For best results with GenAI, law firms should make “IA before AI” their mantra, starting by asking the following questions:
1. Where’s the good data?
Sounds basic, but law firms should start by locating their trusted data. Where does the “good stuff” live?
Firms with a document management system (DMS) have a leg up because they already have a single centralized location for their important files, templates, knowledge assets and best practices documents rather than having them scattered in different places.
Additionally, key business functions may be run in spreadsheets stored in the DMS. These should be identified and marked out uniquely as valuable resources.
2. What qualifies as good data?
GenAI isn’t going to be able to determine on its own which data is most useful and which isn’t. Human involvement is key, particularly when it comes to filtering the data. Speaking of filtering: Giving GenAI access to all files in the DMS will likely overwhelm the underlying large language model (LLM) with too much information.
It’s far better to provide a subset of data, like final approved versions of documents, rather than each and every draft that led up to that final product. Limiting content to more recent data is also a good idea due to the constant evolution of laws and regulations, which can make older documents out-of-date or no longer applicable.
3. How accessible is the data?
Depending on the firm’s security model, some data may be restricted, leading to inconsistent AI responses based on who within the firm is asking GenAI a question and what kind of access levels or permissions that person has.
GenAI isn’t going to be able to determine on its own which data is most useful and which isn’t. Human involvement is key, particularly when it comes to filtering the data.
Adopting a more open security posture for items like best practices and knowledge assets ensures uniform AI responses while balancing security with accessibility. Be sure, however, to treat sensitive business data like billing, salary and employee information with the same confidentiality as client data, only using abstracted data as needed for AI.
4. How can process support the data?
While humans and technology play crucial roles in strengthening the firm’s IA, there also needs to be a change in processes.
Workflow automation can be particularly useful here. Think about the valuable data in a typical set of closing books. Processes can ensure that documents in that closing set are saved in a specified location in the DMS, preventing them from being misfiled or not filed at all because legal professionals are stretched thin and have too many to-do items on any given day.
Similar process automation can ensure that important email attachments don’t disappear when they’re filed into the DMS — the actual documents in the attachments will be properly saved and filed so that GenAI can access them.
It’s best to keep the processes around keeping data flowing in the right direction simple so that people will actually be a part of the solution when it comes to building up and maintaining the information architecture. A great user experience for any supporting technologies is key here so that people can easily get valuable data to the right location, rather than finding workarounds that might just leave the data improperly filed or otherwise stranded so that it can’t be leveraged by GenAI.
THE FOUNDATION MATTERS: DON’T IGNORE IA
Information architecture is just like real-world infrastructure. It’s the critical foundational layer that needs to be put in place before you can build on top of it. And if the foundation is shaky, it doesn’t matter what type of house you build — you’re going to run into problems sooner or later.
By prioritizing a solid IA framework, firms enhance their technological foundation and ability to leverage GenAI to its fullest extent, paving the way for better business results moving forward.
