92% compared to 72%: AI outscored experienced lawyers on invoice approvals.

Legal Invoice Review Should Optimize for Accuracy, Not Maximum Reductions

7 min read

what should the alt text be for this

Blog

Legal invoice review should be measured by whether the department paid the right amount for the right work, not by how much it cut. Reductions are a frequent and legitimate result of review, but a program that treats accuracy as the primary goal rewards the right behavior, as opposed to savings, produces data that can be trusted and is better received by both insie and outsoide counsel attorneys. 

Ask most legal departments how their invoice review program is performing and the answer is usually a reduction figure: total adjustments, write-down rate or dollars cut last quarter. The number is easy to report upward. It says little about whether invoices were reviewed correctly. 

What is legal invoice review actually for? 

Legal invoice review is the process of checking outside counsel invoices against the terms the company and the firm agreed to, then deciding what is appropriate to pay. Those terms live in outside counsel billing guidelines, approved rate structures and matter engagement letters, all set before the work begins. Both sides commit to something: the firm to bill in an agreed upon way, the department to pay for work billed in alignment with those terms. 

In The Right Invoice: Rethinking Legal Invoice Review, Onit’s Jeff Solomon, VP GTM Strategy, made the point directly: Compliance programs do not exist to save money. They exist to make sure the department pays the right amount for the right work.  

Savings remains a byproduct of compliance. When a firm bills for something the guidelines exclude, the correct invoice tends to be a smaller one. However, the reduction is the result of accurate review, not because it’s the objective. 

Why is the biggest reduction the wrong target for invoice review? 

A program measured on write-downs, rather than compliance, often creates three problems. 

It rewards the wrong finds. Reviewers evaluated on dollars cut have an incentive to pursue the reductions that are easiest to defend, rather than the charges that are actually non-compliant. The volume of adjustments becomes a stand-in for rigor. 

It hides inconsistency. Two reviewers can post similar reduction totals while applying the same guideline in different ways. The headline number looks stable while enforcement varies from reviewer to reviewer. 

It strains the relationship the guidelines were meant to support. Billing guidelines are a communication tool between a department and its firms. Compliance with an agreed upon rule is an easy conversation, aggressively going after savings seems more like a monthly negotiation. 

None of this means overbilling should go unchallenged. It means each challenge should rest on what was agreed, not on a target. 

What does accurate invoice review make possible? 

Accurate invoice review produces a high-confidence approval process and cleaner data to rely on afterward. Reviewers approve an invoice because it reflects the applicable terms, not by default. The benefits extend past the invoice in front of them: 

  • Consistent enforcement. The same term is applied the same way across firms, matters and reviewers, which is what gives a guideline meaning. 
  • Reliable spend data. If the department reports on spend by UTBMS task and activity codes, a misclassified line item distorts every report built on it. 
  • Stronger rate negotiations. Accurate billing history gives the department a defensible view of what each firm charged and delivered. 
  • Comparable matter costs. Total case costs can be compared across matters only when each matter was reviewed against the terms that governed it. 

Why is accurate invoice review harder than applying billing rules? 

Accuracy is difficult because many billing terms are contextual. E-billing rules handle fixed checks well, such as catching a duplicate invoice number. Whether a charge is appropriate, though, often depends on the firm, the type of work and what was negotiated for the matter. Onit’s decision framework for business rules, AI and human review covers which checks belong in each layer. 

The terms governing a single invoice can sit in three places: standard outside counsel guidelines, firm-specific or practice-specific guidelines, and matter-level engagement letters. Consider a common example: the general guidelines exclude computerized research, but the engagement letter for one matter permits it. An accurate review honors the engagement letter. A rule that only knows the general guidelines writes the charge down and creates a dispute the department then has to unwind. Why billing guidelines are not enough for matter-level review explains how to bring those terms into the review. 

Accuracy is also measurable. In research by Onit’s AI Center of Excellence, the top-performing model reached 92% invoice-level accuracy, compared with 72% for experienced lawyers reviewing the same invoices. 

Where does AI-assisted first-pass review fit? 

An AI reviewer is most useful in front of the human reviewer, not in place of one. Instead of starting from scratch, the reviewer receives an invoice already checked against the department’s own guidelines and engagement terms, with each finding explained. The reviewer’s job shifts from finding issues to deciding on them. First-pass review also closes the coverage gap that limits most review programs. Accuracy is what makes that wider coverage worth trusting. 

How much the first pass decides on its own is a governance choice. Spend Agent, available for OnitX ELM and Unity ELM, lets Legal Operations set each vendor to one of three modes: 

  • Sentry: Spend Agent identifies findings and calculates recommended adjustments, and a reviewer decides whether to accept them. Eligible adjustments can be shown as recommendations or pre-applied for review. Either way, the reviewer keeps the final invoice decision. 
  • Auto: Spend Agent automatically acts on invoices based on the department’s own guidelines, allowing the reviewer to receive a clean invoice for their final review. 
  • Off: The vendor’s invoices bypass Spend Agent and follow the standard billing workflow. 

These modes are governance choices, not maturity rankings. A department might use Sentry for high-value or sensitive relationships and Auto for lower-risk invoice populations. Engagement documents uploaded to the matter and associated with the vendor are reviewed alongside the general guidelines, and matter-specific terms take precedence where the two conflict. 

How should legal departments measure invoice review instead? 

Measure the quality of review decisions first and reductions second. Start with questions that describe accuracy rather than volume: 

  • What share of invoices are approved without follow-up disputes? 
  • How consistently is the same guideline applied across reviewers and firms? 
  • How often are matter-level exceptions honored on the first pass? 
  • How much reviewer time goes to judgment rather than to finding issues? 
  • Is billing data clean enough to support rate negotiations and panel decisions? 

Savings still belong in the report, measured carefully. Separate findings, recommended adjustments, accepted adjustments and realized savings. A flagged line is not a saving until the adjustment is applied and accepted. Being able to prove negotiated savings to finance and leadership is its own discipline. In an accuracy-first program, savings sit further down the report, as an outcome of review that works. 

Key takeaways 

  • Legal invoice review exists to confirm the department pays the right amount for the right work. Reductions are a byproduct, not the goal. 
  • Measuring review by write-downs rewards easy cuts, hides inconsistency between reviewers and strains outside counsel relationships. 
  • Accurate review depends on applying general guidelines, firm-specific guidelines and matter-level engagement terms together, with matter terms taking precedence. 
  • Better metrics include dispute-free approvals, consistency across reviewers and correctly honored exceptions, with savings reported as findings, recommended, accepted and realized. 

The goal was never the biggest reduction. It was the right invoice. For a practical model of how to get there, download The First-Pass Framework for Legal Invoice Review, which includes a governance worksheet for mapping one vendor or invoice population.