Today, Sayers has built dozens of custom AI Agents that handle everything from confidentiality assessments to contract summaries, transforming how teams across symplr interact with their contracts. The results speak for themselves: a 200% increase in platform adoption and dramatic time savings that have freed up both the legal team and business stakeholders to focus on higher-value work.
We sat down with Sayers to learn about his approach to building AI Agents, the specific use cases he’s developed, and how Contract AI has transformed symplr’s operations.
What drew you to explore AI Agents for contract management?
Sayers: I wanted to address routine challenges that we have with respect to data that’s in our contracts but is not in our business operating systems. As most companies implement contracts, there’s usually just a subset of data that you’re going to put into your systems—like they have this product, and they pay this much, and they get billed on this date. Having those basics is obviously essential for business.
But a lot of times when you support your customers, there’s information that’s in the contract—which is the full embodiment of your commercial commitments—that’s important to understand in order to support the account, or to know how to structure the next deal. I wanted to make sure that our teams had access to the most meaningful contract data relative to our business that we know is in our contracts but not in our systems. We nonetheless have to understand the implications of that data when we’re working with our customers, and that’s what made me begin to create these AI Agents to help everyone easily access the insights within every contract.
Can you walk us through a specific AI Agent you’ve built?
Sayers: One that comes up fairly frequently is when customers ask for a SOC 2 report. This is becoming more common because cybersecurity risk to the healthcare industry has grown, and customers are really scrutinizing their vendors more than ever.
The question becomes “Do we need to put a new NDA in place or do we have something under the existing active contracts that would allow us to share this information?” We also find this to be pervasive because we’ve grown by acquisition. That means a customer could have multiple agreements in place and they may be asking for our SOC 2 report for a particular entity.
So I created a confidentiality AI Agent. What it does is review the confidentiality provisions in any existing agreements with that customer and then advise specifically whether that would cover sharing our security reports. For example, it might say “The agreement would likely cover the confidentiality of security reports because it has a broad provision for proprietary information, and there’s no exclusions that would exempt security reports.”
What was happening before is, an Account Executive would request a confidentiality agreement and it would go into the queue in Legal, and it would take a day, or maybe two, depending on workload. That’s just dead time when we could have just said, “We don’t even need one because we’re covered under the existing agreement.”
Now, with the AI Agent I created in Pramata, the Account Executive can see in seconds whether a customer’s current contract allows them to share a SOC 2 report, and if not, generate a new NDA within minutes.
What other types of Agents have you found valuable?
Sayers: I have Agents for price escalation terms because customers will want to understand what their price increase provisions are, especially when you’ve grown by acquisition and acquired different companies that had different things in place. Instead of having to go through multiple contracts manually, we can quickly see that in one case it’s three percent, and in another there’s nothing explicitly mentioned.
I also have a “recommend a universal price uplift” Agent. The customer success team told me this was something that takes them a long time, so it made sense to build an Agent to automate it. Customers want to have a common uplift provision that applies to all of their solutions, and it’s time consuming to go through all the contracts to see what they currently are and then come up with a recommendation.
The Agent identifies what the price escalators are that exist across all agreements, then gives a recommendation, like recommending a three percent price escalator because it’s reasonable, creates consistency, provides meaningful growth, and is more predictable than tying price increases to the consumer price index (CPI). It even gives rationale so the Customer Success Manager has talking points to go back to the customer about why it would make sense for them to agree to it.
How do you handle more complex contract analysis?
Sayers: One common challenge is when customers want to understand what products and services we’ve actually agreed to provide to them. When we’re trying to service an account and make sure that we understand the different things that we’ve put in place, if I wanted to know what products we sold under all of their order forms, I’d have to click into every single document manually.
Now I have a “summarize orders” Agent that will go through and look at all the orders that have been selected and tell me what the customer purchased in them. I also have a “product license model” Agent because when you look at these agreements, they don’t necessarily indicate whether they are for on-premise or for SaaS products. The Agent will identify all the products they’ve purchased, explain the license model, tell you quantities, their hosting model, the current term end date—and it gives that in a table that we can export if needed.
What about for customers with complex licensing structures?
Sayers: That’s where some of our most powerful Agents come in. We have customers who have grown by acquisition themselves, and they’ll reach out asking “Can you tell me which of my locations are licensed for this particular solution?”
We might have 14 child documents for that customer, and going through all of them to figure out their licensed facilities would be hours and hours of work. Nobody wants to take the time to read 14 agreements and transcribe a list of licensed facilities.
I created a “licensed facilities” Agent that goes through and gives you an overview, shows the initial list of facilities and the ones that were added over time, then provides a final output with all current facilities and their addresses. The Agent checks for locations that might have been removed and gives you the complete current list. To put this into a spreadsheet manually would be hours of time. Now it takes seconds.
How has adoption been at symplr?
Sayers: The feedback has been overwhelmingly positive. Teams are over the moon because it’s been so painful and time consuming to get this information prior to having these Agents. When I showed the data termination Agent at a sales training, people said “My gosh, this is amazing” because it answers questions like, “Do we have to give customers their data back when they terminate?”; “Do we charge them for their data?”; and “Does their license end immediately?”
My goal was to increase Pramata usage across symplr by 20 percent year over year. When I took a snapshot of usage from a 90-day period early last year and compared it to this year, I had beat my goal by 10 times over. Both the number of active users and active sessions went up by around 200 percent, and the Gen AI Agents were definitely a large driver of that adoption.
What advice would you give to others implementing AI Agents?
Sayers: Talk to your stakeholders and your folks who work with contracts, not just in Legal but in Product, in Finance, in Sales, in Customer Success. Try to understand things like:
- When they need to know something in a contract?
- What do they need to know?
- What’s taking a lot of time to research?
- And what are conversations about contracts with customers that are difficult either from a time perspective or in explaining or articulating it?
Rather than letting your imagination run wild about what’s possible, focus on understanding what’s hard to know and what’s time consuming to research. When I talked to sales leaders, they said they always need to know what the uplift terms are because when they’re doing the next deal, they don’t want to upset the customer with terms that are way different from what they’ve traditionally agreed to. When I talked with Customer Success, they wanted quick summaries to know when customers renew and when they have to provide notice, because that helps them plan their engagement cadence.
What do you want others to know, especially if they’re hesitant to get started building AI Agents?
Sayers: There’s nothing to be intimidated by! Pramata was a great partner in supporting me and equipping me with the understanding that I needed to successfully implement these AI Agents and to understand what the limitations are in addition to the art of the possible. Setting that expectation with me was important because it allowed me to then set that expectation internally and help users understand how to get the most from the system.
To read more about Foster Sayers and his team’s success with Pramata at symplr, check out the full success story here.
Ready to see Pramata for yourself? Get a demo now!
