NetDocuments yesterday (9 November) held its EMEA user conference in London, where the focus was very much AI, including generative AI-powered, no-code document and workflow automation tool PatternBuilder MAX, which was made globally available at the end of October.
From conversations at the conference with NetDocuments’ partners, PatternBuilder MAX has attracted the attention of the market, and NetDocuments has been in POCs with a number of law firms since launching it in beta in July. You’ll recall that PatterBuilder was borne out of NetDocuments’ acquisition of Afterpattern in 2022, and one independent legal technology consultant observed privately at the conference how good it was to see NetDocuments leverage that acquisition so well.
In a briefing given to Legal IT Insider during the conference, vice president of product, Jill Schornack and product manager Scott Kelly discussed some of the early emerging use cases for the tool, and how conversations with law firms have reached an inflection point, effectively shifting from ‘should I’ to ‘how should I?’
One interesting observation made by Schornack and Kelly at the outset is that PatternBuilder MAX, which is built on Microsoft Azure OpenAI Service and leverages GPT-4, has been given an exemption to Azure OpenAI’s abuse monitoring, which detects violations of its code of conduct. Schornack said: “Client information or language used in the likes of a deposition might flag a potential violation of Microsoft’s policies, which means it would be sent to Microsoft for evaluation.” The confidentiality of legal documents clearly means that this is not an option. However, Schornack continues: “Microsoft doesn’t want to open the floodgates but they appreciate advocacy from vendors like us and we were granted an exemption.” To learn more about content filtering see: https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy
In terms of customer traction to date, Schornack and Kelly point out that PatternBuilder MAX has only been widely available to the market for a few weeks, but say they have been working very closely with firms during the POC stage.
Schornack says: “The conversation has reached an inflection point: it has moved from ‘what should our strategy be?’ to asking which use cases are day one use cases and how can I use it in my day-to-day professional life.”
The question for firms is often around how to get started and Kelly says: “There is so much to learn and people might feel that it’s overwhelming, which is why we created ndMAX Studio and at Inspire we’re releasing nine apps out of the box. What is really cool is that you have access to how the prompt is built and people can tailor them and make them their own.”
The nine apps – listed in full below – were built in partnership with the early users.
In terms of the early use cases to emerge, Kelly say: “Summarisation is a big one. Entity extraction is now far more flexible; we don’t have to train the model with a giant set of documents, you can use it out of the box. We’re also seeing a lot of use cases around drafting, either net new around a precedent or continuing a draft of a document in Microsoft Word where you need ideas for what the next paragraph is. Those are the big categories for uses cases.”
He adds: “My sense is that we’ll see a lot more generative AI being rolled out in production in law firms next year. Firms are just this year becoming aware and learning. They are evaluating and budgeting. Our customers don’t have to know now what the right solution is. PatternBuilder MAX is for building as opposed to some solutions out there that can’t be extended or modified.”
In a demo of PatternBuilder MAX during the conference, Kelly said: “I suggest you prioritise the low hanging fruit, such as non-billable work, which is very uncontroversial and takes your fee-earner away from revenue generation.”
The cost of PatternBuilder MAX is based on an allocation of compute that NetDocuments think will be sufficient to serve its customers’ use cases. Firms that want to go beyond that can pay Azure directly or go via NetDocuments. Kelly says: “What I will say is that over time compute costs will go down, and in a couple of years we probably won’t be talking about it.”
One of the biggest conversations that firms are currently having internally around generative AI is, however, the cost. Schornack says: “Once firms understand PatternBuilder MAX they are able to sell it internally because they can show that this task used to take this amount of time and now it takes 10 minutes.”
The nine Studio Apps are:
- Summarize: Generates a summary of a document that you select in NetDocuments. Firms can generate a general summary or apply a summary template for specialized use cases (depositions, loan agreements, etc.).
- Drafting: Automatically drafts a new document, allowing users to select a prior precedent document as an example or draft a document from scratch based on custom instructions. This can include analyzing a patent application rejection letter and writing an appropriate appeal, taking a fact sheet and drafting a new lease, or rejections to demand letters.
- Extract: Automatically extracts one or more entities from the selected document, and resulting data can then be saved to a database, written into a document, or both. This app is ideal for identifying and storing contract data.
- Contract Playbook: Instantly analyze an agreement based on the selected contract playbook, including organization-specific contract playbooks that analyze any type of desired agreement.
- Timeline: Generates a timeline of events based on the selected text.
- Translate: Automatically translates text from one language into another, supporting over a dozen languages.
- Compare: Compares two documents based on custom criteria that the user provides.
- Continue Draft: Quickly drafts the next section of a document based on prior content in the document. Optionally, users can include prior precedents to enhance the quality of the draft further.
- Query Document: Ask one or more questions of a document stored in NetDocuments, ensuring quick answers even when buried in lengthy, complex documents.
For more details click here.