Episode 10 now available!

Ep 10 - Building Autonomous AI for Legal Contract Review

Shantanu Singh, Co-Founder, CEO of Obviate.ai

​Building Autonomous AI for Legal Contract Review​

Shantanu Singh didn't follow a straight path to becoming a founder. Cell biology researcher → prosecutor → AI startup co-founder. But every "dot" he dropped along the way prepared him for this moment. When ChatGPT launched, it felt like discovering fire.

In this episode of the Startup Wisconsin Podcast, Shantanu breaks down what makes Obviate.ai different from typical AI wrappers, why most AI products are automated rather than truly autonomous, and how Milwaukee's small but mighty AI founder community creates disproportionate impact.

Episode Topics

• The founding story of Obviate.ai
• How autonomous AI agents differ from traditional Gen AI
• Patent-pending legal reasoning technology
• The "Guided Autonomy™" approach to contract management
• Challenges of building AI for the legal industry
• Enterprise security and data privacy in legal tech
• The future of AI in contract negotiation
• Lessons for founders building specialized AI tools

​

→ Listen to the Interview​

​

Key learnings

Autonomous AI Requires Explainability for Knowledge Work
‍True autonomous AI can't be 100% hands-off in knowledge work because professionals are paid for their expertise and judgment. Shantanu's philosophy is that autonomous systems must balance two constraints: doing work without human intervention while maintaining explainability. When lawyers deliver a memo or brief, they must explain their reasoning. Building AI tools within this constraint, autonomous but explainable, creates genuinely useful products rather than black boxes that professionals can't trust or defend.

Ideas Mean Nothing Without Systematic Evaluation
‍Having lots of ideas is valuable, but Shantanu emphasizes you need your own system for evaluating whether an idea is actionable, viable, and whether you have the capability or time to execute it. He describes his approach as "dropping dots" throughout your life. Trying new things, getting outside your comfort zone. Because you can't predict how experiences will connect until they do. The connective tissue strengthens over time, but only if you're disciplined about which dots to actually pursue.

Margin of Safety Determines Margin of Error
‍Shantanu distinguishes between margin of safety and margin of error in startup ecosystems. Margin of safety is about whether mistakes are reversible. If you can really screw up and it's not an irreversible decision. When your margin of safety is wide, your margin of error for making bad bets is high. But when there's limited time and mistakes mean game over, the margin of error becomes razor-thin. Wisconsin's ecosystem needs to focus on creating wider margins of safety so founders can experiment without existential risk.

AI Wrappers vs. True AI Products
‍Most AI applications are automated, not autonomous. They give you a recipe but you have to bring the ingredients through prompts and iterative interactions. Shantanu built Obviate with four guiding principles: no legal education required, no prompting needed, automatic track changes generation, and interoperability outside any single ecosystem. True AI products abstract complexity from users and make multiple API calls behind the scenes to deliver outcomes, not just streamlined workflows.

Rapid Prototyping De-Risks Your Ask for Help
‍Shantanu's advice to aspiring founders is that you should be able to prototype any idea in three days or less. Tools like Gemini, Claude, and ChatGPT provide unlimited education and customer support for $20/month. They'll teach you GitHub, how to commit code, and debug it visually. Even if you've never coded before. Showing something tangible demonstrates commitment beyond a transient concept and helps others make good bets on supporting your idea.

Good Taste Is AI's Remaining Frontier
‍Shantanu quotes a Hollywood director who said AI won't wholesale replace filmmakers because AI can't have good taste. It lacks the ability to be situationally aware of what constitutes quality. Humans who can develop and project good taste as a user experience, humanizing what they're trying to accomplish with AI, will be the winners. This requires training, study, and exposure. The future skillset isn't technical proficiency but taste-driven curation of AI outputs.

Wisconsin Needs Ecosystem Infrastructure, Not Just Capital
‍The challenge in Wisconsin isn't a lack of good ideas or even access to capital. It's the absence of infrastructure to support dreamers and entrepreneurs. Shantanu believes you could find product ideas just by sitting in any Wisconsin university lecture hall for a week. What's missing is the institutional pipeline from engineer to founder, the connective tissue that exists in places like San Francisco or Chicago, and most critically, learning how to make good bets on which founders and ideas will succeed.

​

What is the Startup Wisconsin Podcast?

A new show where you can learn about Wisconsin's growing tech scene. Hear the stories of startups, founders, investors, and the incredibly generous people making it happen every single day.

​See All 10 Episodes​

​

Upcoming episodes 👀

• Jillian Bichanich (Growing Cubbi and Launching a 2nd Startup)
• Laura Strong (Starting Valency Fund)
• Jessica Silvaggi (UW Milwaukee Research Foundation)
• Michael Ceely (Coaching for Entrepreneurs)
• Ben Camp (Recovery.com)
• Russ From (Wayfi Wireless)
• Angela Damiani (Midwest Founders Community)
• Joey Friedli (Trashpoint)

​

Subscribe to the podcast

Too busy right now but want to listen later?
Let's make this easy for you.

​Subscribe on YouTube​

​Follow on Spotify​

​Follow on Apple Podcasts​

​

​

​

Have questions?
​We've got you. Ask away!
jacob@startupwi.org

​