Bobsled

This is good analysis from Bobsled. https://lnkd.in/g2m7Yipf

"Why not just build agentic analytics in the data platform you already own?" Answer: We've built a context layer that learns with you. Our learning agents watch real usage, flag where the model is falling short, and surface learnings your data team can apply in two clicks — no semantic view edits, no redeploys. Your agents stay in the loop. Your data stays in your platform. And those urgent Slacks from execs "asking for some quick help" stop showing up. Bobsled Solution Engineer Matt Ballantine walks through it in this short demo.

For modern software and AI companies, the in-app dashboard is just one piece of the analytics experience customers expect. Benjamin T. at Aampe saw it early. His customers want their data in their own warehouses, feeding their own tools and models. Building the pipes to get it there would have cost his team a quarter of engineering. He had better things for them to do. "That's engineering time I'd rather spend shipping the features only we can build," Ben told us in a new interview. Ben breaks down how he scoped the build, the edge cases that kept surfacing, and what customers are doing once the data actually shows up where they work. Link to interview in comments 👇

We analyzed 770 data and analytics companies to see how they're deploying AI-powered experiences. 31% have shipped an AI feature. That's past the tipping point on the adoption curve, and the next 6-9 months are when the market really tilts. A few things stood out from the research 👇 1. Two paths are emerging. 23% have built conversational analytics into their product. 16% have launched MCP servers or other AI partnerships. Only 8% have done both. 2. The gap between big and small companies is 10x. 85% of firms with 5K+ employees have shipped AI. Under 50 employees, it drops to 9%. Some of that is resources, but a lot of it is existential pressure. Big companies feel it. Small ones don't — yet. 3. Text-heavy verticals are out front. Healthcare, legal, and commerce lead the pack. No vertical has cracked 50% yet, so everyone still has a window. 4. "Chat with your data" is quickly becoming the killer app. Embedded analytics and AI search are showing up in nearly every roadmap we reviewed. 5. The real endgame is agents. The difference between a commodity data feed and an irreplaceable product is going to come down to how much intelligence gets baked into the data layer itself. If you're in the 69% that hasn't shipped, the next 6-9 months are decision time. Full research in comments. If you're at a data company, where are you on this? Building your own AI experience, plugging into someone else's, or still figuring it out? #dataproviders #daas #ai

Evaluating more than 50 third-party datasets a year? Tired of the legal, compliance, and data engineering overhead just to get an initial look at a dataset — before you even know if it's valuable for your team? Something really interesting is emerging in how firms - particularly those in finance, on the buyside - are using Bobsled. They're creating an independent data evaluation layer — a buffer zone between their production data estate and the entire third-party data landscape. No ingestion into the core stack. No compliance gauntlet. No engineering tickets. Load the data into the buffer zone and start working. Query it in natural language. Understand coverage and relevancy. Bring your own notebooks and skills. Get to “Go” or “No Go” before anyone in legal even knows there was a dataset to look at. Ultimately, we're seeing this enable firms to eval 10x more data with significantly less overhead. A transparently massive win, by any standard. Question, tho: is suggestive of an emergent reference architecture for data eval? Here's an overly-futuristic, obviously-vibed visualization of the emerging architecture we're seeing:

Working on AI driven data analytics for the last few months, I am beginning to realize that the data community has done a disservice to semantic metadata. Because software systems till now could only properly express and process structured metadata, we limited ourselves to table/column definitions with some (largely failed) attempts at expressing metrics in a structured format - yaml or json. But that's merely scratching the surface of the vast amount of contextual information that accompanies data. This context is usually embedded in unstructured formats - PDFs, FAQs, internal docs, chats, customer service logs, meeting notes etc. Often, this context covers years of historical understanding of complex domain logic. Data analysts spend months during their onboarding process to acquire this context before they can extract any meaningful value from Data. Sure, you could surface this rich metadata in your favourite MDS data catalog tool but there was no way to actionize this knowledge. Until now, that is. AI agents with tool calls open up a completely different way to utilize this type of unstructured semantic information. For instance, a properly configured AI agent could read a user document (pdf) using RAG (or even a simple document search) to understand the definition of a metric (with all its complex nuances) and then execute a SQL query that incorporates this definition. Agents are no longer bounded by information structure. IMO, we will see a shift from semantic models to gathering and organizing unstructured knowledge for AI analytics. Structured semantic models have their place in this architecture but in a very limited way. Huge isn't even the word that describes the possibilities this unlocks.

We’re incredibly excited to welcome Matt Neligeorge as Bobsled's new Chief Scientist. Matt has spent his career at the forefront of applied AI and machine learning. He joins Bobsled from Seek AI (acquired by IBM) where he built agents for reasoning over structured data, and previously helped develop the core schema management service at AWS. He holds a PhD in astrophysics from Berkeley and degrees from Cambridge and Harvard. At Bobsled, Matt will lead our work to help the world’s leading data product teams turn their data into trusted and governed agentic experiences for their customers. We’re thrilled to have him on board and can’t wait to share what’s coming next. #agents #agenticanalytics #dataproducts #AI #ML

Is AI a threat to data companies? According to Matt Ober — one of the most plugged-in investors in financial data — the answer is yes… but it's also a massive opportunity. AI is doing two things at once: 🔹 Expanding the market: AI dramatically lowers the cost of using data. That means more users, more use cases, and an order-of-magnitude increase in demand. 🔹 Eroding traditional moats: It also lowers the cost of building data products. What once required thousands of people for QA and support can now be built with 100. The winners will be the companies that can move the fastest—whether that’s building, buying, or acquiring new capabilities. Matt’s full take is one of the clearest visions we've heard so far on the future of data in the AI era. Link in comments.

I’ve spoken with hundreds of people in the data product space this year — and the number of production text-to-SQL deployments in enterprises is still close to zero. The reason isn’t a mystery - accuracy isn’t high enough. The technology is well understood - you need well structured data, context in a semantic model, and governed access to an LLM, but very few companies are honest about the work required to deliver accuracy even after you have all of these in place. Initial demos look amazing, then start hallucinating wildly once you ask more questions. What we’re seeing in practice is that getting to accuracy is an iterative process, combining 3 things: - Well-structured, contextually rich data. - Subject-matter experts who understand the data, what questions to ask, and what the right answers are. - Hands on AI and context engineering to adjust context, prompts, and data models. I’ve yet to see generic agents over random data produce any form of real value, however we are starting to see real success in developing AI experiences tailored to specific data products, which is absolutely where the market is going over the next year.