An associate pulls case notes from your practice management software, pastes them into a chatbot to draft a memo, then copies the result back into the file and checks it against the original because nobody’s sure the AI got it right. That’s not a shortcut. That’s an extra step, and if it’s happening across your firm, it’s probably eating more time than the AI tool is saving.
What’s actually happening
New research from Workday gives this a name: human middleware. It’s what happens when employees become the connective tissue between AI tools that were never built to talk to each other, moving data manually, checking whether two systems agree, and fixing outputs that almost worked but not quite.
The scale is the part worth sitting with. Workday found a quarter of workers spend seven or more hours a week on exactly that, and 62 percent spend at least half their working time translating between disconnected systems instead of doing billable, client-facing work. As a Workday VP summarized it, “too many employees are serving as the human middleware between disconnected AI systems.” For a firm, that’s not just lost time. It’s lost time that was supposed to be spent on a matter.
Why it’s a bigger problem for a firm than the numbers alone suggest
Here’s the part specific to a law office. When staff bridge the gap between disconnected systems by hand, the workaround usually involves moving client information somewhere it wasn’t meant to go, often into a general AI tool the firm never vetted, just to get a draft or a summary the “approved” system can’t produce on its own. That’s a confidentiality problem, not just an efficiency one.
The ABA addressed this directly in Formal Opinion 512 on generative AI, which ties a lawyer’s duty of competence and confidentiality to actually understanding how these tools handle client information, not just whether they’re convenient. An unsanctioned tool being used as a workaround because the sanctioned ones don’t connect is exactly the scenario that opinion is warning firms about.
What to actually do about it
Before adding another AI tool to the stack, look at where your staff are already improvising: which platform doesn’t talk to which, and what they’re pasting into what to bridge the gap. A firm-approved AI tool that actually connects to your practice management and document management systems beats three disconnected ones your staff have to translate between by hand, and it keeps client information inside systems you’ve already vetted instead of wherever someone found a workaround.
If you want a clear picture of where your team is quietly doing that translation work, and what it’s actually costing you in both time and exposure, that’s the review we do.