Some of the main business performance benefits they’re looking for from AI are to reduce manual work and improve accuracy. 35% of respondents indicated that the former was a primary reason end users in their organization were interested in using AI, while 42% cited the latter as a key driver.
But inside many organizations, that enthusiasm runs into a more complicated reality: concerns about the reliability of AI output and uncertainty over which use cases actually make sense. 29% of respondents cited “lack of trust” as one of the biggest challenges or concerns that end users had related to using AI, and 30% cited lack of clarity on use cases.
For legal administrators tasked with supporting the organization as it navigates the AI era — while simultaneously making use of AI tools themselves — addressing the disconnect between “wanting AI help” and “not fully trusting the results” is quickly becoming an essential part of the job.
Where the Disconnect Begins
Closing this gap requires a closer look at how organizations manage their knowledge and how sophisticated they are in those efforts — because that underlying knowledge serves as the foundation for AI.
Firms and legal departments can be thought of as sitting somewhere along a maturity curve, from organizations with fragmented, inconsistent systems to those with centralized, well-governed ones. The less mature an organization's knowledge infrastructure, the more acute its AI challenges tend to be.
In less mature environments, professionals can spend significant time simply locating documents and information scattered across disconnected systems. That alone creates a drag on productivity, but the problems compound from there. When institutional knowledge is scattered, any attempt to pilot AI runs headlong into a "garbage in, garbage out" problem: It becomes difficult to point AI at an organization's best work product when no one is entirely sure where that work lives or which version is current.
Technology culture compounds the issue. In less mature organizations, end users tend to resist new tools, with research indicating that only around 40% eventually adopt them, compared with more mature organizations where professionals often embrace new technology even ahead of a formal rollout.
This combination of fragmented systems and hesitant adoption makes it harder for these less mature organizations to realize the benefits AI is supposed to deliver and to build trust in its outputs.
Building a Foundation Before Adding Tools
For legal administrators evaluating how to get the most out of AI for themselves and others in the organization, the research points to a clear starting point: Get the underlying knowledge infrastructure right before layering on new technology.
That typically means consolidating data into a single, centralized repository rather than continuing to operate across siloed systems — a shift that supports a broader culture of sharing and reusing institutional knowledge.
From there, the instinct to add more tools should be resisted in favor of a more deliberate step: understanding how work actually gets done.
The less mature an organization's knowledge infrastructure, the more acute its AI challenges tend to be.
Any legal administrator — from the IT department to the billing department to HR and beyond — follows well-worn processes without necessarily recognizing them as such, just as many of the attorneys do for their work. Mapping these workflows in detail rather than assuming which tasks are ripe for automation is essential before any AI rollout begins.
That mapping exercise should be used to separate high-volume/low-risk work from high-risk/high-value work. Generating routine documents, such as standard HR forms or customer invoices, tends to fall into the first category — repeatable tasks with predictable outputs where AI can create real efficiency without compromising judgment. More complicated workflows, meanwhile, should remain firmly in the hands of the humans: a multi‑step client onboarding with regulatory checks, for instance, or coming up with a bespoke pricing model for clients.
Note that even in the well-defined high-volume/low-risk processes, there will be edge cases that require human review. Designating a risk steward, compliance lead or other reviewer to stay "in the loop" helps ensure that AI systems continue to perform as intended and gives legal professionals confidence in the results.
Build Trust to Drive Better Outcomes
Ultimately, what legal administrators are looking for is straightforward: AI that eliminates administrative drag without replacing the judgment and expertise that still must come into play, so that both they and the lawyers can trust the outputs that AI is providing them.
Organizations that have invested in strong knowledge foundations, taken the time to map how work actually happens and targeted the right workflows first are already seeing that payoff. For legal administrators guiding these efforts, that combination is what ultimately builds trust in AI and translates AI investments into better business outcomes.
