|Paul Branch, Head of Deal Assurance British Telecommunications (BT) Americas|
|Peter Wallqvist, VP of Strategy at iManage RAVN|
What will it take to get Artificial Intelligence (AI) up and running on a contract lifecycle management (CLM) solution? IACCM research suggests that nearly 80% of CLM deployments currently fail to deliver all the expectations. Does this mean AI for CLM must suffer a similar fate?
Certainly not if you know how to combine human partnerships with AI into your CLM processes.1 Research by British Telecommunications (BT) shows that a CLM system’s deployment is as much about people and processes as it is about technology. AI cannot work alone; therefore, if we expect best results, we must actively approach AI as a human-machine partnership that requires an agile approach to deploying CLM with AI capability.
Why use a partnership approach?
AI is very good at working with data, nevertheless, most data currently stored by most corporations is unstructured.2 Picture, for example, a storage repository filled with thousands of supplier and customer-facing agreements. While most corporations have invested in curating this unstructured data somewhat by entering metadata tags around authorship and content and the like -- most stored data remains a pile of text without any structure.
This trove of unstructured data is a perfect fit for AI. It is terrific at scanning and processing enormous amounts of data to assist with tasks ranging from obligation extraction3 to developing Contract Effectiveness Scorecards.4 Of course, you could do all these tasks manually but the cost is prohibitive – not to mention the huge amount of time you’ll need.
AI does require subject-matter expertise and needs to partner with a human to deliver the best results. For example, if you need to go through thousands of contracts and flag the ones that meet certain criteria, a well-trained AI machine learning application might deliver results with around 85 - 90% precision and recall – but you will still reap 10 to 15% false positives and unrecalled results that a human will need to expertly review.
That need for the human involvement is the main reason you need an AI plus human partnership approach. This does not mean AI eliminates jobs. For most AI operations, you don't replace people with the AI tool; you give the AI tool to the people to replace a lot of the tedious, repetitive work that humans are not particularly good at anyway.
AI’s partnership approach is therefore human led, machine assisted and should be an organization’s mantra. Human involvement must be constant throughout the entire process. This starts right at the beginning with understanding the business, what is required, and what the technology can do. This then helps determine how the AI tool needs to be trained, after which it can then be put into service to address the needs of the organization.
How AI works
Documents are fed to the AI tool to help train the tool, and the diversity of language in documents can greatly impact this training. For example, if you have very similar documents, then you're only going to need a few training samples to teach the AI what to look for. But if you have complex agreements where the verbiage is completely different – the same semantic information is being delivered, but with very different syntax – then you're going to need considerably more training to get a reasonable level of precision.5 So, again human intervention is necessary to provide quality assurance to the output, because it is based on an understanding and measurement of relevance of the AI tool and the desired outcome.
Stay agile and…
If your organization changes from a manual to an AI system you can use several approaches. It’s difficult to pull off a big bang approach6 if you launch the new system all at once after a lengthy preparation phase. If you’ve been trying to use AI to solve every problem within your organization at once, you will likely not be successful. Also, if you have spent too long capturing requirements, they will be out of date by the time you start implementing them.
Staying agile is highly preferable. It means deploying incrementally – making incremental adjustments and improvements, but also making sure the AI platform enabling the deployments allows for an agile approach and supports future projects. Otherwise you could end up with a dead-end technology that supports only a first-use case.
… make a culture change
Organizations also need to approach any systems deployment – especially one around AI – as a culture change, otherwise they will not see the business benefits.
This means realizing that you’re changing the business processes that people have used for a very long time, which in turn, necessitates changing people’s attitudes towards the role of different technologies. These softer issues – people’s attitudes, the organizational culture, and so on – can make or break a system’s deployment and are ultimately much more important than the technical specifications of the system itself.
Regarding technical specifications, it matters whether you select Cloud versus on-premises AI solutions. Your selection depends entirely on your organization. In the context of geographical boundaries such as the European Economic Area,7 data protection and sharing are key considerations. So, most important is your ability to very easily make the solution as operational as necessary, geographically speaking, to meet any pertinent regulations or requirements.
BT and AI solutions you can use
BT has put AI to practical use in several different scenarios, including starting to use it to extract obligations – essentially, we can now identify who is accountable for doing what in an agreement. We can also use AI to:
Manage obligations - Understanding the obligations that BT may owe to customers -- or that suppliers may owe to BT -- ensures that those obligations are effectively owned and managed within the corporation. AI allows us to respond swiftly when those obligations are jeopardized,8 so that the risks caused by that activity do not materialize into issues that may negatively impact the balance sheet. Ensuring that effective governance is in place so that minor risks don't become monumental issues is the hallmark of a well-run contract management function.
