Baker, AI, and Why Employment Lawyers Aren't Going Anywhere

The Ontario Court of Appeal's recent decision in Baker v. Van Dolder's Home Team Inc. has unquestionably become one of the most discussed employment law decisions of 2026. In Baker and its companion decision involving Li v. Wayfair Canada ULC, the Court rejected an overly technical approach to interpreting termination provisions and confirmed that employment contracts should generally be read as a whole, in light of the parties' objective intentions, rather than by focusing on isolated words or hypothetical interpretations.

The reaction from the employment bar has been fascinating.

I have seen a number of posts from management-side counsel celebrating what they view as a welcome return to common sense. From their perspective, Baker represents a corrective to a line of cases that many believed had become increasingly unpredictable, where minor drafting issues or isolated phrases could invalidate otherwise carefully drafted termination provisions. That perspective is understandable and has been echoed widely in commentary since the decision was released.

At the same time, I have seen thoughtful criticism from plaintiff-side counsel. They point out how this decision appears to contradict many earlier decisions, and point out that looking to the parties' intentions may create new avenues for dispute. If courts are looking beyond isolated wording and considering broader contractual context, some may argue that provisions once thought immune from challenge could become vulnerable to new forms of attack.

Of course, not every analysis fits neatly into one camp or the other. Some commentators, including my brilliant colleague George Florea, have discussed possible ways to reconcile Baker with prior Court of Appeal authorities such as Waksdale and related termination clause jurisprudence. Those reconciliation arguments deserve serious consideration, but a detailed examination of them is beyond the scope of this post and will be the subject of a future article.

What has struck me most, however, is not the decision itself. Rather, it is the reaction to it.

In the days following the release of Baker, I have observed a sense of victory for defence counsel and some plaintiff's counsel that are less optimistic with respect to future claims/disputes and how they may be treated by the Courts.

I remain convinced that the field of employment law will remain as active as before, and that these decisions simply represent a continued swinging of the pendulum. Interestingly, one of the reasons for my optimistic view is something that itself causes concern for many: AI.

I have encountered a surprising degree of confusion from both employees and employers arising from the use of AI. Some employee-side litigants have sent lengthy lists of authorities generated after a brief AI search, claiming that the cases support their position. Yet, upon actually reading those decisions, the authorities often stand for the opposite proposition and would, in fact, undermine the very claim being advanced.

The same phenomenon appears on the employer side. Recently, I spoke with a business owner who confidently asserted that Baker would assist their defence. Unfortunately, the decision had little or no relevance to the issues actually in dispute. Their contract raised concerns that more closely reflected the breaches that invalidated contracts in Waksdale and Wigdor. They had input their contract and claim into ChatGPT, which spat out Baker as a possible defence (not understanding that it does not apply - AI still does a very poor job of contractual interpretation), and concluded that it somehow solves their problem without understanding the legal issues to which the decision actually applies.

This is not a criticism of those individuals. Employment law is nuanced. The problem is that many people are increasingly obtaining legal information without the legal training necessary to evaluate it.

The situation reminds me of a conversation I had approximately fifteen years ago with a wills and estates lawyer. At the time, software platforms were beginning to emerge that promised to automate wills and estate planning. Many lawyers were concerned that technology would drastically reduce the need for legal services.

I asked him whether he was worried. His answer has stayed with me.

He explained that very little of his time was actually spent drafting straightforward wills. A simple software tool could probably handle a large percentage of routine situations. The real value of legal counsel arose from unusual facts, overlooked issues, family dynamics, unexpected tax consequences, and the countless forms of complexity that arise when real life refuses to fit neatly into a template.

In his view, the the 3% of the issues that the software failed to address effectively would lead to an increase rather than decrease in litigation.

As it turned out, he was right.

Technology made routine drafting easier. It did not eliminate legal disputes. If anything, new disputes arose when individuals relied on automated tools without appreciating their limitations. The problems were simply deferred until a later stage when they became more expensive and difficult to correct.

I am increasingly seeing the same pattern with artificial intelligence.

Almost every week, I speak with employees and employers who have attempted to solve legal problems through AI-generated research. Often, it becomes apparent within minutes that key facts were omitted, the wrong question was asked, or the answer relies on assumptions that are unsupported by the law. The individual has no reason to know this. Without legal training, it is difficult to distinguish a persuasive-sounding answer from a legally correct one. I can usually tell within seconds what relevant information they either failed to input or how they may have unintentionally misled the AI with their prompts.

A recent file illustrates the point well:

An employer chose not to retain legal counsel and attempted to respond directly to allegations we had advanced. The company believed it was successfully defending itself. Instead, its response contained admissions and evidence that appear to support additional breaches of the Employment Standards Act. What the employer viewed as a defence may ultimately become evidence relied upon in support of broader remedies, including punitive damage arguments. That outcome was entirely avoidable.

The reality is that neither Baker nor the rapid advancement of AI changes the fundamental value proposition of legal counsel.

Courts may continue to refine their approach to contractual interpretation. The law governing termination clauses may continue to evolve. New technologies will undoubtedly make legal information more accessible than ever before.

What will not change is the importance of understanding the facts, identifying the issues that actually matter, recognizing the arguments that have not yet been considered, and applying legal principles to unique circumstances. Those tasks require judgment and judgment remains difficult to automate.

As employment lawyers, we spend very little time dealing with textbook fact patterns. We deal with unusual facts, unintended consequences, conflicting evidence, procedural complexities, business realities, and human behaviour. Those are the circumstances in which legal advice adds value, and those are precisely the circumstances where AI often struggles most.

For that reason, despite the excitement surrounding both Baker and artificial intelligence, I remain confident that there is no substitute for experienced legal counsel. In many respects, as disputes become more complex and information becomes more accessible but less reliable, experienced counsel may be more important than ever.

impact of AI on employment law