The 7% Solution: Why Most Legal AI Projects Fail to Deliver

The research, published in Axiom’s 2026 In-House Legal AI Report, highlights a disconnect between investment and impact. Although 83% of legal departments cannot verify if their AI expenditures yielded a positive return, many continue to pour resources into tools they have yet to master. The data suggests that the most effective teams treat AI as a core strategic program with dedicated ownership, while the majority treat it as an isolated side project.

Structural hurdles further complicate the landscape. In most corporations, the decision-making power regarding AI software lies outside the legal department, leaving general counsels to manage tools they did not choose. Furthermore, two-thirds of teams deploy general-purpose models like ChatGPT or Microsoft Copilot without specific legal configuration. This lack of customization, combined with insufficient staff training, means that even paid-for technologies remain underutilized.

Success, according to the report, depends on five specific habits: investing in ongoing training, running structured pilots, establishing governance before deployment, tying AI to concrete legal outcomes, and seeking external guidance. As firms struggle to extract value, 98% of leaders indicate that outside expertise is essential for navigating tool selection. Currently, legal departments demonstrate a clear preference for alternative legal service providers over traditional law firms to bridge this capability gap.