Manage and track Master Service Agreements - At BT we have successfully used AI to help identify the obligations of various parties in many thousands of its Master Service Agreements. Because Master Service Agreements are the most complex agreements we work with, we need to very carefully manage their highly negotiated, bespoke conditions (specifically written for the stated purpose or customer). Understanding what BT has committed to do – and what operations other businesses have committed to BT– enables us to make sure these commitments happen and to suitably respond should those commitments become jeopardized.
This is critical when it comes to revenue leakage. According to the IACCM, contract revenue leakage accounts for 9.2% of a company’s global revenues. This directly represents the value leakage resulting from poor contract management in organizations with a lack of investment in people, processes and systems. More specifically, the numbers generally mean that 3 to 4% of revenue leakage results from parties not meeting their contractual commitments. These percentages impact bottom line profit. A recent contractlogix article notes: “From missed renewal dates to clauses exceeding approved ranges, there are many reasons why organizations are unable to maximize their contract revenue.”9
Reduce negotiation time - Our company has also used AI technology to apply risk management to contract management with a particular aim towards reducing the amount of time required to negotiate some of its most complex agreements.
Assess risk fast using data comparisons - Improving deal velocity requires repeatable solutions and doing things in a more standardized way. We can use AI to very rapidly assess the risk profile of a contract by looking at the provisions it contains and comparing them to its database of previous contracts. From there, the we can automatically approve the vast majority of the text – including insertions that the customer has recommended, for example, or red lines that the supplier has come back with – and then focus on the two or three key items that are pushing the deal out of a particular risk profile into one that is less attractive.
Parse millions of documents - We found another use for practical AI when the European Union introduced the General Data Protection Regulation (GDPR).10 The company parsed half a million customer and supplier documents through its AI tool to identify the contracts that might require some type of amendment or otherwise require more expert legal attention, in light of the new regulations.
To be clear, the AI cannot come up with recommendations on how to amend the documents or make the amendments themselves – that activity requires a human. However, it could quickly triage that mass of documents into a more manageable pile for human review, saving vast amounts of time and effort. Again, this is the essence of a human-machine partnership approach.
A Worthy Investment
While AI might not be magical – or mystical, or even miraculous – it's still a technology that is very much worth the investment for most organizations. Approaching AI in a realistic fashion – understanding its capabilities as well as its limitations; understanding that it requires humans and machines to work together – will allow organizations to put it to use in a very practical fashion and enjoy the benefits of a new and fruitful partnership.
- Collaborative Intelligence: Humans and AI are Joining Forces, Harvard Business Review, H. James Wilson, Paul R. Daugherty, July August 2018 edition.
- Datamation article: Structured vs Unstructured Data.
- definition: Obligation Extraction.
- definition: Contract Effectiveness Scorecardmeasures effectiveness of a contract for all parties.
- Wikipedia article explaining precision and recall.
- Wikipedia article explaining big bang adoption.
- definition: European Economic Area.
- definition: jeopardy managementpertains to an app that warns when due dates may slip.
- contractlogixarticle: 4 Ways to Improve your ROI from Your Contracts.
- Wikipedia article explaining GDPR.
About the Authors
Paul Branch, Head of Deal Assurance at British Telecommunications (BT) Americas has worked in the Telecommunications/IT industry for over 30 years, with more than 20 years spent in various buy-and sell-side contract and commercial management (CCM) positions. He is a Fellow, Vice-Chair and Officer of the IACCM. Paul believes that the application of Artificial Intelligence (natural language processing and machine learning in particular) is set to revolutionize the work that we do. He is very active in applications of this technology in BT.
Peter Wallqvist, Peter Wallqvist, VP of Strategy at iManage RAVN is Vice President of Strategy at iManage RAVN, where he is responsible for positioning new and enhanced practical AI solutions that empower professionals to increase efficiency, improve productivity, and mitigate risk. Peter was co-founder of the AI company RAVN Systems, which iManage acquired in 2017.
ABOUT BRITISH TELECOMMUNICATIONS
BT Global Services provides information and communications technology services. Its portfolio ranges from desktop, network equipment and software, transport and connectivity, IP-based e-business solutions, managed network services, and systems integration to consultancy for global requirements. The company offers integrated data and value-added services to multi-site corporate entities across the globe.
ABOUT iManage at RAVN
iManage RAVN Insight takes enterprise search and knowledge management to the next level through AI. Leveraging the RAVN AI technology, Insight finds, analyze and identifies organization information buried in disparate systems regardless of location. iManage RAVN Insight’s Knowledge Graph surfaces connections between people, expertise and knowledge to identify hidden experts, related projects and relevant clients. iManage RAVN Insight activates expertise location for law firms throughout the entire enterprise. Professionals work smarter and can build knowledge applications that solve high ROI problems across the enterprise.