That is CAST’s conservative estimate of the time it would take to repair the technical debt accumulated across the world’s software systems. Traditionally, tech debt was often a conscious trade-off. Developers took shortcuts to get products to market faster, knowing they might need to revisit the underlying architecture later. AI is changing that dynamic. Coding agents can now make thousands of decisions in an instant, introducing dependencies and complexity that developers may never fully see or understand.
Law firms have built up a remarkably similar kind of debt in their support teams.
Over decades, firms have introduced new tools, processes and team structures to solve immediate problems, often without revisiting how they fit into the wider movement of work. An extra inbox here, a manual handoff there, a new system layered onto an old process. Each change may have worked perfectly well on its own, while making the overall workflow progressively harder to see and understand.
Call it workflow debt. The accumulated cost of years of decisions about how work enters the firm, how it is triaged, assigned, delegated, completed, checked and returned.
For a long time, that debt could remain largely hidden. Experienced support professionals knew which inbox to watch, who could actually take on a task and how to move work around the gaps between systems.
The duct tape held because people knew where to apply it.
AI is stretching it.
BigHand’s 2026 Legal Workflow Leadership Report found that 96% of firms are already using AI in support services. Yet only 27% have redesigned roles and workflows in response, and just 41% have formally reviewed support task types to identify where AI or automation can deliver the greatest impact.
That gap matters because completing tasks faster does not necessarily make the workflow faster.
Workflow debt, just like technical debt, can remain hidden for years while a system continues operating under familiar conditions.
AI may reduce a 30-minute drafting task to five minutes. But what happens before and after that task? If requests still arrive through fragmented channels, if triage depends on someone interpreting an email, if assignment relies on knowing who has capacity, or if AI-assisted work creates another layer of supervision and checking, the bottleneck has just moved, not disappeared.
This is the “Theory of Constraints” in practice. The performance of an entire system is limited by its biggest constraint — the weakest link in the chain. Improving one stage does little for overall throughput if the pressure simply shifts somewhere else.
The research suggests some of that strain is already showing. Thirty-five percent of firms report increased reliance on lawyers for administrative work. Forty-six percent say AI-generated outputs require additional supervision, while 45% report more time being spent checking and verifying work.
That is a troubling paradox. AI is being introduced partly to remove low-value work, yet without redesigning the workflow around it, some of that work can simply be redistributed back onto lawyers and support professionals.
It also explains why visibility matters so much.
For many firms, work is visible at the point it is requested and again when it is completed. What happens in between can be much harder to see. Requests move across inboxes, teams and individuals. Priorities change. Ownership shifts. Delays are often discovered only when someone asks where something is.
That makes workflow debt difficult to identify, let alone repay.
Before adding another layer of automation, firms need to understand how work actually moves. How does it enter the support function? How is it categorized and prioritized? What determines where it goes? Can teams see who owns it and where it is sitting? Are similar requests handled consistently? Where are delays, rework and unnecessary handoffs occurring? And do leaders have enough data to measure demand, capacity and turnaround?
Once firms can answer those questions, AI becomes far more useful. It can help clean up intake, support smarter routing, remove repetitive steps and accelerate the right parts of the process. But automation should follow an understanding of the workflow, not substitute for one.
That is the real challenge AI is exposing in law firm support services. Firms can complete individual tasks much faster without improving the speed, consistency or manageability of overall work delivery.
Workflow debt, just like technical debt, can remain hidden for years while a system continues operating under familiar conditions. Paying it down starts with identifying where the real constraints sit and redesigning the process around them.
AI can accelerate work. But the firms that see the greatest return will be those that make sure the workflow itself is ready to move faster.
