UPDATE
DreamLegal In Conversation with Kanti Prabha Co-Founder, Sirion
Sirion co-founder and President Kanti Prabha explains how agentic AI differs from generative AI in contract management, why the vendor is responsible for every AI output on its platform, how CLM and legal AI fit together, how to measure CLM beyond time saved, and her prediction for the next two to three years.
Kanti Prabha is co-founder and President of Sirion, the contract lifecycle management (CLM) company she started in 2012, where she leads product and technology. In this DreamLegal interview she explains the difference between agentic AI and generative AI in contract management, why she believes the vendor is responsible for every AI output on its platform, how CLM and legal AI fit together, and why she expects contract management to stop being human-driven within two to three years.
Key takeaways
Agentic vs generative AI: a generative AI assistant is "chatty" and answers your input; agentic processes "think, act, sequence, decide" and choose between pathways, which suits complex work like contract management.
Vendor liability: "the vendor is responsible for all AI that is presented through their platform." Outcomes should be predictable, repeatable and traceable, with the human as the final decision-maker.
CLM vs legal AI: CLM serves the whole business (pricing, deliverables, regulation for HR, finance, IT, legal and risk); legal AI serves lawyers in depth. They overlap and should work together in one interface.
Measure more than time saved: revenue, right spend, less value leakage, and both parties keeping the promises in the contract.
Her prediction: contract management will become system-driven, the boundaries between systems will blur, and contract data will become intelligence.
What is the difference between agentic AI and generative AI in contract management?
DreamLegal asked: From a customer's point of view, what is the key difference between agentic AI and generative AI? Where does an assistant stop, and where does agentic take over?
Kanti Prabha: I'll keep it very simple. A generative AI assistant is chatty. It takes your input, tries to understand it, responds to it and may give you some information back.
Agentic processes replicate human behaviour. They can think, they can act, they can sequence, they can decide, they can weigh options and they can choose different pathways.
"The closer you get to truly agentic processes, the more they mirror human behaviour and human decision-making."
That makes them far more useful when we are dealing with complex domains or categories, contract management for sure.
Which CLM module matters most: repository, authoring or contract intelligence?
DreamLegal asked: CLM was known for the repository from the start. Then redlining arrived as an active contracting use case, and then modules such as review and intelligence. In your experience, which module still holds the most value?
Kanti Prabha: All of them. The world is changing very rapidly, and contracts represent the pace of business, so they are changing rapidly too. Many years ago, contracts used to run for five years, seven years or ten years on average, and we would see some that were fifteen years long. Companies now want the shortest contracts, because the world is changing, tech is changing, and what a vendor can provide to you could be very different within three months.
So the pace of contracting, and the ability to write contracts in the fastest time possible, with the lowest risk and the tightest terms to protect your business, are very important. Knowing what you've already signed, living with that risk and fixing it matters from a repository perspective. Creating better, faster, tighter contracts matters just as much.
Most importantly, there is learning from contract intelligence: why we took certain decisions in the past, why we negotiated something in a particular way, how our business responds to contracts, and how contracts need to respond to the business. The intelligence piece is as important. I would say all three areas will become even more interesting, and even more autonomous, over the next several years.
DreamLegal asked: Is there a priority among these three? Does the buyer side more often start with one, which then leads to the other two?
Kanti Prabha: For agentic systems to be truly agentic and mirror human behaviour, they need to mirror human intelligence and domain intelligence. If you have a very strong knowledge base that is domain-centric, authoring will be better, the creation of new contracts and amendments will be better, and your repository will be better. So having that solid knowledge base and intelligence base is probably mission critical for being truly agentic.
Compare the tools in each area: contract repository software, contract negotiation software and contract review and analysis software.
CLM vs legal AI: is legal AI replacing contract management?
DreamLegal asked: Legal AI came in as a category in 2023 and has taken a lot of the legal tech attention. What has it changed for the CLM market? Has the definition of CLM changed, or is CLM losing importance against legal AI? Or is it still the most important legal tech tool for in-house teams?
Kanti Prabha: That's a great question, and I think it's important to go through it carefully. First, we're talking about contract management and the contract asset. The contract is a business asset that defines the relationship between two organisations. It is written in legalese, and one part of it is legal-heavy in nature, but many other parts contain pricing, deliverables and regulations, aspects that are relevant to different people within an organisation. It could be HR, delivery, tech, IT, finance, legal or risk. Everyone owns a piece of the contract and derives something different and pertinent to their own role.
Contract management as a category is designed to serve the business as a whole, with all of these personas and business objectives. Legal AI is a department-specific set of capabilities that makes lawyers superhuman, built around the problem statements they are working on.
Contracts overlap, so legal AI will always overlap in some form with contract management. Contract management is very horizontal, touching multiple persona types. Legal AI is focused on the needs of lawyers, which are deep. They need to work on contracts, but also be aware of regulatory research and do their normal matter management. So there is an intersection.
