ThoughtSpot is an AI-powered analytics company that enables organizations to access, analyze, and act on data through natural language search, agentic AI, and business-focused semantic intelligence.
Many companies have run AI pilots, but relatively few have successfully scaled AI into production. In your mind, what accounts for that gap?
There are several reasons. First, agentic AI is fundamentally different from traditional machine learning. Machine learning is deterministic, while agentic AI is probabilistic, and it involves giving agents permission to take actions on behalf of humans. That represents a generational shift, not just a technological one.
Organizations often focus on the technology itself and underestimate the mindset shift and change management required. The companies that succeed tend to invest not only in a strong data foundation but also in preparing their people and processes for this new way of working.
So organizations struggle because AI adoption is not being approached systematically across the business?
I think that’s a valid observation. One of my predictions this year is that boards and leadership teams with higher AI literacy will achieve higher returns on AI investments because they approach it holistically rather than as a collection of experiments.
Misaligned expectations can create problems. People hear terms like “vibe coding” and assume AI can instantly solve complex infrastructure challenges. In reality, success requires a solid data foundation, trusted semantics, and effective change management. Companies such as Hyatt, Navan, Verisk, and Elevance demonstrate that even complex or regulated organizations can do this successfully.
Data quality is often described as the biggest challenge in AI. How does ThoughtSpot’s Spotter Semantics change the equation for your customers?
Generative AI is probabilistic. It predicts the next word or number, which means it must interpret what a user actually means when they ask a question. If someone asks, “What are my sales this quarter?”, the answer may depend on whether they mean a fiscal quarter, a calendar quarter, revenue, commissions, or something else entirely.
Spotter Semantics provides that business context. It understands the language, definitions, and intent behind questions so that users receive consistent and trusted answers. Without that context, large language models can generate inconsistent or inaccurate responses. Our semantic layer has been proven in the market for over a decade and continues to evolve for AI-driven use cases.
Don’t organizations need near-perfect accuracy before they can trust agentic AI systems?
The first question I would ask is whether we expect AI to be more accurate than humans. Humans make mistakes all the time. After a long flight, I know I’m more likely to make spelling errors or transpose numbers. So it’s important to keep expectations realistic.
At ThoughtSpot, our semantic layer helps us achieve accuracy levels of at least 95%. Beyond that, transparency becomes critical. Users need visibility into how an answer was generated, and there should be a human-in-the-loop process to refine results when necessary. Combining transparency, context, and feedback mechanisms is what ultimately helps organizations reach the confidence they need.
I believe organizations should adopt a crawl, walk, run approach. Initially, an agent might simply identify a problem and recommend an action. That already creates value without requiring organizations to hand over decision-making authority. As trust grows, companies can gradually grant more autonomy. An agent might send alerts, initiate workflows, or communicate with managers. Eventually, organizations can allow agents to take more direct actions, but only after they are confident in both the technology and the associated change management processes.
Looking ahead over the next 12 months, what are your priorities at ThoughtSpot?
Innovation is moving faster than many organizations can absorb. One of our priorities is ensuring customers view us as a partner rather than simply a vendor.
That means helping them rethink their business models, workflows, and operating processes while ensuring their data and AI strategies remain tied to measurable business value and outcomes.
