt like they were meant for someone else. Our team ran mixed-methods research and gave Salesforce four recommendations for keeping humans in control of AI-driven offers.
NDA Disclosure: Our recommendations spanned transparency, user control, and guardrails for AI-driven offers. Per our agreement with Salesforce, I cannot disclose specific insights. This case study is about how we got there.
Overview
The Brief
Marketers were pulling customer lists, building rules, and checking for conflicts by hand, then launching offers one by one. Even after all that work, customers still got promotions that felt disconnected from their real lives. Salesforce wanted to know if AI could fix that without becoming the very black box that made marketers distrust the system in the first place.
Process
Getting People to Talk to us
We were undergrads with no marketing connections and no budget for incentives, trying to reach mid-career CRM managers who had no reason to respond to a cold message. Cold outreach mostly went nowhere. So we adapted: I built our survey and personally recruited more participants than anyone else on the team, and we brought in proxy users and end users alongside the marketers we could actually reach, instead of pretending we'd talked to more "real" ones than we had.
Process
My Contribution
I created and ran our survey, ran interviews, and led part of our field research, going out to cafes to watch how people actually reacted to real promotions in the moment, not just how they described their habits afterward.
Process
What the Survey Told us
Loyalty programs are already part of daily life: ~75% of respondents used one at least half the time. But actually redeeming offers lagged behind, with only 52% saying they redeemed them often.
The biggest barrier was relevance. 56% had wondered why they received a particular offer at least sometimes, yet people weren’t opposed to personalization. 40% were comfortable with companies using purchase history, and 44% found AI-driven recommendations more useful than generic ones.
People wanted the personalization. They just wanted to understand it.
The best offers were simple: a discount on something they already buy (84%), easy redemption (56%), and rewards that didn’t expire too quickly (44%). Underneath it all was one consistent preference: people wanted control over how often they heard from a brand.
Process
What the Interviews Told us
Overview
Reflection
Splitting time across a survey, interviews, and fieldwork taught me how differently each method surfaces the truth. The survey told us what people wanted in aggregate, the interviews told us why, and watching people in cafes told us what they actually did when a real offer showed up on their phone, which wasn't always what they said in either of the other two. That gap between stated and observed behavior is the reason I trust field research more now than I did going in.