What can Artificial Intelligence (AI) do for contract management and what will it require? Our author proves that AI is already pushing contract management processes toward faster and smarter results. Chances are you will want to be in on it.
While Artificial Intelligence (AI) techniques can streamline many business processes, Contract Management is particularly well suited as an application of AI. AI capabilities will play a rapidly increasing role in improving contract management and upleveling the strategic role of contract managers.
Both predictive AI and Natural Language Processing have extensive application in contract management. Contract managers can prepare to realize the benefits of AI by ensuring that their entire corpus of contract content and process data is captured in high resolution within a robust contract lifecycle management system.
What is AI really?
Leaving aside visions of Skynet, an artificially intelligent system is designed to assist and facilitate the human decision maker. An artificially intelligent system gets better at its job based on the information fed into it. Instead of being wholly dependent on hard-coded programming to improve, such a system learns and improves on its own.
An ordinary widget-making machine will churn out exactly the widget it has been programmed to make, time after time, without improvement. An artificially intelligent widget-making machine, in contrast, would use the information gained from repeatedly producing widgets to make or suggest improvements in the product or the manufacturing process.
In one respect at least, AI systems learn much the way humans do; they use data from the past to predict the future. Predictive Intelligence (aka machine learning) based upon regression or neural networks predicts what will happen next, based on information about similar scenarios in the past.
Another form of artificial intelligence is Natural Language Processing. NLP interprets the meaning of text in, again, a more human-like fashion. As intelligent beings, we make meaning out of the words we hear and read, not just from the words themselves, but from the context in which they appear. This is an essential human skill because the exact same words -- depending on tone, situation, who is speaking, and the words that precede and follow -- may be anything from an endearment to a joke to a threat. AI NLP, while not yet sophisticated enough to detect the subtleties of sarcasm, also interprets the meaning and intent of text based upon context.
Together these forms of artificial intelligence open the door to contract management systems that support contract managers in an entirely new way; automating functions that previously required human intervention and allowing contract managers to uplevel their roles to more strategic functions. At ContractRoom1, we believe that the use of AI in contract lifecycle management and contract negotiation will significantly improve the efficiency of the contracting process and reduce the legal risk for all parties.
Let’s touch on the importance of data as the “feed stock” of AI, and then look more deeply at how each of the different forms of artificial intelligence can be applied specifically to contract management.
The Value of Data
Artificial intelligence, like our own human intelligence, is fed by data. Data about known scenarios and their outcomes are used to train AI models to process future data. The more examples of historical data are available to train the models, the better the models become at predicting future outcomes. A human or an AI system that has been exposed to only one sheep that happens to be black might conclude that all sheep are black. A system that has been exposed to several sheep would not make the same mistake, but still might not have enough information to understand what other characteristics are correlated with black wool. However, an AI that has been exposed to a sample of millions of sheep could identify what characteristics are associated with black wool and make predictions about whether any particular future sheep might be black based upon those characteristics.
The more data points available within each historic example, the more likely the model can identify the key combinations of information that are useful for understanding and prediction. In our farmyard example, color is only one datapoint among many about each individual animal. Adding other datapoints enriches the model by allowing the AI to discover meaningful associations within the data.
This means the value of AI in contract management will rise with the number of datapoints captured in your contract management system, as well as the number of contracts that system has processed. Capturing the contract text itself is not sufficient to enable useful application of AI.
Contract management systems should also capture additional data about contract content, as well as data about the contracting process. For example, in addition to contract text, contract management systems should capture the structured discrete terms within the contract language, such as the effective date or payment schedule. Additionally, values associated with the contract but not explicitly stated within it should be captured, such as the associated business unit or lifetime contract value.
Data from the process side may be even more important in enabling AI for contract management. Contract management systems are the natural place to capture data about the negotiation workflow -- information that is typically lost outside of the contract management system. This data includes who was involved in the negotiation process, what changes they made, who reviewed and approved each change, how long each review took, and who disagreed and slowed down the process.
