When a law firm commits serious money to artificial intelligence, the conversation in the partners’ meeting usually turns quickly to return on investment. Baker McKenzie has decided that the most useful measurement comes earlier than the final financial calculations. Before the firm can put a credible number on productivity gains, cost savings, or improved client outcomes, it first has to know whether its lawyers have actually absorbed the tools into their working routines. To answer that question, the firm has turned to its global training operation, making education the primary gauge for whether an AI investment is likely to pay off.
Seen one way, that decision sounds modest. Seen in the context of an industry that has spent years chasing meaningful adoption metrics, it looks more like a strategic repositioning. Baker McKenzie’s message to the market is that AI value cannot be assessed from the procurement side of the table alone. Value is created at the point of practice, and the most reliable early signal of that value is a lawyer’s demonstrated willingness to use the technology in real work.
The Measurement Problem at the Heart of Legal AI
The difficulty of proving AI returns has become one of the most discussed topics in legal technology. Traditional firm software delivered returns that were comparatively easy to track. E-discovery tools reduced document review hours that were already being logged and billed. E-billing systems generated immediate, visible savings on invoice processing. Generative AI is different in kind. It touches the most valuable asset a law firm sells — the professional judgment of its lawyers — and its effects are spread across thousands of individual tasks rather than concentrated in one discrete workflow.
That creates a measurement challenge on several fronts. Time savings are difficult to isolate because an AI-assisted piece of work still involves a lawyer’s review, editing, and judgment. Quality improvements often surface only over the long term, in the form of stronger work product, fewer revisions, or faster turnaround times. And cost reduction, the easiest metric to cite, can be undermined by the fact that AI tools are frequently layered on top of existing workflows instead of replacing them outright.
Law firms are also learning that seat counts are a poor proxy for value. A firm can deploy hundreds of AI licenses and still see modest real-world impact if lawyers open the tools once, find the experience underwhelming, and quietly return to established habits. Utilization data captures whether someone clicked a button; it does not capture whether the underlying work actually changed. Baker McKenzie’s training-led approach tries to close that gap by focusing on something deeper than raw logins: competence, confidence, and repeated practice.
How Baker McKenzie Uses Training to Gauge AI Investment Value
Baker McKenzie has built its approach on a direct chain of reasoning. Lawyers who understand a tool are far more likely to use it. Lawyers who use it repeatedly become more efficient with it and discover new applications on their own. And that accumulated usage is what eventually generates the productivity improvements and service enhancements that justify the original investment.
Training, in other words, operates as both the foundation and the measuring instrument. By making AI education a firmwide priority, Baker McKenzie gathers a steady stream of information about which practice groups are engaging, which tools are being adopted, and where resistance remains strongest. Participation rates across offices and departments become leading indicators of AI value, visible months before financial returns work their way into matter economics.
That data has practical consequences. A tool that draws strong voluntary engagement across multiple practice groups is treated differently in the firm’s investment plans than one that requires constant pressure to keep utilization numbers acceptable. Likewise, a practice group showing weak training participation and little subsequent usage signals a need for more targeted support, better tool configuration, or a reassessment of whether the product fits that group’s workflow at all.
The Mechanisms Behind the Approach
For the strategy to work as a genuine gauge of value, several elements have to come together in a disciplined way. The first is scale. AI training at Baker McKenzie is not reserved for a small circle of innovation-minded partners or distributed haphazardly through internal newsletters. It is organized firmwide, reflecting the view that AI competence is no longer a specialty practice skill but a baseline professional capability.
The second element is relevance. Generic presentations about how large language models work do little to change behavior. Training carries weight when it is tied to realistic work scenarios — drafting memos, reviewing contracts, summarizing discovery, preparing due diligence reports across practice areas. Lawyers absorb the technology when they can see it acting on tasks they actually recognize from their daily work.
- Adoption measurement: Training completion is tracked alongside matter-level usage so leadership can see whether education translates into action rather than curiosity.
- Feedback loops: Training sessions generate structured feedback about tool performance, and that feedback feeds directly into procurement decisions and vendor roadmaps.
- Practice group granularity: The firm can observe variation across its global network, identifying where AI becomes part of the daily rhythm and where adoption stalls.
The third element is accountability. When training is framed as a core component of professional development, it carries expectation rather than mere suggestion. That does not require coercion. The more effective lever is building AI fluency into the normal expectations of practice at the firm, so that avoiding the tools becomes the path of greater resistance.
Why Training Participation Is a Leading Indicator, Not Just a Compliance Checkbox
Read carefully, training data offers a sharper picture of an organization’s relationship with AI than most financial dashboards do. It reveals not only who shows up but who comes back for advanced sessions, which tools sustain interest beyond the initial novelty, and which offices are moving faster than others. For a firm like Baker McKenzie, with a presence across dozens of jurisdictions and legal systems, that geographic variation is itself strategically valuable.
