Agiloft has unveiled its Summer 23 release. A core part of the update is AI Trainer. This new AI model training capability will empower non-technical users to fully customize how they review and analyze contracts. The solution provides an interface allowing subject matter experts to train AI with the terminology and information specific to their firm and industry that is held within their contracts. Users can create the terminology used when Agiloft imports contracts to improve the AI understanding as it reviews hundreds and thousands of documents.
AI Trainer does not require data scientists, nor does it need developers to enable it. It uses a no-code platform to create labels, models, and AI Projects. Projects link models to the data, and users can then ingest the data, using the models to process contracts in volume for the Agiloft platform to better understand the content. The solution will allow organisations to accelerate review cycles, extracting relevant information about each contract in their repository.
What is in this new release?
Enterprise Times spoke to Andy Wishart, Chief Product Officer, Agiloft, about the new release. He explained further, saying, “AI trainer gives the power to our customers to create specific models that are going to find the content that is unique within their agreements, is to be used, in addition to out of the box pre-trained AI. But we believe that to get the full power of AI to find data and clauses within our customers’ contracts. It requires both pre-trained AI, as well as custom AI models.”
AI models will contain multiple labels associated with a term or clause. According to Wishart, labels can be associated with discrete data points of whole clauses contained within contracts.
Customers will start with a very small set of standard pre-trained platform labels, including contract title, party name, and addresses effective date, duration, end date and the governing law. Wishart indicated that Agiloft will add to this list over time.
Enterprise asked Wishart whether Agiloft would look to provide a base set for different verticals and horizontal solutions. He indicated that there were no plans yet to do so. However, partners can create pre-trained models for industries like biotech or financial services.
How will this help customers?
Wishart also revealed how the new update will assist existing customers. Wishart answered, “One of the use cases is legacy import where a customer may be new to CLM. They have a set of contracts that might be sitting within a SharePoint site or a file store somewhere, and they have a lack of data about those contracts.
“They want to get those contracts into the CLM. But they also want to get some of the key data points and the clauses out of those contracts. So creating custom labels that specifically target the key terms and the clauses unique to their organisation can help with that onboarding of content.”
Wishart concluded, “No two contracts are quite alike, nor are any two organizations. That is why relying on pre-trained, generic AI models alone simply does not get the job done. We are introducing AI Trainer to ensure more organizations can use our best-of-breed AI to surface, analyze, and report on their contracts effectively.
“It provides legal and contracting teams with an easy-to-use, self-service tool that helps them codify their expertise to enhance the automation of the contracting process. AI Trainer empowers the very teams who are closest to the contracting process and gives them a way to train and individualize the systems they use to uncover and categorize key terms and clauses in their contracts.”
An Intelligent trainer
The AI Trainer can be prompted with new information by users, and will also identify additional data that users may wish to tag appropriately. Wishart views this as a differentiator.
He added, “As users annotate documents within the AI trainer, we can learn from those annotations on the first 15 documents and then make suggestions whenever they open up a new document. (The AI Trainer) then says to the user, we think this is an example of what you want to teach me. So we’re learning as the user adds additional documents.
“We speed up the training process by making suggestions. We recognise the pattern, and then we ask the subject matter expert, is this the pattern I’m looking for? They can then, just with one click, accept that pattern. Then, they’ve trained that document. So we’re using AI itself to speed up the training process.”
Eric Laughlin, Agiloft’s CEO. Commented, “CLM should be viewed as an enterprise solution, rather than just legal technology, so a legal department’s partners – in sales, procurement, and finance – need to be able to easily find and interact with the invaluable data that is contained within their contracts.
“Using AI to identify party names, dates, and common clauses is necessary but insufficient. The organizations that create competitive advantage will be those who are able to extract and use the information that is uniquely valuable to their environment and unlikely to be covered by pre-trained models – for example, legal and commercial terms unique to their industry niche or company. That’s why we are so excited for our customers to use AI trainer – it puts that power in their hands.”
AI Trainer is available globally. However, it is an add-on module to the standard platform, and pricing starts at US$30,000 per year.
What else is in the Summer release
Wishart summarises some of the other key improvements saying, “We have a Gmail app that will be available within the Google marketplace. Using the Gmail app, customers can connect to their Agiloft instance from within Gmail to be able to draft new emails and link in contracts in a similar way that we have for our Microsoft Outlook app.
“We have a few usability enhancements. One is called the modal action, which just makes creating the forms that users complete a little bit more intuitive. And a few smaller enhancements. But those are largely the main ones.”
In addition, Agiloft has combined the configuration of Outlook and Gmail into a single configuration wizard. The integration with Power BI is improved, with users now able to bring Agiloft table data into Power BI to easily place it in dashboards and reports for further analysis.
There are also several improvements to the MS Word integration. The full release notes, including other updates in Agiloft 25, are available here. Agiloft will explain more about the new AI Trainer, part of its Summer release, in two Webinars on August 15th at 1 pm EDT and August 16th at 2 PM BST.
Supporting the Summer 23 new release and AI Trainer
In addition, Agiloft will be releasing a module within the Agiloft University for AI Trainer as part of the release. Wishart also explained how Agiloft will further support the launch of AI Training with its customer success team.
“Within our professional services organisation, we have the AI success team. The AI success team provide best practice advice regarding the approach in training models and labels. To train our customers on the terminology between labels and models, train them on the datasets they will need for those 20 to 50 documents, and ensure there’s variability within those but not too much variability.
“So some of the best practices that come from them as practitioners. Our professional services team and our legal knowledge engineering team build our models. They are attorneys that have worked on commercial negotiation of contracts throughout their careers, they’re specialists. We provide that as an additional professional service, as well.”
Another step towards fulfilling the vision
Agiloft continues to drive development across four pillars within its platform. These are “Autonomous & Self-Governing, Standardized and Interoperable, Intelligent & Optimised, Digital and Connection”. While there are elements of each within this update, AI Trainer is part of the Intelligent and Optimized pillar.
Overall, Agiloft has a vision to build connected, intelligent, and autonomous contracting processes that enable companies to unlock the value of contract data and accelerate their business. Wishart postulated where the vision might take Agiloft, saying, “The ultimate end point of that vision is a fully autonomous and self-governing contract and process that could see two GPTs in the future, negotiating a contract on behalf of the organisations that they represent.
“That GPT will be fine-tuned on the commercial levers, the risk profiles of the companies they represent, to try and reach this theoretical optimal contractual agreement that could then be passed for human review and execution.”
AI Trainer is a step on the journey towards this vision. Another step Wishart explained could be using “Generative AI and classic machine learning to quickly identify the contractual, non-contractual and the systemic risks. And to do things like optimising contract and managing obligations, and contract remediation as well.”
Beyond that, “Generative AI could be used in the creation of amendments to then remediate that change in legislation or regulation.”
Enterprise Times: What does this mean
The absence of industry verticalisation may seem a barrier to adoption, a small one. This is an enhancement worth noting and looking for the early case studies and benefits the solution can bring. The applications across both horizontal functions and vertical industries could provide some extremely attractive investments for partners for their go-to-market engines. Customers with specific contracts can now generate even greater value from their legacy contracts without the assistance of legal or developers.
