Legal tech or outside counsel: where to draw the line on document review

An economic equation under strain

The trade-off between in-house tooling and outside counsel is not a new debate, but recent figures give it a particular edge. According to the Thomson Reuters Institute's 2026 report on the state of the US legal market, worked billing rates rose 7.3% in 2025 while inflation hovered around 3%. The institute's Law Firm Rates Report 2026 documents the longer trend: over the past decade, firms have raised rates at least twice as fast as inflation, and average Am Law 100 rates have crossed the $1,000-per-hour threshold.

This dynamic produces a mechanical effect that Thomson Reuters describes as work moving downstream: clients are redirecting part of their legal work toward cheaper firms, their own teams or tools — not because the quality of large firms is in question, but because not every task justifies the same hourly rate. A telling detail from the same report: despite the headline rate increases, average spend per billed hour actually declined slightly in 2025, a sign that clients are now sorting what they send to their counsel.

Document review sits on the front line of that sorting. Reading an NDA, extracting the terms of a credit agreement, checking an amendment against the existing documentation: these tasks are necessary, recurring and hour-consuming — but they rarely call on the expertise that justifies a senior partner's rate.

What legal departments are already insourcing

Surveys of in-house lawyers confirm that the shift is underway. According to a joint study by the Association of Corporate Counsel and Everlaw published in 2025, 64% of in-house counsel surveyed expect generative AI to reduce their reliance on outside counsel, and 50% expect lower outside counsel costs. Respondents identify contract drafting (78%), contract management (71%) and legal research (62%) as the first candidates for insourcing.

These intentions are gradually translating into practice. The State of Contracting survey conducted in late 2025 among more than 450 in-house legal professionals found that the share of teams actively using AI for contract review doubled in a year. These figures call for caution: the same survey shows that active users remain a minority and that most teams are still at the evaluation stage. Adoption is real but early — which also means reference practices have not yet settled.

The common thread across these studies is instructive: in-house lawyers are not planning to replace their outside counsel, but to redefine what they send them. Mechanical, high-volume work moves to tooled-up internal teams; high-stakes or highly technical work stays with the law firms.

The specific case of private debt funds

For a private debt fund, the equation has a particularity: most teams have few or no in-house lawyers. The trade-off therefore plays out not between outside counsel and a legal department, but between outside counsel, investment team time and tools. And documentation volume grows mechanically with the market: Morgan Stanley puts private credit at roughly $3 trillion at the start of 2025, with a projection of around $5 trillion by 2029. More deals means more NDAs, more term sheets to compare, more credit agreements to monitor and more amendments to process over the life of each position.

These documents are not all alike. A sale-process NDA is a short, repetitive document whose points of attention are known and finite in number. A 300-page credit agreement negotiated on a sponsored deal, by contrast, concentrates most of the legal value of the position. In between, term sheets, investment committee summaries and covenant monitoring occupy an intermediate zone: extraction and comparison work that is largely mechanical, but whose conclusions feed investment decisions.

This gradation is precisely what makes the trade-off workable. The question is not whether a tool can replace a law firm on a deal — it cannot — but how much of a fund's documentation volume consists of tasks that no one should be paying for at a corporate law firm's hourly rate.

What outside counsel provides that no tool replaces

It is worth being precise about what tools do not do. Outside counsel brings, first, negotiation judgment: knowing which points to concede, which to hold, and how the counterparty will react. It brings live market knowledge — which terms are clearing right now, with which sponsors, on which segments — that no tool trained on historical documents can reproduce in real time. It brings the ability to structure novel arrangements, where tools reason by comparison with what already exists.

Law firms also carry professional liability. A legal opinion binds its author and its insurance; the output of an analysis tool binds no one but its user. For questions where an error is measured in recovery points — ranking of security, enforceability, governing law in a cross-border structure — that difference is not symbolic. Finally, drafting the final documentation remains lawyers' work: precise definitions, cross-references between articles and the mechanics linking documents together determine how negotiated terms actually apply.

What the tool does that no firm will do at that price

Conversely, certain tasks are structurally ill-suited to the hourly-billed external counsel model. Systematic portfolio review is the clearest example: re-reading the 40 credit agreements in a portfolio to check exposure to a given clause — following a court decision, a regulatory change or a liability management exercise at a comparable borrower — represents hundreds of billable hours if sent to a firm. Few funds commission it, and the question then simply goes unanswered. A document analysis tool runs that sweep at low marginal cost, with a high reliability level on extraction and identification tasks.

The first pass on incoming documents follows the same logic. Triaging the NDAs received in a process, flagging those containing a non-solicit or an unusual exclusivity clause, extracting the economic terms of a term sheet and comparing them to the fund's previous deals: tools like MyClauze, Ontra or Kira are built for this first-level work, which produces a structured base from which teams — and their counsel — work faster. Consistency is a secondary but real benefit: a tool applies the same reading grid to the hundredth document as to the first, where a manual review spread across several juniors produces uneven outputs.

The right question is not "is the tool as good as the lawyer?" but "does this task justify a lawyer's hourly rate?". For a substantial share of a fund's documentation volume, the answer is no.

False trade-offs and real limits

Framing the choice as a substitution would be an analytical error — and yet that is how it is often sold. Three limits deserve to be stated clearly. The first is the need for supervision: the outputs of an analysis tool must be reviewed by someone competent, which assumes the fund retains, internally or at its counsel, the ability to detect a faulty extraction or an overly literal reading of a clause. A tool without a qualified reviewer displaces risk rather than reducing it.

The second is the risk of false confidence. A clean, well-structured analysis report can create an illusion of completeness, while tool reliability varies by task: high on term extraction, more uncertain on interpreting interactions between clauses. Teams need to know what the tool can do and, above all, what it cannot — a distinction vendors do not always have an interest in emphasizing.

The third is the real cost of tooled-up insourcing. Beyond the license, there is configuration time, the building of internal reference bases, team training and process maintenance. The ACC/Everlaw study notes that implementation difficulties remain the obstacle most cited by legal departments. The savings against external fees are real on high-volume tasks, but they are neither immediate nor automatic.

Drawing the dividing line

In practice, the dividing line is built around three criteria: volume, repetitiveness and stakes. High-volume, repetitive documents with limited unit stakes — NDAs, first passes on term sheets, data extraction for reporting — are the natural candidates for tooling, with proportionate internal review. High-stakes work — credit agreement negotiation, structuring, opinions, emerging disputes — stays with outside counsel, who bring a value the tool does not replicate.

The intermediate zone is where the articulation matters most. A tooled-up fund does not send fewer questions to its lawyers; it sends better ones. Arriving at counsel with a documented deviation sheet — this deal departs from our practice on the EBITDA definition, on MFN sunsets and on two baskets — concentrates billed hours on analysis and negotiation rather than reading. That is probably where the most durable gain lies: not replacing the law firm, but no longer paying it for the part of the work that does not require its expertise.

The trajectory of hourly rates makes this trade-off less theoretical every year. Funds that structure it early — knowing precisely what they assign to the tool, what they keep in-house and what they send to counsel — build a cost and speed advantage that others will pay full price for.