Company Background

Favor is a same-day delivery and food ordering platform based in Austin, Texas. They deliver food, groceries, and everyday essentials across Texas through a network of Runners (delivery drivers). Acquired by H-E-B in 2018, Favor combines local business partnerships with fast, contact-free delivery. Learn more at favordelivery.com.

‍

“We’re spread out all over the great state of Texas. We handle the H-E-B family grocery deliveries, and also have our own delivery platform.” – Evan McMillan, Fleet Support Team Lead

‍

The Challenge

Favor’s Fleet Team is a niche group with processes that differ from the broader support team. As a result, Guru’s AI-generated search results were crowded with irrelevant information from other teams, making it harder for fleet agents to find the right answers quickly. The time-consuming search process discouraged team members from checking resources altogether.

‍

“People would sometimes say, ‘Guru’s AI isn’t working for us,’ but it’s really that they weren’t using it correctly. Once we fine-tuned it and adjusted the prompt, the quality of answers changed completely.” – Evan McMillan

‍

The Solution

Favor used Guru’s Knowledge Agents to create a dedicated AI agent focused solely on Fleet team support.

1. Evan McMillan, Support Team Lead, customized the agent’s prompt to focus only on fleet-related content.
2. He adjusted the agent’s tone to match Favor’s friendly, approachable culture.
3. The agent was integrated into Slack, making it easier for the team to ask questions and get answers in real time.

‍

“Once we got the ability to make individual agents, it was really exciting to be able to hone in on the specific type of work that we’re doing. The agent really started to understand our business better and gave more focused answers.” – Evan McMillan

‍

“Hooking it into Slack was the game-changer. I was hesitant at first to have something answering questions automatically, but once we tried it, it completely unlocked the potential. Seeing good answers come through in Slack really built trust.” – Evan McMillan

The Approach

1. Defined the Business Context – Evan provided detailed context about Favor’s business model and fleet support processes so the AI agent could better understand how Favor’s customer segments (drivers, customers, and merchants) work together. This helped improve the relevance and accuracy of the agent’s answers.

‍

“Explaining how the business works so the AI agent can understand what Favor is, what we’re trying to do, and how our customer segments work together really helped improve the quality of answers.” – Evan McMillan

‍

2. Tailored the AI Agent’s Focus – Since the fleet support team operates differently from the broader support team, Evan narrowed down the agent’s sources to only fleet-related content. This eliminated irrelevant answers from other teams and allowed the agent to deliver more precise responses.

‍

“My team is specialized, so the default AI agent would pull from broader company knowledge, which wasn’t always relevant. Limiting it to just fleet-related sources made a huge difference.” – Evan McMillan

‍

3. Refined the Agent’s Tone – Evan set the agent to respond with a more conversational and supportive tone, aligning with Favor’s company culture. He positioned the agent as a “training coach” to help agents not just with answers but with learning how to handle situations.
   
   
4. Integrated with Slack – The agent was embedded in Slack, making it easier for team members to ask questions and receive answers in real time. Having the agent within a familiar communication channel increased adoption and engagement.
   
   
5. Iterated and Improved – Evan tracked different versions of the agent’s prompt in a spreadsheet, allowing him to easily revert to earlier versions and refine responses over time. He also experimented with formatting instructions, like listing steps clearly when relevant, to make answers more actionable.

‍

‍

‍

The Outcome

✅ Increased confidence in using Guru—more team members now say, “Hey, I wonder if that’s in Guru. I’ll ask the AI agent real quick.”

✅ Improved efficiency in handling fleet-specific issues that don’t follow standard procedures.

‍

“It’s been a huge help for those oddball questions that don’t fit into a standard process—it’s saving us a lot of time.” – Evan McMillan

‍

✅ Faster, more accurate answers—especially for niche fleet-related questions.

✅ Higher engagement with Slack-based AI support, leading to better adoption and trust.

‍

“Seeing good answers pop up in Slack really built confidence. Now people are asking more questions and trusting the agent’s responses.” – Evan McMillan