There is a number that recurs in the professional services literature on AI-assisted billing: 8 hours per fee earner per week. It is the average gap between the work actually done and the work actually recorded for billing. The number comes from comparing reconstructed time entries, generated from digital activity across email, calendar, documents, and calls, with the time entries the fee earner actually submitted. The gap is consistent enough to be predictive. The recovery is not theoretical, but the mechanism deserves more scrutiny than it usually gets.
Where the eight hours actually come from
Three sources, in declining order of magnitude. First, micro-tasks: ten-minute emails, fifteen-minute calls, twenty-minute document reviews that do not get recorded because the act of recording them takes nearly as long as the task itself. Second, context-switching tasks: client work that happens during nominally administrative time and gets coded as administrative because the fee earner did not separate the two. Third, after-hours work: drafting and review that happens in the evening or on weekends and gets either omitted or compressed into the next billable day's entries. Each of these is a recording problem rather than a working problem. The hours were already worked. They were just not captured cleanly enough to bill.
The hours were already worked. They were just not captured cleanly enough to bill.
The Clio data on this
According to Clio's 2024 Legal Trends Report, the average lawyer records 2.9 hours of billable work in an eight-hour day, leaving roughly 5 hours unbilled per day [1]. The 5-hour figure is striking but combines two distinct phenomena: time genuinely spent on non-billable activities (administration, business development, training) and time spent on billable activities that does not get recorded. AI-assisted timekeeping addresses only the second category. Industry analyses suggest the recoverable portion is in the 1 to 2 hours per day range, or 5 to 10 hours per week, which is consistent with the 8-hour benchmark. The 2025 Clio update found that AI adoption among legal professionals had reached 79 percent, with growing firms using time-saving automation tools nearly three times as much as shrinking firms [2]. The link between adoption and growth is correlational, not necessarily causal, but the directional signal is clear.
What AI-assisted time tracking actually does
A working AI-assisted time-tracking system monitors digital activity in the background, identifies likely billable tasks from activity patterns, and proposes time entries with narrative descriptions for the fee earner to review and approve. The key word is propose. The fee earner remains the decision-maker about what gets billed, but the cost of recording drops to near zero, which removes the under-recording driven by recording friction. Reuters Future of Professionals Report 2025 found that productivity tools in law have the potential to save lawyers nearly 240 hours per year [3], much of it in this category. The mechanism is not magic. It is the elimination of the small frictions that aggregate into large under-recording.
Why the productivity narrative misses the point
Most coverage of AI in professional services focuses on productivity: AI lets lawyers draft contracts faster, accountants review filings faster, consultants build models faster. The 2024 Clio Legal Trends Report found that up to 74 percent of hourly billable tasks could be automated with AI [4]. But the productivity gain creates a pricing problem under hourly billing: faster work means fewer billable hours, which reduces revenue under the same fee. The billable-recovery story is the opposite. It recovers revenue from work that was already happening, without changing how long the work takes. For firms that have not yet migrated to flat-fee structures, billable recovery is a cleaner near-term win than productivity, because it does not threaten the existing revenue model.
The math for a mid-size firm
For a firm with 40 fee earners averaging a EUR 250 effective hourly rate, recovering 8 billable hours per fee earner per week represents EUR 4,160,000 in annual recovered revenue at full realisation, or roughly EUR 3.5 million after typical 85 percent realisation. The implementation cost for AI-assisted time tracking is modest by comparison, typically in the range of EUR 50 to 150 per fee earner per month for the leading platforms. The payback period is measured in weeks, not quarters. The reason this is not yet universal is not economics. It is adoption.
The economics are not the constraint. Adoption is. And adoption is mostly a privacy-architecture conversation.
The adoption barrier
AI-assisted time tracking requires the fee earner to share digital activity data with the system. Privacy concerns are real, both for the fee earner and for client confidentiality. The leading platforms have addressed this with local processing, encrypted storage, and client-side filtering, but the perception lags. The 2025 Clio data showed that only 40 percent of legal professionals are using legal-specific AI solutions, down from 58 percent in 2024, as more lawyers shift to generic tools like ChatGPT that lack the privacy and data-protection features of purpose-built legal AI [2]. The trend is concerning. AI-assisted time tracking specifically benefits from legal-specific deployment because the privacy architecture is built into the product. Firms that deploy the right tools see the recovery. Firms that deploy general-purpose tools see less, and risk more.
What the partner needs to do
Three things, in order. First, choose a tool that is privacy-architected for professional services use, not a general productivity tool. The financial benefit is similar across tools, but the risk profile is not. Second, set the firm-level policy on what activities are eligible for automatic time-capture and what activities require explicit manual entry. The line varies by firm but the policy needs to be explicit, not implicit. Third, monitor the post-deployment realisation rate, not just the recorded-hours rate. The recovery is real only if the recovered hours actually invoice and collect. If the hours go up but the realisation rate drops proportionally, the recovery is illusory. The firms that have done this well track both numbers weekly during the first quarter of deployment and tune the policy based on what actually invoices.
The closing observation
Eight billable hours per fee earner per week is not a hypothetical. It is the consistent finding across deployments in mid-size professional services firms. The recovery is durable because it does not depend on changed behavior; it depends on reduced recording friction. The firms that capture it are not working their fee earners harder. They are billing for work that was already happening. For Belgian and Iberian professional services firms specifically, where partner economics are under pressure from both client procurement and talent costs, this is one of the most accessible margin-improvement levers available. The investment is small, the implementation is well-understood, and the result is measurable within a quarter.