A rich data set can be used to power AI models that provide insights back to the contract managers, adding greater speed, control and insight to the contracting process.
Let’s look at how the different types if AI, if correctly fed by rich data, can contribute to these gains.
Predictive AI for contract management
Predictive intelligence, using patterns discovered from past data to predict the future, can contribute to at least two different types of insights to help make better decisions and close contracts sooner.
Contract content insights
Content insights, as the name implies, are related to the content of the contract itself. They consist of predictions about present content based on past examples.
For example, when changes are made to a particular section or field value within a current contract, AI could use past data to help predict whether the current changes are likely to be approved by all parties. For example, it might be predicted that any change to a force majeure clause, regardless of its details, is likely to be approved. On the other hand, changes to an indemnity clause are likely to require more negotiation.
In some cases, predictions can be made about what the final value for a field is likely to be. For example, if "commission percentage" is a field within the contract and the starting value is always 10%, other contract content can help predict whether a proposed change to 15% is likely to be approved or not. (In that case, one party is almost certain to approve it!)
Contracting process insights
Another set of insights will be related to the contracting process itself, particularly the length of time it will take all parties to get to "yes." Based upon the contract template used, the companies and people involved in the negotiation, and the behavior of those people and companies during past negotiations, an AI could predict how long the current process will take.
The most valuable use of such predictions is not a simple estimate of timeline, but rather being able to predict the timing impact of contract changes before you propose them.
For example, suppose one of your negotiators proposes a change to a particular section. With enough historical data, the AI could predict how much time that additional change will add to the negotiation process, and even whether the change could impact the final contract value. Would they still want to propose the change if it will likely take two weeks longer to close? Or what might happen if it will take two weeks longer to close, but likely increase the total contract value by $100,000? With machine-learning based AI techniques, contract managers can help negotiators make much better decisions about whether, “the juice is worth the squeeze.”
Natural language processing
The predictive insights described above can all be achieved without sophisticated understanding of the meaning of contract language. However, incorporating natural language processing (NLP) into AI-powered contract management can provide even greater value. In particular, NLP will allow contract groups to identify when contracts have deviated from their defined standards. The approval process -- and therefore the negotiation process -- will move much faster when negotiators can quickly be informed how close, in substance, the current version of the contract is to standards.
NLP is useful here because not every change in words produces an equal change in substance. For example, if a counterparty is proposing a language change that does not materially change the meaning of the section, natural language processing can identify that and inform the negotiator. While every company typically prefers their own language, a negotiator who knows the change is not material will feel freer to approve the change to close the contract quickly. Determining the materiality would be made by combining NLP with learning from a corpus of past contracts with variations in language that have been accepted or rejected.
This same judgement about materiality can also save work internally. If a negotiator from your own company is proposing a language change that doesn't materially change meaning, the AI could recognize that the change is not material and recommend reverting it, rather than adding the work and delay of including it and risking pushback from the counterparty.
Conversely, the AI could also recognize when proposed new language does materially change meaning. Since material changes are more likely to present a risk to your company, different approval rules could be triggered. In addition to reducing risk, this use of NLP also speeds up the contracting process. Material changes can require a high level approval, while less material ones can be approved further down the organization or automatically.
This use case for NLP will become even more powerful when a company has a history of contracts executed with the same counterparty. Proposed changes to new contracts will often match changes your company previously reviewed and approved or rejected on past contracts from this specific counterparty. This context is relevant because a negotiator may allow changes from one counterparty that would be rejected from another. NLP, enriched with this additional context, will be able to identify those situations and inform the worker based on past decisions – reducing negotiation time and costs.
NLP can also add new intelligence to reporting and risk management, based on language meaning rather than just string matching. For example, legal managers frequently want to identify risk in existing contracts, so that better language can be put in place upon renewal. Today this is done by identifying places where particular clauses have been used or deviated from. It doesn't require much sophistication to identify and report upon the contracts that have had a language deviation in a particular clause or section. However, this form of reporting will give many false alarms, because it detects any deviation from the specific string of words being searched for. NLP, in contrast, could identify only contracts where a materially different version of a particular clause or section has been used.
