DocQMiner: contract analysis through AI

Case study

DocQMiner: contract analysis through AI

Evolution of recognizing and processing unstructured data

In order to comply with international regulations, companies with lease contracts will be required over the next few months to go through thousands of contracts one by one. It’s an immense task, certainly when you consider that an analyst will spend around 90 minutes on each contract. That time can be drastically reduced using machine learning technologies. To assist companies with tasks like this one, Deloitte has developed DocQMiner: a self-learning application that reads through and analyses contracts.

Smart Suggestions

A number of our clients were faced with a challenge, says Marc Verdonk, Partner and Innovation Manager at Deloitte Risk Advisory. “In accordance with the new IFRS 16 accounting standard, virtually all lease contracts must be listed on the balance sheet from 2019.” For a telecoms company, for example, which leases every mast and every plot of land on which that mast stands, this means it will be required to go through hundreds of thousands of contracts in all manner of different languages. “Our team began working on the question of how machine learning could extract the relevant data points from those contracts. Not to replace contract analysts, but in fact to assist them with smart suggestions.”

The result was DocQMiner, a user-friendly application that can be used by analysts to review contracts. “The application features a bot, which we have named Robin. You feed in a number of contracts, and Robin gives you suggestions for the data that you will need, such as the start date of a contract,” explains Verdonk. The analyst sees the highlighted suggestion and indicates whether or not it is correct. “Robin learns from this, which means that subsequent contracts are analysed a little more smartly each time, and the reliability of his predictions increases. The analyst is constantly training Robin during the review process.”

DocQMiner works using state-of-the-art machine learning and natural language processing technologies. Verdonk: “It features a neural network that understands language, sentence structure and how words relate to one another,” continues Verdonk. That knowledge is converted into figures, which are used to calculate and make predictions. We also add information to it. By feeding in annotated contracts, showing what data we are looking for, Robin learns how to recognise lease contracts.” Because the system’s recognition becomes more and more accurate, but does not store any privacy-sensitive client data, it can be taken from one client to the next. “It means our clients benefit from the experience it has already gained.”

AI email alert

Receive the latest AI cases

Sign-up

Broader application

The use of new technologies also gives rise to a new form of doing business. Verdonk: “With DocQMiner, we are not selling the amount of work that goes into it, but rather the result,” says Verdonk. “Instead of stating the number of hours that we expect to spend on it, we make an agreement regarding the number of contracts that we will analyse. That results in a low price for the client, and an incentive for us to do it as efficiently as possible.”

Although DocQMiner was developed for clients who are required to comply with the IFRS 16 or US GAAP accounting standards, there are numerous other conceivable applications in which large quantities of contracts need to be read through. Verdonk: “One example is Brexit repapering, when all contracts will need to be revised if companies’ headquarters relocate. Or the General Data Protection Regulation, which will require clauses in many purchase contracts to be checked. We can use DocQMiner in all kinds of situations.”

 

DocQMiner- contract analysis through AI

*) This case is part of the series of 16 Artificial Intelligence projects from Deloitte. Other cases in the series are in random order:

  1. TAX-I: A virtual legal research assistant
  2. AI Benchmark 
  3. SONAR: Find labelling errors in databases
  4. Transaction detector with regard to the Dutch work cost regulations
  5. GRAPA: assistance with risk strategies
  6. Chatbot as a handy search tool for the online technical library
  7. Argus: an eye for detail
  8. PostNL: optimising delivery times
  9. Virtual assistants: beyond the hype
  10. HR agent Edgy: the future of Human Resources
  11. Using machine learning to assess risks for insurance policies
  12. Predicting payment behaviour
  13. DocQMiner: contract analysis performed in no time at all
  14. Combating welfare fraud with machine learning
  15. Using machine learning and network analytics to search for a needle in a haystack
  16. Clustering unstructured information in BrainSpace

Sign up for the email alert to get all cases through email.

Vond u dit nuttig?