This article is sponsored by iManage
ChatGPT might have been the first company to burst open the generative AI floodgates, but an increasing number of vendors are incorporating this technology into their offerings, leaving law firms to wonder: what now? Should we be getting on board with generative AI? Are we even ready to get on board with generative AI?
In situations like this, it pays to have a roadmap. By carefully following it, firms can ensure that they can get started with generative AI and take advantage of its benefits without making too many wrong turns along the way.
Now: Ask the questions
Before jumping into generative AI simply to jump into generative AI, firms should take the time to ask: What business problems are we actually trying to solve – and where can AI assist with those problems? What are the high value applications of AI, and which applications are more or less cosmetic?
This might sound like straightforward advice, but some firms are in such a rush to get started with generative AI that they don’t take the time to really understand the strengths and limitations of this technology – as well as what prerequisites are necessary to get the most out of it. Firms that pause to think about what exactly they’re trying to accomplish will be setting themselves up for success as they journey down the road.
Next: attend to information architecture
As a next step, firms need to check the state of the information architecture within the firm. The large language models (LLMs) that power generative AI and generate content need to be trained on data, which begs the question: what are the trusted data sets within the organization? For that matter, where are the trusted data sets?
A document management system (DMS) gives a firm a good foundation to build off of, but it’s just a starting point. Generative AI won’t necessarily be able to find the “signal” within the “noise” of millions of documents within a firm’s DMS.
A better approach is to expose LLMs to a small subset of data – such as the final versions of documents from within a specific time range, rather than all versions of documents, stretching back years and years.