Manufacturing’s next AI phase is about integration, accountability and quality

The most interesting of these may be MCP, because it addresses one of industry’s oldest frustrations: fragmented enterprise data. Propel argues that manufacturers have spent years trying to reconcile engineering, quality, supply chain and commercial information sitting in disconnected systems, often through expensive custom integrations. MCP, by contrast, is presented as a simpler route, allowing AI to query business systems through prompts and act as a unified intelligence layer across functions. If that promise holds, manufacturers that adopt early could reduce one of the biggest structural barriers to agility.

Bank of Canada’s take

There is a useful Canadian parallel here. The Bank of Canada reported in June 2026 that AI adoption among Canadian firms remains at an early stage for production purposes, even though personal use of AI by business leaders is already widespread. In other words, interest is no longer the constraint; operational integration is. Statistics Canada has likewise found that Canadian firms with complementary capabilities, such as cloud computing, data analytics, robotics and ICT training, are significantly more likely to adopt AI successfully. That aligns closely with the MCP idea: AI delivers greater value when it can sit on top of connected digital foundations rather than operate as a stand-alone novelty.

The second trend identified by Propel is the evolution of AI from productivity tool to “co-engineer.” This is an important distinction. A chatbot that drafts content or summarizes notes is helpful, but a co-engineer is embedded in design review, quality assessment and iterative development. Propel’s view is that the surrounding context, memory, tooling and workflow scaffolding determine whether AI remains a conversational assistant or becomes something more akin to an engineering partner operating within defined human parameters. That does not remove accountability from people, but it does compress development cycles and expand the number of design options teams can test before committing.

Canadian industry appears to be moving in the same direction, albeit unevenly. Deloitte Canada’s 2026 AI report says organizations are shifting from simple productivity gains toward wider workforce and operating-model redesign, while KPMG Canada argues that the country’s new national AI strategy is intended to push firms from experimentation to enterprise-wide deployment. The federal government has also been explicit that manufacturing and robotics sit among the priority sectors for this next wave of adoption. The strategic implication is clear: the firms that treat AI merely as an office efficiency layer may fall behind those that embed it directly into engineering and operations.

Propel’s third trend—the expansion of the digital thread is less flashy, but arguably more impactful. The company argues that a product is no longer just hardware; it is now a blend of physical item, software layer, service model and sometimes subscription revenue. That means product data cannot remain a static engineering artefact. It needs to become a live business asset, linking development decisions to field performance, quality events and commercial outcomes in real time. This is the kind of shift that can materially affect time to market, product margin and the speed of post-market improvement.

Again, the Canadian comparison is telling. Statistics Canada has noted that AI adoption is associated with stronger productivity prospects when firms pair it with complementary digital capabilities, rather than treating it as an isolated software purchase. Government and advisory commentary around Canada’s “AI for All” strategy also frames AI as a lever for economy-wide productivity and competitiveness, not just task automation. For manufacturers, that means the digital thread is not simply an IT modernization project; it is increasingly part of the commercial architecture of the product itself.

AI accountability becomes a business requirement

The fourth trend is about AI accountability moving from best practice to business requirement. This is perhaps the most urgent. Propel’s argument is that the question is no longer whether companies use AI, but whether they can show who was responsible for each AI-influenced decision, how unexpected behaviour is detected, and how agent performance is monitored over time. This is a sensible reframing. “Human in the loop” can become a passive slogan; “human in command” demands traceability, auditable oversight and clearly owned judgment calls.

Canada’s policy direction reinforces this. The federal AI strategy places trust at the center of its agenda, while legal analysis has pointed out that Canada is pursuing a more incremental governance route after the demise of the proposed Artificial Intelligence and Data Act. That may leave some details to emerge through privacy, sectoral and other legislation, but it does not reduce the commercial need for accountability. If anything, it increases the onus on manufacturers to build governance into workflows now rather than wait for regulators to write every rule for them.

The final trend may end up being the most commercially powerful: quality as a revenue signal. Propel says AI-enhanced quality is shifting from early adoption to competitive necessity, especially when connected quality systems identify failures earlier, shorten feedback loops and strengthen customer relationships. This is a persuasive argument. In complex manufacturing, quality has too often been treated as a cost center or compliance obligation. Yet when quality intelligence is linked to design, service and field data, it starts to influence repeat sales, brand trust and lifecycle profitability.

For Canadian manufacturers, this should resonate. Much of the country’s AI-in-manufacturing discussion has focused on reliability, productivity and competitiveness, with predictive maintenance and smarter quality control often cited as the practical use cases most likely to generate near-term return. The broader point is that AI in manufacturing is maturing. The next phase will not be won by the companies with the most pilots, but by those that connect data, govern decisions, and turn quality and engineering intelligence into measurable business advantage.