Training also surfaces the gaps that utilization metrics hide. A lawyer may use a generative AI tool frequently for simple administrative tasks but avoid it for complex work where results genuinely matter. Training assessments that distinguish between levels of competency can highlight whether lawyers are using AI for low-value convenience or for substantive professional work. The latter is where the real return lives.
How does Baker McKenzie use training to gauge AI investment value?
Baker McKenzie uses firmwide training as a leading indicator for AI investment value. Instead of relying only on financial projections, the firm tracks whether lawyers complete AI training and apply those tools in their daily work. High participation signals genuine adoption, and adoption is the precondition for measurable productivity gains and cost savings. The approach treats training not as a compliance procedure but as the first data point in an ongoing evaluation of whether an AI tool deserves continued investment.
No single metric tells the complete story, and Baker McKenzie’s approach is best understood as one element of a broader evaluation framework. Direct financial analysis still matters, as do matter-level time studies, client feedback, and quality review. But training occupies a particular position in the sequence because it is the earliest reliable signal. By the time financial results become visible, the strategic decisions about which tools to scale and which to retire have often already been made.
The Risks and Limits of an Adoption-Led ROI Strategy
An approach centered on training carries its own hazards, and any firm copying the playbook should recognize them. The most obvious is the gap between attendance and mastery. A lawyer can complete every session and never touch the tool again. Completion rates can be inflated by well-designed content, convenient scheduling, or even subtle pressure from practice group leaders. Without usage data to corroborate training metrics, the numbers can flatter an organization into believing adoption is healthier than it really is.
There is also the risk that training becomes a metric to be gamed. If participation is linked to performance reviews or compensation, lawyers may complete courses mechanically without changing their behavior. The discipline that makes this approach work is the insistence on linking education to application — checking not only whether lawyers attended, but whether the tools they learned about are showing up in their matters and their time entries.
Another limitation is-speed of technological change. The generative AI landscape is evolving quickly, and training content risks becoming outdated within months of being written. A training-driven evaluation model therefore requires the firm’s education function to move at product development speed, continuously updating curricula, rolling out new use cases, and retiring material that no longer reflects current tool capabilities. The firm whose training lags behind the market will end up gauging value based on yesterday’s technology.
Finally, there is a strategic risk embedded in any adoption-first philosophy. If the firm concentrates too heavily on whether lawyers are using the tools already available, it may underinvest in capabilities that are not yet ready for firmwide deployment. The most valuable AI investments may require patience — small trials, failures, and course corrections — before they produce something worth training thousands of lawyers on. A pure adoption framework needs to make room for experimentation alongside measurement.
What Other Firms Can Learn From Baker McKenzie’s Approach
Baker McKenzie’s strategy offers lessons that extend well beyond its own operations. The most important is the reframing of the core question. Investment committees have traditionally asked whether they bought the right tools. The training-led model shifts the emphasis to whether the organization has created the conditions for those tools to succeed. That shift has practical consequences across the firm’s entire technology operation.
Procurement decisions, for example, begin to factor in adoption risk as a central consideration. Product evaluations pay closer attention to ease of use, learning curve, and the quality of vendor-provided training materials. The firm’s knowledge management and professional development teams become active participants in technology selection rather than late-stage support functions. Training ceases to be a cost center that follows the investment decision and instead becomes central intelligence for the entire process.
For other firms searching for ways to measure AI value, the approach points to a series of concrete actions.
- Define adoption before you define ROI. Determine what successful use looks like for each tool, then build training and measurement around that definition.
- Connect training records to usage data. Attendance tells you about intention; usage tells you about behavior. Both are necessary.
- Segment the data. Adoption patterns differ across practice groups, offices, and seniority levels. Aggregated numbers will hide the most important variations.
- Use training feedback as procurement intelligence. Lawyers are candid about what works and what frustrates them. Capture that candor and route it into vendor evaluations.
- Treat training as an ongoing investment, not a launch event. The tools will change, and the training must change with them.
The broader legal market has been searching for credible ways to evaluate technology value ever since generative AI became one of the fastest-growing line items in firm budgets. Some firms have implemented AI-specific time tracking, asking lawyers to log minutes saved by automated assistance. Others have introduced matter-level comparisons between similar engagements, with and without AI support. Baker McKenzie’s training-centered method offers a complementary route: measure the moment where value begins, which is the moment a lawyer learns to use a tool with genuine skill.
That does not mean training is a perfect proxy for financial return, nor that it should replace traditional financial discipline. The strongest evaluation frameworks will combine education metrics with utilization analytics, matter economics, and client feedback. What Baker McKenzie has recognized is that the sequence matters. A law firm cannot measure what its lawyers do not know how to use, and it cannot capture the full value of an AI investment until the people doing the work have become fluent in the technology.
The real prize is a law firm where the ROI question begins to answer itself because the work has already changed. As AI tools grow more sophisticated and new models arrive each quarter, an organization’s capacity to learn and adapt becomes its most durable competitive advantage. Baker McKenzie’s training-led approach is, at bottom, a bet on that capacity — and a signal to the rest of the legal industry that the most important investment metric may not be the size of the AI budget, but the depth of the learning culture behind it.
