Kira: AI-assisted Document Review Platform

Kira can make due diligence review more efficient gives a firm an opportunity to redesign the process for significant profits

Website:

Overview

  • A short video on how Kira works
  • Kira is an artificial intelligence (AI) assisted document review platform, which is designed to help legal teams expedite the due diligence (DD) process where a large volume of documents is under review
    • In most transactions, there will be a number of key or material concerns within the underlying documentation that parties expect their counsel to analyze during the DD process, whether on the buy or sell side
    • Transactional teams can expedite the DD process by focusing only those material clauses, and using an AI-assisted document review platform allows that team to readily identify those material clauses automatically
  • Kira has a library of "smart fields" (legal concepts / clauses) covering all common transactional practices, and which can used to quickly identify material clauses within documents under review
  • Lawyers can review documents within the platform using Kira's document review pane and note taking functionality
  • It boasts industry leading optical character recognition (OCR) software, which automatically renders documents searchable from within the view pane
    • This is really helpful as, when reviewing a document and Kira has not picked up a clause that you expect to be present in the agreement, you can simply search for key words you would expect to find in the missing clause. 
    • The OCR also immediately identifies poor quality documents, allowing the matter manager to isolate them early and seek better quality versions
  • Kira defaults to operate in English, however it recognizes multiple languages and will automatically classify those documents based on language upon processing. Similarly, the system can deploy a limited number of trained models on documents drafted in a variety of languages
  • Kira also has a number of project management features, allowing matter managers to quickly assign and keep track on the progress of the review. This includes includes batching documents to review based on practice area or complexity to project team members, and the ability to track the number of documents reviewed by reviewer

How Kira's AI Works

  • Kira deploys its AI by providing users with its library of "smart fields": these are algorithms trained by lawyers to identify legal concepts or specific clauses within documents, and this library of smart fields cover the most common practice areas in transactional work 
    • These models are in essence legal concepts and / or transactional clauses such as change of control, assignment, and data retention clauses, which a user can ask Kira to identify within a set of documents
    • Kira's library of clauses is extensive, with over 1,000 trained models ready to be deployed out of the box, some covering foreign languages and different legal systems

Kira was founded in Ontario, Canada, by former M&A lawyers and as such the focus of the training of model clauses focused on English language, common law, transactional documentation. Kira now has a small number of trained clauses in other languages which can be deployed to DDs, however this offering is small. So, if you have a requirement to review documentation which is not in English

  • A matter manager can select the appropriate smart fields and create a "Work Sheet" for the project or set of documents: a work sheet is a list of smart fields the matter manager is asking Kira to identify, which is usually based on the type of document under review
    • i.e. an Employment Work Sheet containing smart fields which identify key conditions of employment in a set of employment agreements
    • Multiple work sheets can be created covering varying document and agreements as needed per project
  • The work sheet is then applied to the document set under review, and Kira automatically identifies those clauses for the reviewers
    • Reviewers have the option to view the identified clauses within Kira or extract them to a Word or Excel spreadsheet and review them separately, free of the remainder of the document 
    • Material clauses are automatically highlighted and extracted to the right side of the lawyer's screen, and a lawyer can then report on the clause in an adjacent free text field marked "summary"
    • If Kira does not have a smart field to cover a material clause, the matter manager has the ability to train Kira to identify the clause in future DDs 

How Kira's AI Training Works:

  • Training Kira to identify new, custom built smart fields, requires that reviewers identify positive examples of the legal concept or clause within documents or agreements
    • This is a simple process in which the user, within Kira's document review pane, highlights text containing the concept or clause using their mouse cursor
  • Kira's AI needs to learn from a number of positive examples of the legal concept before it can identify it automatically. The number of positive examples will vary based on the concept you are training Kira on
  • In essence, Kira learns from not just the what that makes up a legal concept, but also from Kira has seen the what, how often, and from the circumstances 
    • Kira learns from the words used, the position of the words within the document, the words used immediately before and after the highlighted text, and the amount of information being highlighted

Training is easy! It simply requiring a lawyer to highlight the appropriate text on screen using their mouse cursor. To train a new smart field quickly you need to provide Kira with as much information as possible: so long clauses will be trained quicker. Similarly, those contained within agreements whose structure and wording does not change often will ensure that the new smart field is accurate and trained using a minimal amount of examples

  • A law firm or legal department in this way, can build up its own library of custom built smart fields which becomes its own intellectual property
    • But be careful: if Kira's knowledge lawyers have not created the smart field it's likely because it is a niche clause that does not come up often, and so the limited use may not be worth the time to train it
    • Similarly, you need to ensure that the quality of training is consistent: every person training a new model must not only understand the legal concept/clause being trained, but also understand how Kira's training works
      • Garbage in garbage out principle applies here

Is Kira's AI Accurate?:

  • In short, YES!
    • There is a caveat to this in that no tool (nor humar) is not perfect - and Kira is improving the accuracy of their smart fields all of the time
    • Deficiencies can be observed where, on very limited occasions, Kira may miss a clause due to a number of possible reasons but the main ones are:
      • Unique drafting; and,
      • Poor quality of document (i.e. a scan of a fax)
    • Similarly, at times and for the same above reasons, Kira may not identify (and extract) the entirety of a clause

When these tools were first introduced there was a lot of speculation, and skepticism, as to the accuracy of the AI involved. Over time and in practice this speculation and skepticism, has not been borne out. Kira, and tools like it, are not infallible, but they are as accurate as a manual review in the areas in which these tools have been trained.

