“The value of Guru can’t be overstated, especially in a world like ours where there are so many spinning plates. Making our job simpler is a huge return.”

‍ — Christian George, Associate Product Manager, Steno

‍

Company Overview

Steno is a legal services and tech company that provides court reporting and litigation support services to law firms across the United States. A core part of their operation is a team of 100+ remote videographers who sit in virtual deposition rooms alongside attorneys, managing the technical infrastructure that keeps high-stakes legal proceedings running smoothly.

‍

Christian George, now an Associate Product Manager at Steno, saw an opportunity to use Guru’s Knowledge Agents to scale support for these videographers — without scaling headcount. What started as a support efficiency play became the proof of concept that earned leadership buy-in for AI across the organization.

The Challenge: 200 Questions a Day, No Scalable Answer

Steno’s videographers work in one of the most demanding real-time support environments imaginable: a virtual room with lawyers who are laser-focused, under high-pressure, and intolerant of technical delays. When something goes wrong, videographers need answers immediately — not in five minutes, not via a ticket queue.

Problem 1: Speed Is Non-Negotiable

The defining constraint of Steno’s support challenge is time. Videographers sit in live deposition rooms with attorneys who are actively questioning witnesses. Any technical hiccup — audio issues, Zoom configuration problems, exhibit management questions — needs to be resolved in seconds, not minutes.

‍

“One of the defining features of the work of a videographer at Steno is that we have to have them sit in a remote meeting room with lawyers who are very focused, and they don't want to waste any time.”

— Christian George, Associate Product Manager

Problem 2: Volume Growing Faster Than Headcount

Steno’s internal support team fielded roughly 200 questions per day from videographers in the field, all flowing through Slack. As Steno grew, so did the question volume — and the company faced a choice: continue scaling support headcount linearly, or find a way to handle the answerable questions automatically.

‍

“Those kinds of help questions were coming in more and more rapidly as Steno continued to grow, and it would have required us to continue to increase our support headcount.”

— Christian George, Associate Product Manager

Problem 3: Half the Questions Were Already Documented

When Christian audited the support channel manually, he discovered that roughly 50% of questions had documented answers — they just weren’t being surfaced fast enough. The other 50% genuinely required a human to intervene (jumping into a room, taking over a session). This gave him a clear, data-driven target: if an AI agent could handle the documented half, it would free the human team to focus entirely on the problems that actually required them.

‍

“I actually went through and I scanned our channel where we have all these help questions and looked at what required somebody to actually jump in versus an answer we have documented. It turns out that number is about fifty percent.”

— Christian George, Associate Product Manager

The Approach: Built in Under a Week, Embedded in the Workflow

Christian’s approach was fast, practical, and designed to minimize change management. He configured the knowledge agent prompt himself, connected it to the right sources, and embedded it directly into a Guru-centric Slack channel, where videographers go first to ask questions.

1. Connecting Sources for Complete Coverage

The Vid Help Assistant draws from sources including Google Drive folders, Google Sheets, Guru collections, and public websites like Zoom’s help center to ensure the latest information is always feeding into the knowledge agent.

‍

The Zoom integration was particularly strategic: since videographers work primarily in Zoom, connecting Zoom’s public help center as a live source meant troubleshooting answers stay current automatically — without Steno having to manually maintain their own Zoom documentation.

‍

“It saves us a lot of headache having to keep our own Zoom troubleshooting cards active and up to date when we can rely on the provider themselves.”

— Christian George, Associate Product Manager

2. Context-Aware Prompt Design

Christian’s prompt design reflected a deep understanding of his users’ constraints. He told the agent that video specialists do not have Zoom admin permissions — preventing answers that instruct users to change account-level settings they can’t access. He configured responses in detailed, bulleted formats for quick scanning during live depositions. And he added a thumbs-up/down reminder to every response to drive the feedback loop that would improve the agent over time.

3. Minimal Change Management

Rather than asking videographers to learn a new tool or change their habits, Christian embedded the agent directly into Slack where they already go to ask questions. The instruction was simple: start here first, thumbs up if it helps, thumbs down if it doesn’t. If you need someone to physically take over, escalate.

‍

“We essentially told the team, unless you know you need somebody to come take over the room for you, start in the Guru channel first and give it a thumbs up if it helps you, and a thumbs down if it doesn’t.”

— Christian George, Associate Product Manager

Results: Better Self-Service = More With Less

• 🚀 48% Reduction in support tickets
• 🔥 100% of daily questions handled autonomously
• 📈 2x Productivity — one person doing work scoped for two

The Vid Help Assistant achieved a 48% reduction in support tickets — just shy of Christian’s explicit 50% target. The human support team now only needs to intervene when a physical presence is required in the room, even as volume continues to increase month-over-month. The productivity gains were concrete: a role originally scoped for two hires was filled by one, with the agent handling the answerable volume.

But the deeper impact, as Christian described it, wasn’t just about ticket counts. Freeing the human team from repetitive, answerable questions lets them focus on harder problems, respond faster to genuine emergencies, and maintain better situational awareness across the entire operation.
‍

“The best thing about being able to deflect half of our questions to an agent is that it frees up our actual human support team to do the harder stuff, to be more on top of communications, respond faster, and generally just have a better eye on the operation because there are things that they don’t have to think about because they know they’re taken care of.”

— Christian George, Associate Product Manager

What’s Next: From Support Agent to AI Strategy Lead

The Vid Help Assistant’s success proved to Steno’s leadership that Knowledge Agents deliver measurable ROI — and that proof of concept has opened the door to much broader ambitions. Christian is now charged with leading efforts to improve Guru AI Agent strategies on the Product side of Steno and spinning up additional Knowledge Agents across the organization.

His advice to anyone starting out is characteristically practical: think of your agent as a new coworker, and write the prompt as if you’re giving instructions to someone supporting your team while you’re not looking. If you had a magic genie, what would you tell them?

‍

getguru.com - Your AI Source of Truth

‍