Kanti Prabha, Co-Founder and Chief Product & Technology Officer, Sirion – Interview Series

Kanti Prabha, Co-Founder and President, Sirion, is an enterprise technology leader with more than two decades of experience spanning product development, legal technology, intellectual property, operations, and SaaS. As a member of Sirion’s founding team, she has played a central role in shaping the company’s product vision, design, and development, serving in leadership positions including VP of Product Management and Chief Operating Officer before becoming President. Her expertise includes building enterprise-grade B2B SaaS products, product strategy and management, customer success, enterprise sales, and scaling teams and organizational culture. Earlier in her career, Prabha was the first employee and a member of the founding team at UnitedLex, where she developed technology-driven solutions for corporate legal departments and managed intellectual property initiatives, following previous experience in patent services at Evalueserve. She also serves as an advisor to akirolabs and previously advised Clarisights.

Sirion is an AI-powered contract lifecycle management (CLM) company that helps enterprises manage contracts from creation and negotiation through ongoing performance, obligations, and value realization. Its platform combines conversational AI with purpose-built AI agents that can centralize and structure contract repositories, answer questions about contract terms, generate and review agreements, identify risks, propose redlines, and monitor obligations and service-level agreements. Sirion is increasingly positioning its technology around agentic contracting through its agentOS platform, which enables organizations to deploy pre-built contracting agents as well as create customized agents connected to enterprise data and workflows. The platform is designed for large organizations seeking to turn contracts from static documents into structured, searchable sources of operational intelligence while maintaining governance and human oversight.

Your path to founding Sirion was unconventional: you studied chemistry, moved into intellectual property law, worked on patents, and then began seeing contracts differently while working with Hewlett-Packard. What did you recognize in those early experiences that ultimately convinced you there was an opportunity to build Sirion? 

I did not grow up thinking I would build a company around contracts. I studied chemistry, but realized fairly quickly that I was more interested in solving business problems than spending my career in a lab. That eventually took me into intellectual property, where I worked on patents and later led teams developing technology and legal solutions for large enterprises.

The real turning point came when my Sirion co-founder Ajay and I started working with HP. They had thousands of contracts that needed to be organized and managed following a large acquisition. We initially saw it as a repository problem, but once we started looking closely at the contracts, it became obvious that storage was only a tiny part of the opportunity. Contracts told you what a company had bought, what it had sold, what it had promised, what its suppliers had promised, what it was paying, where it had risk and where value was potentially being lost. Yet almost none of that information was easy for the business to use. That was the insight behind Sirion. Contracts were a largely untapped source of business intelligence.

You’ve described contracts as a form of “liquid intelligence” within an organization. What valuable information is typically trapped inside contracts, and how can AI make that intelligence useful to teams beyond legal, including finance, procurement, sales, and operations? 

Contracts are one of the few places where a company’s intentions actually become commitments. They tell you what you agreed to pay, what a customer agreed to buy, when prices can change, what service levels must be met, when an agreement renews, what happens if someone underperforms and who owns which obligation. That is incredibly valuable information, but historically it has been trapped in documents that most people only open when they need to negotiate something or when something has gone wrong.

AI changes that because it allows us to turn unstructured contract language into information the rest of the business can actually use. Procurement can understand supplier obligations and pricing. Finance can see commercial commitments and potential leakage. Sales can understand renewal terms and customer commitments. Operations can monitor whether what was promised is actually being delivered.

This is what I mean by making contract intelligence liquid. Instead of information sitting inside a PDF, it can flow into the decisions and systems where it is needed most. Instead of the contract being the end of a transaction, liquid intelligence allows it to be an active source of context for how a business operates.

Sirion has evolved from contract lifecycle management software into an increasingly AI-driven platform. How has the emergence of large language models changed what is technically possible with contract data compared with the machine learning approaches available when you started the company?

When we started Sirion, AI in contracting was primarily about extraction and classification. Machine learning could become very good at identifying a clause, a date, a party name or an obligation, but you had to train models for relatively specific tasks.

Large language models dramatically expanded the reasoning layer. Now a system can understand a question expressed in natural language, interpret the context around a clause, compare language against a company’s preferred position, explain why something represents a risk and suggest what someone should do next.

But I don’t see this as LLMs replacing everything that came before them. Contracts are too complex for that. At Sirion, we use a multi-model approach that combines large language models with specialized models and deep contract intelligence. Different models are good at different things. That combination is important because enterprises don’t simply need an AI that can generate convincing language. They need precision. The most exciting change is that we can move from extracting information from contracts to reasoning over that information and, increasingly, acting on it.

Sirion is now introducing specialized AI agents across tasks such as contract extraction, issue detection, redlining, and searching large contract repositories. How do you envision these agents working together, and how far should enterprises ultimately allow AI to act autonomously during the contracting process?           

I don’t think the future is one giant AI agent that does everything. Contract development involves many specialized tasks, and each requires different context and different levels of judgment. One agent might extract terms from thousands of agreements. Another identifies language that deviates from policy. Then a redlining agent can suggest a precise change based on the company’s playbook. The real power comes when those agents can work together around an outcome rather than forcing the user to manage every individual step.

For example, someone might ask, “Can we accept this supplier agreement?” The system should be able to identify the most relevant terms, compare them with policy, surface the issues, find precedent, and recommend changes, all while explaining what it did and why.

Where autonomy stops should depend on risk. I am comfortable with AI autonomously handling many repetitive, rules-based actions. But a novel liability provision or a strategically important negotiation may require human judgment. Autonomy’s end goal should be to remove unnecessary work while preserving human control where accountability and relationships matter.

