Case study · Spot AI
Mash built Spot AI a knowledge system that could read video and images
From Mash’s knowledge-systems era, 2025 to early 2026. Before Mash became a Personal Relationship Manager, it built custom knowledge systems for a small number of technical B2B revenue teams: bespoke pipelines into each company’s own scattered sources, and answers delivered where those teams already worked. This is a record of that work.
It is not the product on sale today. What carried over is the engineering underneath — reading years of messy, scattered history and making it answer questions reliably. Mash now does that for the people in your own network instead of for a company’s product documentation. See what Mash is now.
- 45%of previously unanswerable questions answered using video and image data
- 90%+of questions asked as an image understood well enough to answer
About Spot AI
Spot AI helps businesses see, understand and manage their physical operations. Problems in the physical world surface by surprise and finding them requires visual context, so Spot AI’s product — and its support material — is built around video.
The problem, and two things that made it unusual
Like other technical revenue teams, Spot AI’s GTM organization was searching across training material, Notion, Google Slides, ticketing systems and Slack, then escalating when that failed — delaying answers for prospects and customers, and interrupting the most experienced people in the business. The team estimated that sales engineers and solutions architects were repeating themselves on around 84% of the relevant queries.
Two things about Spot AI did not fit any existing knowledge tool:
- The questions were pictures. People asked for help by posting an image of what they were looking at — a product feature, an issue, a situation — usually with no accompanying text for a system to work from. On evaluation, this covered roughly 45% of the relevant questions being asked.
- The answers were video. Spot AI had a large library of demo, integration and tutorial video that was the best source of truth it had, and no knowledge system could read it.
Spot AI is an AI-first company whose leadership expects teams to use AI to operate more efficiently. The team had spent over a year looking for something that worked, and reviewed five options before choosing to build with Mash.
What Mash built
Mash built two things that did not exist in the platform before this engagement.
The first understood images both as questions and as knowledge: a screenshot posted in a channel became something the system could reason about and answer, and images inside the company’s own material became part of the knowledge base.
The second read video. Mash transcribed it, parsed the frames, and time-stamped the related context so an answer could link to the exact moment in a video held on an internal system. Processed video and imagery then fed into the answers Mash gave.
The results
Around 45% of questions the system previously could not answer became answerable using the video and image data, and more than 90% of questions asked as an image were understood well enough to be answered. The sales engineering team adopted it first and began sharing answers outward to the wider organization.
Next: Rootly — Answers in Slack instead of a queue behind two subject-matter experts. Or see what Mash is today.