What is fascinating is that legal AI coming in strongly has given contract management an impetus for innovation. "Force" is the wrong word, but it has encouraged all of us to think more innovatively and innovate faster. Contract management, given that it is a multi-persona business asset and not just a legal asset, is here to stay. But we learn a lot from these systems that are born AI-native, and all of us should make sure we have the same technological architecture and principles within our products, no matter when we were born.
I also think these systems need to work together. Contracts are born and managed through their journey in CLM. Where external information, regulations or policies need to apply to those contracts, legal AI plays a great role by providing a single user interface. While you're drafting a contract and want some legal research, you can get it right there. These systems can talk to each other and exchange information smoothly. It's great for the user that agentic systems talk to each other and don't need the user to jump into multiple systems to get the job done.
Ultimately, we are here at the service of the user, and getting all of these data sources together in one place, in one common business process, is what benefits us and our users.
Sirion has announced this kind of link with a legal AI platform: Legora and Sirion link agentic AI to enterprise contract intelligence. Browse both categories: CLM software and legal AI software.
Where should a legal team with no legal tech start?
DreamLegal asked: Take a buyer who is new to legal tech and has no stack at the moment. Should they start with general-purpose legal AI and then move to more specific tools? Or should they look at the core work they do, find specific tools for it, and then broaden their AI use cases? Not from a vendor's point of view, but from yours.
Kanti Prabha: The key is to keep the user at the heart of the conversation. Make sure the user doesn't have too many conflicting tools, and that a clear technology picture has been drawn. IT teams do a great job making sure the right tool for the right need is the golden tool for a particular purpose. Look for a golden collection of tools that are truly complementary to each other and can talk to each other.
Users shouldn't have to go and look for their own tools. The company should provide that overall construct. That way you make sure the right breed of tools is chosen, and their interconnections are supported by the organisation instead of being left to users to figure out. I think that is also a very cost-effective path, and it allows you to invest in the best tools in the market and then connect them accordingly.
Who is liable when legal AI makes a mistake?
DreamLegal asked: Errors, hallucinations and inaccuracies persist across AI software. From the legal tech side, what should the vendor's liability be by market standards if a major mistake is made, whether or not the tool was involved? How should users look at it, and does Sirion handle it in a particular way?
Kanti Prabha: It's a very important question. Sirion has always talked about the concept of responsible AI, and about being responsible for everything we put out there. We are in a serious domain: contract documents, and the answers that come out of them. If we draft a document without human controls, we need to be responsible for those outcomes.
"I truly believe the vendor is responsible for all AI that is presented through their platform."
There are responsible ways to use AI. Applying raw AI and letting it do its thing, without the right safeguards and transparent, deterministic outcomes, is something one should avoid. Over the last two to three years, several good architecture practices have come into play, and I think Sirion is at the forefront of applying all of them.
Our customers are enterprise-grade customers. We serve some of the largest companies in the world, and we are very aware that any AI we put out there needs to be more than AI for the sake of AI. It has to be focused on the right outcomes, which are predictable, repeatable and traceable. We are not relying only on raw LLM power, but on knowledge gathered over several years and from hundreds of subject matter experts. All of that helps with more accurate data, more determinism and transparency to the user.
The product should show what it has produced, give the user all the information on how it got there and what decisions were taken to produce those outcomes, and then let the user, the human, be the final decision-maker. We learn from how humans look at data and make decisions, and we make systems responsibly autonomous over time, as opposed to a system where there is no way to figure out why it behaved a particular way. I think we should absolutely avoid that.
There are ways to stay away from hallucinations and errors, and to create production-grade agentic processes at scale that are grounded in knowledge and use the best of what AI has to offer.
Related: Crosby to insure AI agents for professional liability.
New AI contract tools vs established CLM: is implementation still a barrier?
DreamLegal asked: The market is full of tools. Some are good and some are cheap wrappers. Newer contract tools are faster to get started with, while CLM carries the tag of being implementation-heavy. Is there a real difference in value between these newer contract tools and legacy CLMs? And is implementation still a barrier that makes legal teams hesitant to start on CLM?
Kanti Prabha: One should always learn from the players out there, no matter how young they are. If one can achieve faster, smoother, predictive implementations that are autonomous in nature, one should absolutely aim for that.
Some of the systems built before AI had the luxury of spending many years observing human behaviour. We have so many human decisions collected over the years that we have watched, learned and understood from. We sit on a very wide range of knowledge, both industry-agnostic and industry-specific. I believe that if systems like ours, which were built before AI-native systems came in, learn and use that knowledge wisely, we can have really quick implementations, grounded in an understanding of the particular customer and their industry.
Organisations should move to faster implementations, no matter when they were born, and have systems that can prescribe the right implementation and reduce mundane human effort. Move human effort into better processes and more decision-making, and do more with what we already have.
So I feel legacy systems have an opportunity. We should absolutely evolve, and quickly. You cannot stand still at this point in time. But we have advantages that newer systems do not have, and we should use them to the customer's advantage.
See how buyers compare the established platforms: Icertis vs Sirion, Ironclad vs Sirion, Conga CLM vs Sirion and all three side by side in Icertis vs Ironclad vs Sirion, or the full top 10 CLM software ranking.
What do customers value most about Sirion?
