And other lessons from recent Canadian AI M&A

Within 90 days of acquiring System Dynamics Corp., Toronto’s Beacon Software had a six-figure AI project with a Caribbean utility. For anyone trying to understand where enterprise AI revenue is coming from, and will come from, the sequence is more telling than industry commentary, tech prognostication, or a survey asking executives whether they plan to adopt it. Deals talk.

Getting to this point had taken different kinds of cycles. Relationship cycles. System Dynamics had served the utility for 40 years. When its new owner wanted to know where AI might be useful, someone could pick up the phone and ask the CFO. 

That conversation helps explain what CPP Investments bought on November 5 when it invested US$50 million in Beacon. The Toronto company acquires businesses that make software for particular industries, from recreation to utilities, then puts a shared team of engineers to work on problems their customers already have. The investment brought Beacon’s Series C to a reported US$425 million at a US$1.75 billion valuation. Across its 45 companies, Beacon has something a new AI vendor could spend years trying to assemble: people on both sides of a sale who already know how to work together. 

What the CFO wanted was help reconciling bank transactions. According to Beacon, the resulting project covered operations in Jamaica, Trinidad and Guyana. The utility bought a multiyear subscription, with part of the price tied to outcomes. The relationship got Beacon the sales opportunity. The contract gave the utility a way to hold its supplier to the result.

Canada has a long history with the first half of this strategy. Constellation Software built a business acquiring specialized software companies and giving their operators considerable independence. Beacon is betting that a central engineering team can learn something at one company that makes the next customer’s problem cheaper to solve. How much of the work travels? Bank reconciliation may be familiar across utilities. Accounting systems, approval rules and the people trusted to change them can differ at every one. 

Ada + Dialpad : what the deal tells us

The proposed merger between Toronto’s Ada and Dialpad, announced November 4 and expected to close around November 12, approaches the same issue from the customer’s side. Ada builds AI agents for customer service. Dialpad operates the communications network and software that businesses use to handle calls and messages to including its VOIP roots. As he was founding Ada, Mike Murchison spent a year working across seven customer service teams. He saw people who wanted to help and often understood what a customer needed, but had to wait for an approval or track down a colleague in another department. The person doing the coordinating was frequently the customer. 

Murchison uses a missing delivery to explain the merger. The customer needs the item tomorrow. An excellent answer about where the original parcel went is of limited value. Someone has to locate a replacement, authorize the extra cost and arrange delivery. Ada and Dialpad believe they can connect that work across the call, the software and the person authorized to make an exception. This is more useful if the problem is detected in advance, a function pattern-based AI excels at. The question is no longer “does the replacement arrive, and does the customer have to call twice?” It’s “is the issue detected in advance, and addressed upstream on the supply chain.” Customers keep their account teams, pricing and support channels. Even a company selling a new way to manage support understands the value of being able to call the same person on Monday. 

Toronto chipmaker Taalas gives the hardware side of this story a familiar cast. Its three founders had worked at AMD and later helped lead Tenstorrent before starting Taalas to build chips that hardwire an AI model into silicon. On August 6, AMD agreed to buy the company, putting the price at approximately US$2.9 billion in cash and stock, including up to US$1.4 billion in contingent payments. AMD expects the deal to close by year-end. Where Beacon’s utility sale grew out of a customer relationship, Taalas shows how professional relationships can come full circle. Engineers who once worked for AMD developed an approach outside the company that AMD now wants to offer alongside its own Instinct GPUs, a huge win for the Canadian ecosystem with a short window.

Canadian marine technology company Kraken Robotics is another datapoint. On November 5 it disclosed $102 million in new orders booked since its August update. More than $55 million came through the Covelya businesses Kraken acquired in July. The total also included subsea batteries and sonar systems, with orders across the combined company coming from both returning and new customers. Acquiring Covelya expanded what Kraken could supply. It also brought in businesses that defence and offshore energy buyers already knew well enough to place orders with. How much additional business those relationships generate for the combined company is a question for subsequent results. The combination of AI, and expertise and relationships, equally weighted,is the key to real utility and progress, expansion and growth.

This distinction exists in Beacon’s numbers, too. It’s a holding company, conducting a set of AI rollups. A campground software company, Let’s Camp, acquired in summer 2024, has increased annual recurring revenue by 62%, according to Beacon. A founder who had been handling sales, product and customer service himself gained colleagues, payment tools and the capacity to take on work he previously had to pass up. This is the combination of domain expertise, offloading of repeatable functions to a trained system, and reorienting of founder time to where it is likely most useful: relationships.

The revealing question for Beacon is what happens with System Dynamics at the second utility. Can the engineers reuse enough of the first project to make the next one more profitable? A relationship built over 40 years helped them discover what the need was,and how AI could be deployed to address it. The lesson from several of these stories: technology leads, but expertise builds, and relationships close.