In the future, more sophisticated NLP techniques will identify particularly risky proposals to contract language. This requires a much deeper understanding of meaning than the above examples. However, we do expect NLP technology to jump from simply identifying whether new language is materially different to detecting when new language increases contract risk.
Highly standardized contracts, such as non-disclosure agreements (NDAs), will be one the first places where risk can be identified by NLP. We expect identification of unusual clauses within an NDA to be one of the first examples of sophisticated content analysis and risk identification using NLP in contract lifecycle management.
Contracts by third parties offer additional use cases for NLP in contracting. Companies that enter into a negotiation where the counterparty has provided the initial draft of the contract will still want that contract captured and categorized in as detailed a manner as contracts they themselves originate. Currently, that process of identifying sections, terms, and fields with a third-party contract is a laborious manual process. NLP can make this process much more efficient. Given a library of pre-defined clauses to compare against, an NLP algorithm could identify which template clause matches most closely to each paragraph in the third-party document.
Getting ready for more AI in contract management
As the above examples show, AI has a great deal to contribute to improving contract management. It’s no surprise then that contract management solution vendors are actively engaged in expanding existing AI capabilities and adding new ones. So how can contract managers prepare to realize these benefits?
Digitize for Real
The crucial step in preparing for AI is collecting the data to make AI techniques possible. Data about your contract process as well as contract content is the raw material that an AI requires for learning. A surprising number of contract managers are still handicapped by “half-digitized” contracting processes that do not place the entire contracting process within a contract-specific system. Redlining documents via Microsoft Word and email, for example, wastes the opportunity to collect valuable data about your negotiation process. Likewise, simply storing completed contracts digitally does not place them in a form that an AI can use to effectively analyze them to make future predictions. Implementing a robust contract lifecycle management system that covers your full contracting process and content is the most effective way to adopt currently available AI capabilities and prepare for new ones.
Maximize data resolution
When configuring a contract management solution, the richness of the data collected will be determined by the number of agreements executed in the system multiplied by the number of data points collected per agreement. An analogy would be the number of photos you take multiplied by the pixel resolution of each photo. The higher “resolution” of data points collected per contract, the more useful your total corpus of data will be to present and future AI capabilities. This includes data points on both contract content, metadata about that content and data about the contract process such as approval speeds. Providing the richest possible data creates a corpus that machine learning to can use to discover correlations and make predictions that would not otherwise be possible.
Modern cloud solutions also offer contract managers the opportunity to benefit from learnings drawn from a broader pool of contracts. Just as cloud CRM solutions use the cross-customer pool of anonymized sales data to improve sales processes, modern cloud contract management solutions can use anonymized data from a larger pool of contracts to improve capabilities and train AI.
Potentially, AI can turbocharge Contract Lifecycle Management solutions to to continually uplevel the strategic role of contract managers. Although AI capability is in its early days for current solutions, many exciting AI innovations are in the pipeline. Contract managers need to educate themselves and ensure that their contract processes are fully digitized and AI-ready.
ContractRoom enables the future of agreement by offering “the world’s smartest Contract Lifecycle Management platform.” ContractRoom is based in Silicon Valley with additional offices in Caracas, Venezuela and Guadalajara, Mexico. More information is available at contractroom.com.
ABOUT THE AUTHOR
Emil directs product and business vision at ContractRoom (www.contractroom.com) , a SaaS Contract Management solution provider for Fortune 1000 and enterprise clients. With 20 years as an entrepreneur in tech, media and design, he co-founded Megazines Publications and America Riches and held executive positions at NBA and Batanga Networks. At age 14, he launched his first profitable startup. He is Vice-Chairman of Miami Music Project and a board member of EndPoverty.org. Stefanutti earned his degree in Design from the Istituto Europeo di Design in Milan, Italy.