July 2026 legal tech recap for in-house teams: consumption pricing, Microsoft picks Harvey, agent guardrails, the EU AI Omnibus in force, and more AI sanctions.

Legal Tech Rewind is our month-by-month look back at 2026 for in-house legal teams: what happened, why it mattered, and what it looks like with the benefit of hindsight. Published October 2026.
AI usage got big enough that vendors started rethinking how they charge for it, clients started asking where their AI savings went, and the agent conversation moved from capability to control.
Harvey co-founder Gabe Pereyra told Artificial Lawyer the number of tokens processed on Harvey had risen 14x in six months and that a new pricing model was in the works. Legora was also moving toward pricing that reflects consumption. Pereyra added that law firms passing token costs on to clients "might become more common."
Not everyone agreed. LexisNexis, reporting its fastest growth ever (Legal revenue of £959 million in the first half, up 10% underlying, with 90% of new business coming from AI-related products), said token costs were under 1% of its costs and it had no plans to move to consumption billing. It prefers three-year subscriptions. LegalOn went the other way too, shipping more than 100 attorney-built AI workflows under flat per-user pricing.
Why it mattered: your AI budget used to be a seat count. Now it might be a meter, and a meter that sits inside your outside counsel's invoices as well as your own.
Since then: in September, Gartner predicted that consumption-based pricing will make up over 35% of net new corporate legal tech spend by 2028. If you haven't set a position on AI pass-through charges in your outside counsel guidelines, see our 2027 budget template.
The Thomson Reuters Institute's Future of Professionals 2026 research (July 21) found that 70% of corporate respondents expect AI productivity gains from their law firms, but only 6% say most firms are delivering them. Almost a third said they're reconsidering firms that can't show AI value.
Why it mattered: in-house teams finally had data to back up what they suspected: firms are adopting AI faster than they're passing on the savings.
Microsoft's Corporate, External, and Legal Affairs group, about 2,000 lawyers and compliance professionals, selected Harvey. Harvey said more than 1,000 in-house legal teams now use it, and it reported more than $100 million in net-new annual recurring revenue in Q2. Separately, Harvey bought asset-management startup Benchmark, its third acquisition of the year, and Legora bought litigation fact platform Wexler, its fifth.
Why it mattered: the company that sells Copilot picked a specialist legal agent to run on top of it. "Copilot plus a specialist legal AI" is starting to look like the default architecture for large legal departments.
Why it mattered: the honest answer to "who's responsible when the agent gets it wrong?" was "it depends on your contract." Most contracts didn't say.
Flank launched Record, an agent-maintained contract system of record for in-house teams. Agents connect to e-signature, email and shared drives, capture metadata at signing, file and tag documents, and reconcile duplicates daily, with a lawyer supervising. Separately, Theorem relaunched as a free procurement marketplace for in-house teams to run legal tech and outside counsel RFPs.
Why it mattered: the oldest CLM problem is that nobody keeps the repository up to date. Agents might finally solve that, at least for teams willing to supervise them.
The EU's Digital Omnibus on AI was signed on July 8 and entered into force on July 27. High-risk obligations moved to December 2, 2027 (stand-alone systems like hiring and credit) and August 2, 2028 (AI in regulated products). But most Article 50 transparency duties, such as chatbot disclosure and marking AI-generated content, still applied from August 2, 2026.
In LeDoux v. Outliers (W.D. Wash.), Judge Tiffany Cartwright sanctioned a plaintiff's lawyer on July 24 for dozens of inaccurate or hallucinated citations across at least five filings. A month later, the court excluded the plaintiff's expert because his report cited AI-hallucinated academic articles, and the claims were dismissed with prejudice.
Why it mattered: AI errors now sink cases through expert reports too, not just briefs. Your expert-vetting checklist needs a citation check.
July's lesson: if AI is metered, someone has to read the meter. Make sure it's you.
More from the 2026 Legal Tech Rewind: ← June 2026 · August 2026 → · The full 2026 Rewind: 9 lessons
This article is general information, not legal advice.