What legal leaders can learn from the first wave of AI-assisted review

When it comes to the legal industry, AI adoption depends on who you ask. An 8am 2026 Legal Industry Report found that 69% now use general-purpose AI for work, up from 31% a year earlier. Firms haven’t always been as fast to adopt here. 

Only 46% have implemented those tools at an institutional level, and 43% of respondents said their firm has no formal AI policy and no plans to write one.

The barriers respondents named had less to do with budget than with exposure. Data security came first at 46%, followed by ethical concerns, privilege concerns and a lack of trust in results, each cited by roughly four in ten. Just 24% called cost a significant obstacle.

That gap between personal enthusiasm and institutional caution is sharpest in e-discovery. The Secretariat and ACEDS 2026 Artificial Intelligence Report, published in July, found that 91% of the litigation and e-discovery professionals it surveyed had used GenAI in the past year, and 64% expect their organization’s AI investment to grow over the next 12 months. 

Read together, the reports point to a fairly specific set of conditions for adoption at scale. Legal teams want the speed of AI-assisted review, but they will rely on it for consequential calls only when a lawyer can check the reasoning behind each decision and when client data stays inside an environment the firm controls.

This condition sit at the center of a deployment announced this week Baltimore based legaltech company Altorney regarding McCarthy Tétrault, the Canadian business law firm with more than 800 lawyers. The firm has adopted MARC, a document intelligence platform built by Altorney, for document review and analysis across litigation, investigations, antitrust, tax and data breach matters. So far it has applied the platform to relevance, privilege and personal information analysis.

MARC classifies documents for relevance, responsiveness, privilege, personal information, trade secrets and other matter-specific categories, and attaches its reasoning to each result.

For Susan Wortzman, Partner at McCarthy Tétrault and President of MT>3, the firm’s e-discovery division, that rationale is what makes the output usable.

“MARC provides detailed rationale for every coding decision, making it easy for lawyers to validate its accuracy and spend their time applying complex legal judgment,” Wortzman said. “Throughout, all data remains secure within our own private environment, giving our clients confidence that their data is securely managed.”

She added that the Altorney team works closely with MT>3 “to develop bespoke workflows for highly challenging and unique matters, working within high pressure deadlines,” and that “the cost savings are impactful for our clients.”

Where the analysis happens in the workflow matters as much as how it is explained. MARC can run before documents enter a review platform or on data already hosted there. Running it upstream gives teams a chance to keep irrelevant material out of downstream review, which is typically where volume, and therefore cost, accumulates. 

Running it on hosted data supports quality control on review already under way. The platform can be deployed in a client’s cloud, a client-controlled data center or Altorney’s own secure hosted environment, which speaks directly to the data security and privilege concerns that top the 8am findings.

“We built MARC to give legal teams better intelligence wherever their data resides,” said Shimmy Messing, CEO and co-founder of Altorney, which was founded in 2021.

“McCarthy Tétrault’s active use of MARC for first-pass review reflects the need to turn large, complex document populations into informed legal action earlier in the workflow.”