The latest time warp between social media feeds and our day to day experience talking to customers and prospects about workflow is upon us. On the one hand the promise that generative AI will take away all the pain of structuring your data and workflows and bring powerful automation to the masses (see MS CoPilot etc); on the other hand, operations professionals implementing intake projects to move work from email and Excel to a modern workflow or CLM platform. Let’s address AI once we’ve tackled the basics: where are we? What can we do today and how are businesses utilising workflow automation?   

 We are in the midst of a 20-30+ year transformation, moving work inputs and outputs from paper artifacts to ‘digital’: understanding our data, ascribing meaning to it, and using it to drive decisions or actions. We have had hiccups, for example mass PDF scanning and formatting which allowed greater access but gave no better understanding as to what was in our contracts. Every large scale regulatory change (GDPR? LIBOR?) we fail to learn that lesson en masse as we pay as an industry countless $mm in discovery to extract contract data because we’re still not sure.  

Our transformation is somewhere in the messy middle. We have progressed leaps and bounds in both capturing data upfront (structured intake), and extracting it (content extraction/ML products). Workflow practices have evolved alongside the technology, where common flows such as contract intake and negotiation, or claims processing, have reached the stage that you’d broadly recognise similar stages across different providers or corporates, but the terminology might differ. Most ongoing transformation efforts still involve shifting from early workflows which replicated manual processes into email and Word/Excel, into newer platforms which structure the data, documents or workflow, allow easier collaboration – sometimes in a more pleasing online environment – and integrate across enterprise systems. This is Autologyx’s own mission, alongside speeding up the change effort with ease of configuration.  

You might be struck that generative AI aside, the biggest recent trends in technology have not, on the whole, been ‘new’ features: they have been about ease of configurability and making these features commonly available outside of core IT functions; hence the ‘citizen developer’ and ‘no/low-code’ movements. This is a worthy cause we believe in: what use is a nicely designed online intake process if it means every legal function has to have ongoing access to python or javascript developers? Adoption is everything, and ease of configuration sits at the heart of this alongside change management skillsets. The answer to ‘what can we do today’ is perhaps unexciting: not hugely different to what some businesses could do for the last c.15 years; the biggest change is now everyone can. 

The following is within reach of non-technical business analysts:  

• Portal / intake configuration: landing pages, intake forms, look & feel 
• Complex form design (including embedded conditional logic between sections or fields) 
• Workspace and record configuration (e.g. record view or layouts) 
• Analytics set-up (either embedded or exported to BI tools such as PowerBI or Tableau) 

• Workflow triggers, rules (e.g. triage of work, documents or data depending upon conditions) 

The ability to configure combinations of the above covers most legal workflow and collaboration, and once implemented brings dual benefits of structured outputs and understanding of how work is being delivered. Businesses are putting this to use in both internal and external facing use cases: contract lifecycle management, collaborative external-facing workflows around specific matter types, or dedicated customer-facing knowledge products are all large areas in their own right which share the same basic features.  

So what about AI? New generative LLM capabilities are taking the world by storm, so much so that until the dust has settled we might be left wondering where to invest our time. Technology providers have started to use it to enhance existing products, where use cases such as natural language question/answers, extraction, analysis and summary fit alongside creating or storing contracts or other documents.  

Our journey towards structure and understanding will at the very least speed up dramatically; but I would also look to areas of huge productivity gain off the current ‘map’ of structured processes – for example writing an email taking certain factors into account is an integral day-to-day task which these products are very well suited to. Perhaps what we currently call ‘human in the loop’ workflow automation – the combination of technology and human judgement to get the job done – has just had a big boost to the one part of productivity which automation couldn’t previously touch: accurately replicating the ‘human judgement’ part by generating content based on data sets so large that we can’t tell the difference. Time will tell where the biggest impacts will be, but the trend is always towards more structure and understanding, not less.