How to Make Kira Work:

  • Kira is not a silver bullet solution to expediting the DD process and cannot, by itself, render an otherwise inefficiently run matter profitable
  • Instead, Kira is a tool which matter managers can deploy as part of a process which is employs some legal project management principles to ensure smooth and timely delivery of the DD, and a profitable outcome for the firm
  • Matter managers should first scope the matter with the client to understand what is material to the client and the transaction, what level of diligence and reporting is required, and who/what should be doing the reporting
    • In limited circumstances, the client might be happy for Kira's AI to analyze the documents and extract the material clauses for a lawyer to then cast an eye over and highlight any potential issues
    • Similarly, a client may wish to use the heat map functionality within Kira to report on an exceptions/deviations only basis. This operates by using an approved template document as the base and layering similar documents on top of the base to get a heatmap highlighting deletions, amendments, or additions made to the documents under review
      • For more on scoping, see our article here and speak to us about Equisphere, and our methodology for more effective matter management and delivery
  • Once the matter has been scoped, and assuming that there is sufficient volume to warrant deploying Kira, you need to assemble your team based on the level of diligence reporting you need to deliver and what you are reviewing
    • DD exercises are generally tight on budget, so wherever possible deploy the cheapest members of your team to do as much of the work as possible. For example, subject matter experts ("SMEs"), who are generally more expensive, should not be conducting the first level of review of standard/boiler plate agreements, as this will disproportionately and unnecessarily eat into your budget
      • Instead, use the tagging system to have your first line reviewers tag documents/clauses which require review by an SME. Utilizing the tagging system as a method to escalate documents to SMEs ensures that your team is properly leveraged and the budget is well managed
  • Now you need to upload the documents to the platform, and Kira's OCR technology will process them
    • Kira integrates with the most common data room providers (Intralinks and Merrill), as well other platforms such as HighQ, which allows for the quick and efficient download and processing of documents within the platform in just a few clicks
      • Setting up the project within Kira so that it has a coherent structure is important, and helpfully Kira can natively bring in folder structures already built in the above platforms so there is no duplication of effort

This feature is really helpful. Often a lot of time is spent organizing a data room into a structure which makes sense, due to the volume of documents involved. The ability to take all or just the material parts of that structure into the Kira saves a lot of time, and ensures that there is no inconsistency between the main data room and the Kira project

  • As noted above, set up the necessary work sheets and apply them to the documents. Kira will then set about analyzing the documents and will provide you with an overview of what is in them
    • Kira's dashboards will automatically provide a quick overview of key pieces information relating to the DD upon processing of the documents. These dashboards can edited and added to so that they display custom, project specific views, and are useful to keep the team aware of the progress of the DD project
  • Use Kira's project management features to batch documents to reviewers, and to track the progress of the DD. These also allow the project managers to filter the documents by tags placed on them, which ensures that any escalation to SMEs is effectively managed
  • Be specific in what your team needs to report on, and where possible direct what they will report: if doing a red flag DD for example, provide the review team with the means within Kira to indicate a red flag or particular issue whether using the flag button or (better yet) an agreed set of tags
    • Using an agreed set of tags at the beginning of a DD ensures that all are reporting in a consistent manner, and real issues can be brought to the fore early. This consistency of approach also ensures less training for new hires, and it can also be reutilized on later DDs which should improve efficiency
  • Once the review is complete you can extract the users reports into Word or Excel formats. There are number of options for what information to extract, including the summaries only or the tags used
    • As mentioned earlier, Kira also integrates with HighQ, allowing matter managers to push information to an i-Sheet in the HighQ Collaborate platform. This requires an extra module within the HighQ system, however it opens a lot of options to further expedite the production of the DD report, especially where the firm also has Contract Express
      • If you don't have HighQ but you do have an document automation platform such as Contract Express, you can expedite the production of your DD report using the extracted information from Kira contained in an Excel spreadsheet

The extraction process is Kira's weakest element. Most large law firms have spent a lot of time and money in creating beautiful branded templates into which their DD, and other reports, go. Kira does not allow for their users to extract information directly into their firm templates, which means that there will be a considerable amount of time spent copying information extracted from Kira and pasting it into a new template.

Plans and Pricing:

  • We do not have specific information on pricing from Kira, however we know that Kira's model is that customers can purchase a bundle of documents which they can process through the system on a yearly basis for a fee (i.e 10K documents for X amount of $)

Used by Member Firms:

  • Bowmans (member firm for South Africa)
  • Arthur Cox (Ireland)
  • Baker Botts L.L.P. (USA)
  • Morrison & Foerster LLP (USA)
  • Womble Bond Dickinson (USA)

Contact Us:

  • There is more we could tell about Kira, and tools like it, and how to get the best out of those tools but instead we would prefer to talk to you. So, if this article has been helpful and you would like to learn more, schedule a consulting call with us here