DreamLegal asked: As a founder of Sirion, what is the one thing customers love most about it? It could be a feature, a particular trait, the support system or the implementation.
Kanti Prabha: There are two things we do very well. One, we know we are here because of our customers and for our customers. Their needs from a contract management system change over time, and we should change with those needs and be extremely responsive to them. If they need faster implementations, a better user experience, or AI that doesn't just suggest a response but takes processes into its own hands, we will do those things. The customer is at the heart of our evolution, and it has always been that way. That shows in our roadmap and in how quickly we innovate, and also in how we interact with our customers and listen to them.
The second thing is not easy to see from the outside. We never try to build trivial technology. We take every problem statement we get and try to solve it properly for the largest companies in the world, for the most complex use cases at hand. We try to cover all types of scenarios, so that when we put something out there, it's worth its salt. It's not demoware. It is real, and it will work for all of our customers, no matter who they are, how big they are or which industry they're from.
So these are the two core principles: the customer voice, and building thoroughly and responsibly. These are areas we have always been proud of, and I feel it is something our customers really love about us.
How should you measure CLM ROI beyond time saved?
DreamLegal asked: A lot of ROI frameworks for legal tech rely on time saved. Since CLM has an impact across multiple departments, which other metrics should a legal team, or the whole business, look at? We commonly hear about revenue leakage, obligation management and deal velocity. Which metrics do legal tech customers often miss that truly speak to the value?
Kanti Prabha: I'll come at it from contracts as an object. It is one of those assets that can help us do better business if it is written, managed and understood correctly. Better business comes in the form of more revenue, right spend, less leakage of value, time saved, and people being able to do high-value work instead of repetitive work. Quality goes up when a system can use broader, product-based intelligence instead of relying on one person to run a deal.
Time is one dimension, and it's a great one. Whenever you save time, you unlock time for much higher-value activities. But there are several other areas with strong and visible impact on both the top line and the bottom line, and contract management systems can address them well.
When two companies come together in a contract, if they keep their promises to each other intact, they will always do better business. The vendor does better business, the customer gets more value from the vendor, and their business in turn improves. For me, that is way more valuable as an outcome than just time saved. Time saved is great, but there are many higher-order outcomes we can aim for if it's done right and implemented correctly.
The future of contract management: her prediction for the next 2 to 3 years
DreamLegal asked: The market is changing quickly, with new sentiments coming in daily. If you had to make one prediction for the next two to three years, what would it be? We're not asking about Sirion's strategy, only the one thing you're seeing that will surely happen.
Kanti Prabha: For our category, contract management, I feel it will be completely revolutionised. It will no longer be human-dependent or human-driven. It will turn to higher-order outcomes that are system-driven, and humans will move to using contracts for business outcomes that were not possible before.
I feel we will see different systems contributing their data to common interfaces, with the user truly at the heart of all business processes and each system there to serve the user. The boundaries between systems will start to blur. Today a user goes to one system for one process and another system for another process. Those boundaries will blur, and we will truly meet the user where they live.
And data will transform into intelligence like never before. We are already seeing a lot of it, and I think Sirion is at the forefront of reimagining contract data as contract intelligence in ways that were not possible before.
"We should all be prepared for technology to do things and take over processes completely, and we should be ready to move to higher-order thinking ourselves as well."
For a wider view of the shift, read the rise of AI-powered contract lifecycle management platforms and Leah's agentic contracting platform.
Building a CLM company in 2026
DreamLegal asked: When we were researching you, one bio line stayed with us: "proud mom of two girls." How does it feel to be a co-founder of one of the most successful CLM companies, at a time when legal tech has finally found its value? Is it overwhelming, or an exciting journey?
Kanti Prabha: 2026 has been one of the most rewarding, exciting and exhilarating years for building products. What we have in our hands are tools, skills and ways of thinking that allow us to go 10x, 100x faster than we have ever been able to build. Those who have been able to break out of past ways of thinking and remove constraints from their thought processes are able to run much faster and build outcomes for their end users in special ways.
I feel very blessed to have this opportunity to re-look at an entire category completely from the ground up. My daughters see every day how excited I am and how enthusiastic I am about working on technology I have never worked on before. I've introduced them to the power of LLMs and how they think, and my girls experiment with it too, within the guardrails of what kids should be doing.
All in all, there is absolutely no sign of slowing down. It is a very different level of energy that Sirion possesses today, and so do I.
About Kanti Prabha
Kanti Prabha is co-founder and President of Sirion, where she leads product and technology. She co-founded SirionLabs in 2012 and is based in Toronto. Sirion builds contract lifecycle management software for large enterprises. Kanti Prabha on LinkedIn · sirion.ai · Sirion CLM on DreamLegal
This interview was conducted and edited by the DreamLegal team. Answers are in the guest's words, edited for length and reading.
Compare tools
Add the tools from this article, then search for more.
Sirion CLM
Was this update helpful?
Products mentioned
Stay informed
Make your next legal technology decision with more clarity.
Join DreamLegal for independent research, market perspectives, and practical evaluation guidance for legal teams.
Create Free Account