The legal market has spent the last two years chasing new AI products, but many lawyers still have not changed how work really happens. The next phase of legal innovation, says iManage adviser Ivy Grey, should be less about adding more tech and more about optimizing AI-enabled workflows for attorneys.
Building Accountability Into the AI-Enabled Workflow
“Workflow” is often pictured as a technical diagram filled with the steps and building blocks to be assembled into an automation. But identifying those building blocks takes you only so far. You need to get them into the proper order. And the more important part of workflow is getting the handoffs right — understanding how work gets handed off to another lawyer, to a junior associate, to a piece of software, to an AI agent.
That’s the piece that needs attention now, and it’s why “accountability by design” is essential for increasing lawyers’ adoption of artificial intelligence, and actually improving how work gets done.
Trusting a person or an AI agent to act on one’s behalf requires a way to build accountability into the natural course of doing the work — affirming, as part of drafting or reviewing, that a fact has been checked and a citation is accurate. The more typical approach bolts accountability onto the end of the process: Do the work, then audit it, then document that the audit happened.
That structure is no longer tenable in the AI era.
Accountability Leads to Efficiency
Consider a litigator drafting the factual background of a brief, working from “the record” — everything filed with the court, plus supporting material pulled from a document repository. Suppose the narrative states that a mortgage payment of $3,510 arrived on a specific date. In the paper era, that sentence needed little scrutiny. The underlying check existed, physically, and no one worried its amount had been invented.
Once an AI system drafts that sentence, the calculus changes. The amount, the sender and the date all need confirming against the source, preferably with a link back to the document.
The same discipline applies to legal requirements, not just facts. Suppose the argument is that the payment arrived after a contractual cure period expired, thus the contract was breached, and the creditor can now pursue its collection remedies. An AI tool might accurately conclude that the prerequisites have been met and that foreclosure is now proper.
Confirming the analysis and application of lien law is correct remains a professional obligation.
A junior associate, for example, should check the facts and mark them as verified by inserting their initials. A senior lawyer should never have to personally recheck that kind of foundational point. They should be able to trust that the associate has done the legwork and that they can pick up where the associate left off as the work is handed off.
Without an accountability structure like the above in place, AI creates a strange inefficiency: Multiple lawyers can spend excessive time verifying every line. Time spent checking for invented case law, invented facts or other hallucinations to make sure that whoever was entrusted to do the work didn’t solely rely on AI and performed the necessary verification in a way that’s recorded and visible.
That’s a phenomenon known as the “verification-value paradox.” It exemplifies how quickly the expected efficiency gains from AI disappear if accountability by design isn’t part of the workflow.
A Few Best Practices Go a Long Way
Workflow optimization requires taking a look at the handoff itself.
Poor handoffs are among the most common (and most avoidable) pitfalls when leveraging AI for various pieces of workflow.
For instance, lawyers routinely tell a colleague or an AI tool to “just do X” without mentioning the 25 steps that have to happen first. A good handoff spells out the goal, the sources typically needed, and how to make a judgment call if a question arises. That context isn’t simply a courtesy. Providing context turns a vague instruction into a task someone else can successfully complete.
This gets at the heart of a distinction many lawyers blur: Delegation is not assignment.
- Delegating a task means handing it off with enough context to do it well, and a clear enough picture of “done well” to judge the result when it returns.
- Assigning a task means handing it off and hoping for the best.
A useful rule of thumb: If a lawyer cannot judge whether a task was done well or poorly, they do not have the competence to delegate it, thus it was assigned, not delegated.
Related: “Delegating Legal Work: 7 Tips to Scale Your Practice” by Karen and David Skinner
Respect how people already work.
If information normally comes from email, the workflow should pull it from email rather than rerouting lawyers to a new application. Keeping the steps familiar and building accountability into the ones already there (rather than creating new, burdensome processes) increases the likelihood of success when reengineering business processes and attorney workflows.
Lawyers change behavior when a new approach is aligned with how their work gets done. A workflow that builds verification into work lawyers are already doing, in tools they already trust, stands a far better chance of sticking.
Image © iStockPhoto.com.
Sign up for Attorney at Work’s daily practice tips newsletter here and subscribe to our podcast, Attorney at Work Today.