Contracts can contain some of a company’s most commercially and legally sensitive information. You’ve said Sirion would rather delay a product launch than compromise on security. As generative AI becomes embedded deeper into enterprise workflows, what security or privacy shortcuts concern you most?

Contracts can contain pricing, intellectual property, employee information, acquisition details, strategic commitments and extremely sensitive information about customers and suppliers. Companies need to know exactly where that information is going when an AI system touches it, how it is being processed, who can access it and whether it is being used to train another model.

There is enormous pressure right now to ship AI features quickly, but the shortcut that concerns me most is treating enterprise data as though it were ordinary training material. I understand that pressure, but enterprise AI cannot be built on “we’ll fix governance later.” Security has to be part of the architecture from the beginning.

That means strong access controls, encryption, auditability, clear data boundaries and very deliberate decisions about how third-party models are used. At Sirion, we have absolutely made decisions where security slowed us down, and I am comfortable with that tradeoff. In enterprise software, particularly when you’re handling contracts, trust takes years to build and one careless decision to damage.

One persistent challenge with generative AI is that a response can sound authoritative even when it is wrong. In contract intelligence, where a subtle error can have financial or legal consequences, how do you design AI systems that users can verify and trust? 

This is one of the most important differences between consumer AI and enterprise AI. If I ask an AI assistant for a restaurant recommendation and it gets something wrong, the consequence is limited. If a legal team asks whether 5,000 supplier contracts contain a particular termination right and the system confidently misses 200 of them, that is a very different problem.

We design for verification. If the AI tells you something about a contract, you should be able to trace that answer back to the underlying language. If it recommends a redline, you should understand what policy or playbook position informed that recommendation. Explainability is fundamental to adoption because trust comes from making AI inspectable.

Sirion emphasizes AI responses grounded in an organization’s own contracts and precedents rather than relying solely on general-purpose models. Do you think the next major phase of enterprise AI will be increasingly driven by systems that combine foundation models with highly specialized proprietary company data?

Absolutely. Foundation models are extraordinarily capable, but they don’t know how your company personally operates. They don’t automatically know which concessions your legal team approved last year, how you structure pricing with a strategic customer, which suppliers repeatedly miss service levels or when your organization requires an escalation. That context is where enterprise AI becomes genuinely useful.

Combine contracts’ powerful record of decisions and commitments between organizations with unique company policies, negotiation playbooks, historical agreements and operational data, and an AI system can move from giving a generic answer to giving an answer that reflects how your business works. I think the competitive advantage in enterprise AI will increasingly come from this context since most enterprises already have access to similar foundation models. The differentiator will be how effectively they can connect those models to trusted proprietary information and give them enough context to reason accurately without compromising security.

You have spent more than a decade building Sirion and have held roles spanning product, operations, and company leadership. What did you learn about scaling enterprise AI software that you would not have understood during the company’s first few years?

One thing I’ve learned is that building the technology is sometimes the easy part. The friction comes in when you begin changing how a large organization works. In the early years, you naturally focus on whether the product can solve the problem. At enterprise scale, you realize that a technically excellent answer is only useful if people trust it, it fits into their existing workflows and it can handle the enormous complexity of a global organization.

AI makes that lesson even more important. You can build an impressive demo very quickly today. Building something an enterprise is willing to trust with millions of data points and workflows is different.

As AI makes it possible to continuously analyze obligations, risks, pricing terms, renewals, and performance across thousands of agreements, could contracts eventually become a real-time operational data layer for the enterprise rather than documents that are primarily consulted when something goes wrong?

I think that is exactly where we are headed. For a long time, the contract has been treated as the finish line. Everyone negotiates, signs, celebrates and then the document goes into a repository. Six months later, someone opens it because there is a dispute, a renewal or a question about what was promised. That model doesn’t make sense when AI can continuously understand what is happening across the contract portfolio.

In an ideal world, imagine knowing that a supplier has missed an SLA before the next invoice is approved, or seeing that a pricing adjustment should have occurred three months ago. Imagine a sales team being alerted to an upcoming renewal alongside the customer’s negotiated commitments, or finance seeing commercial obligations reflected directly in its systems.

Contracts define relationships between companies, so contract data should be connected to the systems where those relationships are actually managed. I believe the contract will increasingly become an operational data layer that helps the enterprise understand not just what it agreed to, but whether those agreements are delivering the value everyone expected.

Looking ahead, what do you believe contract management will look like once AI agents become substantially more capable? Does the traditional contract lifecycle management category eventually disappear into a broader enterprise intelligence layer, and what role do you see Sirion playing in that transition? 

I hope to see the boundaries of CLM become much less visible over time. Today, users still think in terms of steps: I need to create a contract, review it, get approval, sign it, store it and then manage it. As agents become more capable, people shouldn’t have to orchestrate all of those systems and workflows themselves. They should be able to express an intent and let specialized agents handle much of the work behind it.

But the bigger change happens after signature. If AI can understand every obligation, commercial term, risk and relationship across millions of agreements, contract intelligence becomes useful far beyond legal. For me, that’s the exciting part: contracts finally becoming everyone’s business.

So I don’t think contracts disappear. Quite the opposite. They become more important because they provide trusted context for enterprise AI. Our goal at Sirion is to make that intelligence usable wherever work happens. CLM may eventually feel less like a standalone destination and more like an intelligence and execution layer connecting the commitments a company has made with the actions it takes every day.

Thank you for the great interview, readers who wish to learn more should visit Sirion.