“We just realized we were wasting a lot of potential by just having to manually do things.”

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

Faire is a wholesale marketplace connecting independent retailers with brands around the world. Their CX team relies on Guru as the central source of truth for everything from policies to tariff updates — making the accuracy of that content mission-critical for support quality and agent confidence.

Like many fast-growing companies, Faire’s CX knowledge lived in Guru — but keeping it current was a manual, time-intensive process. And as long as content couldn’t be trusted, the rest of Guru’s potential stayed out of reach.

The Challenge: A Vicious Cycle That Blocked Everything Else

A Small Team, a Big Maintenance Burden

When Carly Byrge stepped into her enablement role at Faire, she inherited a Guru instance with trust scores in the forties and fifties — and hundreds of unverified cards waiting to be reviewed. With a lean team and no automated process to help, Carly took on the bulk of the verification work herself.

‍

“I spent probably an average of six hours a week trying to stay on top of and catch up with all of our unverified cards. It was a lot.”

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

By the end of summer, she’d manually brought the trust score to around 75% — a meaningful improvement, but not a sustainable one. The time cost was real, and the ceiling was clear.

People Didn’t Trust the Content

The low trust scores weren’t just an administrative problem — they were affecting how people did their jobs. When content felt outdated or unreliable, users stopped trusting it. That meant they came to Carly directly with questions instead of finding answers in Guru, adding to her workload and undermining the very purpose of having a knowledge base.

‍

“We did see a lot of agents just not trusting things. It’s almost like I wasn’t fast enough.”

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

The organization wanted to push deeper into features like Knowledge Agent chat and Slack-based auto-answers, but all of it depended on the content being trustworthy first. The team was stuck in a loop: too much time on maintenance to make progress on adoption, and too little adoption to justify investing in the content.

Potential Blocked by Manual Work

Leadership was actively interested in getting more value from Guru — pushing for broader agent chat usage, expanding Knowledge Agents, and improving the quality of AI-generated answers. But as long as Carly was spending the equivalent of a full workday each week just keeping up with verifications, there was no bandwidth left to move forward.

The Approach: Start with What Matters Most, Then Build

Faire’s approach to Guru’s quality automations was deliberate — they started with the Knowledge Agent that mattered most to their team and used what they learned to guide future expansion.

1. Prioritizing the CX Policy Assistant

Rather than rolling out quality rules across every collection at once, Carly chose to start with the Knowledge Agent that served the broadest CX audience: the CX Policy Assistant. This agent pulls from Faire’s most actively used internal content like their help center, policies, and operational guides that agents rely on every day.

‍

Carly enabled both auto-verify and auto-unverify rules for this collection, setting conditions around card view frequency and positive engagement signals. The defaults proved to be a solid fit for most of the CX content, requiring only a few small adjustments after reviewing early results.

2. Using the Quality Log to Stay in Control

Carly built a lightweight weekly QA habit. Every Monday or Tuesday, she reviewed the previous week’s quality log — spot-checking five to ten cards, with particular attention to anything flagged by the Knowledge Agent as low confidence.

‍

“I love having those confidence levels — low, medium, high. I tend to check the low confidence ones more. 

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

This approach gave Carly enough visibility to catch edge cases early without turning QA back into a part-time job. When a rule fired incorrectly the log showed exactly which rule triggered it and made the fix straightforward.
‍

“The Quality Log is really helpful for somebody like me. I'm not an engineer. I'm not an AI-expert, but a lot of it makes sense. And if the automation isn't doing what I want it to do, it's easily fixable. So yeah, I feel like the QA part of it has been really easy on my end. Instead of having to spend six hours a week, I'm only spending like less than an hour a week kind of like qa-ing that stuff. So tons of time saved for me.”

— Carly Byrge, Associate Program Manager of Ops Enablement
‍

3. Refining Rules Based on Real Content

One early lesson: rules written for broadly structured content sometimes need tuning when content gets specific. A time-bound unverify rule designed to catch outdated cards inadvertently flagged a tariff update card that included a historical timeline of changes.

‍

“It explained to me why it did that. I saw it was because of the time-bound rules, and I knew exactly which one I had to edit. That hasn’t happened again.”

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

The experience also changed how Carly thinks about authoring. She now considers verification rules when writing new cards — flagging when date-specific language might conflict with existing rules so the content stays stable.

4. Expanding to a Second Agent

After getting comfortable with the CX Policy Assistant, Carly extended quality automations to a second Knowledge Agent supporting Faire’s marketplace operations team. The content there is more procedural and less subject to change than CX content, so the rules are configured more conservatively. The automation handles the bulk of routine verification, with only occasional manual review needed.

Results: Restored Trust and Regained Time

⏱️ 5+ hours back per week by reducing verification time

📈 20% gain in Trust Score (From 76% to 96%)

🔓 Unlocked new potential productivity gains with AI

‍

With quality rules enabled on only a single Knowledge Agent, Faire experienced a meaningful shift: from hours a week spent manually combing through unverified cards, to spending less than an hour a week spot-checking the Knowledge Agent’s decisions in a transparent quality log.

The impact wasn’t just time saved — it was what that time made possible. For the first time since Carly took on her role, she had space to focus on things beyond content cleanup. Now she can spend time thinking strategically about new opportunities where Guru and AI can streamline operations and increase productivity.

‍

“Yes, I 100 percent feel like Guru is in a better place since enabling the quality automation. I feel like we’re in a much better place where I’m not stressing out about it. Our team’s not stressing out.”

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

The broader signal from the team supported what their new 96% Trust Score (previously 76% before enabling quality rules) already showed. Users who had been skeptical began re-engaging with Guru, both in the web app and through a Slack channel where the CX Policy Assistant surfaces answers automatically. Feedback from the CX team and across the org was 

‍

“I have heard feedback from a couple CX people who have explicitly told me that they feel like Guru just feels better.”

— Carly Byrge, Associate Program Manager of Ops Enablement

What’s Next

“Not only am I excited — I call myself the Guru nerd at Faire — but we’re getting people who don’t even have stakes in it to see the value. That’s what we’ve been needing for the past couple of years. It got people really interested. People that don’t even work with Guru, that are senior managers that don’t necessarily know what it is. It had people really impressed.”

— Carly Byrge, Associate Program Manager of Ops Enablement

Before quality automations, Faire was held back on what they could do with Guru.  Now, with a stable foundation of a healthy knowledge base, Faire has freed themselves up to take advantage of all the other time-saving and efficiency-improving capabilities that Guru has to offer. They plan to improve existing Knowledge Agents, scope opportunities for new ones, and enable the entire team on chat, research, and insights abilities. 

‍

Faire went from a knowledge base held together by one person’s manual effort to a governed, trusted system that the team can actually rely on. And they’re just getting started.

‍

“I feel really excited and hopeful that in the next six months we’ll just see more and more improvement.”

— Carly Byrge, Associate Program Manager of Ops Enablement

‍

‍