In 2016, I started asking whether our business would still be relevant in three or four years if artificial intelligence kept improving. We had built a successful company around professional language services for law firms, financial institutions and other regulated organizations, and it was becoming clear that technology would eventually handle a meaningful share of the work our team handled.
Ignoring that possibility felt riskier than investing in it, so we began building AI into what we offer our clients.
At the time, I assumed the hardest part would be technical. If we built technology that was good enough, organizations would want to use it. Ten years on, that is one of the assumptions I would revise. The technology matters a great deal, but getting professionals to trust it, use it properly and fit it into the way they already work has proven every bit as difficult as building it.
The other surprise has been the effect on our need for people. Becoming a more AI-centred company has increased our reliance on professional expertise in several important areas. The human side of our business is growing, and we continue to hire and develop professionals who bring subject-matter knowledge and judgment to work that increasingly involves AI.
I think that experience holds a particular lesson for legal professionals. Whether AI can perform work that lawyers once did themselves is largely settled, since in many cases it already can. The more useful question for the profession is what becomes more valuable for the lawyer to do as AI takes on a larger share of the work.
The human role is larger than review
“Human in the loop” appears so often in discussions of AI that the phrase has started to lose its meaning. It can suggest that the AI does the real work while a person waits at the end to catch mistakes, which I think undersells what the professional contributes.
Inside our company, we have come to describe the change in the language profession as a shift from producing the work to orchestrating it, and I see a useful parallel for lawyers. A conductor does not play every instrument, yet the performance depends on someone who knows how it should sound and keeps each section working toward it. A lawyer working with AI may no longer produce every first draft or perform every repetitive step by hand, while still setting the objective, weighing the risks and deciding whether the work is ready to put before a client, court or regulator. Those responsibilities are broader than the ones they replace.
If a lawyer’s whole contribution shrinks to checking hundreds of pages of machine-generated work line by line, the firm has probably automated the wrong part of the workflow. A better objective is to use technology to cut repetitive production and concentrate professional attention where judgment adds the most value.
Expertise becomes part of the system
Specialized AI improves as experienced professionals correct its output, flag exceptions and set out how a situation should be handled in future, and the quality of that input matters far more than its volume. A small set of curated, firm-specific documents will usually teach a system more than a large archive of raw material, which can degrade the output instead of improving it. Deciding what goes in is itself a matter of professional judgment, and only people who know the practice can make those choices well.
Our experience with new linguists offers a useful comparison. When we bring a translator onto a client account, they arrive fluent and still spend months learning that client’s glossary, its preferred phrasing and the terms it wants handled a particular way. The strongest translators go further and learn how the client’s departments work together and where its business strategy is heading, so that what they produce in the target language reads as though it had been written in that language from the start. A capable AI system needs similar context, instruction and correction before a firm can expect consistent results, and the professionals who provide it are helping to define what good work looks like for that firm.
Start with what is at stake
When firms acquire powerful technology, there is a temptation to ask what else can be automated. I would begin by asking what happens if a given task goes wrong, and who would notice. Some errors surface immediately and cost little to fix, while others sit undetected in a signed agreement or a regulatory filing until someone relies on them.
Answering that question tells a firm more about where AI belongs than any list of tasks the technology can perform, and how carefully a firm answers it says a good deal about its AI maturity.
Know what to check and when
For lawyers, one of the most useful skills to build is managing the work that passes between professional and machine.
Before the AI starts, someone has to settle what context and sources it should rely on. Afterward, the lawyer needs to know where review should concentrate and how corrections get captured so the same problem does not keep coming back. A conductor who knows how a piece should sound can pick out the one instrument playing slightly off-key, and in the same way, a lawyer with a clear picture of what the finished document must achieve is far better placed to notice the clause that no longer fits with the rest.
Invented case citations get most of the attention, yet the errors I worry about more are small drifts in language, such as “commercially reasonable efforts” becoming “best efforts,” which read fluently and change what a party has agreed to do. Catching those means reading for what might have shifted in a document that looks complete, and that habit has to be taught. It asks a good deal more of a lawyer than typing a prompt into a chatbot and reading the answer.
Protect the way lawyers develop judgment
Much of a lawyer’s judgment is built by doing the work, including drafting, being corrected and watching how senior colleagues approach a problem. If AI takes over some of those tasks, firms will need to think deliberately about how the next generation gets the repetitions that once built expertise.
Nobody should preserve inefficient work just because juniors have always done it, though efficiency and professional development remain separate goals. AI can become part of training itself, giving juniors more time to question outputs and understand why something is right or wrong, provided firms design for it. One practical approach is to have juniors build the skeleton of a memo or argument themselves, deciding its structure and where it needs to go, and then use AI to flesh it out under their direction. An architect works in much the same way, settling the structure of a building and how people will move through it before anyone chooses the finishes.
Without that attention, a firm could become very efficient at producing work while thinning the pipeline of people able to judge whether that work is any good.
Why we are still investing in people
When we started investing seriously in AI, I worried that technology might eventually make our existing business irrelevant. That concern was partly justified, since technology has changed our industry enormously and will keep doing so.
The experience has also changed how I think about the relationship between AI and professionals, with AI increasingly handling production while people supply the context, judgment and accountability, including the decision about when to trust the technology. The professional is the one who understands the larger picture and the outcome the client needs, decides how each part of the work is refined to reach it and ultimately carries the liability if it falls short.
After years of building AI, we are still investing in human expertise, and I expect most law firms will reach the same conclusion. Their lawyers may personally produce less of each first draft while carrying more responsibility for directing the work and answering for the result, which makes human professional judgment the core of the job.
Gary Kalaci is the founder and CEO of Apertera (formerly Alexa Translations), which provides AI-enabled language solutions to legal, financial and regulated organizations. A lawyer by training, he began his career as a translator and court interpreter. Since founding the company in 2002, he has built it into a provider serving most of Canada’s major national law firms, all of its major banks and leading securities regulators.
