Skip to content
Book with Me
AI Mid-Market Transformation

The Strategy Solves the Wrong Problem

Paul Korber
Paul Korber

What the AI for All strategy, released June 4, means for mid-market organizations that have already started and are not seeing results

93 percent of Canadian business leaders say their organizations are already using AI. Two percent are seeing a return on that investment.

Read that again. Not 2 percent of laggards. Not 2 percent of businesses still running on spreadsheets. Two percent of organizations that have already committed budget, assigned people, and started moving.

That is the actual state of AI in Canadian business right now. The problem is not getting started. The problem is getting results.

The federal government released AI for All, Canada's national AI strategy, on June 4. It is a serious document with genuine ambition behind it. The target of raising AI adoption from 8 percent of Canadian SMEs today to 60 percent of all businesses by 2034 is an honest acknowledgment of how far behind the curve most Canadian organizations actually are.

Read AI for All carefully, though, and a pattern emerges. Four of the six pillars are fundamentally about awareness, literacy, training, and workforce development. The operational commitments lean heavily toward education programs, online tools, and individual capability building.

For a micro-business that has never touched a digital tool, that is the right medicine.

For a mid-market organization with 50 to 500 people, a management team that has already deployed AI tools, and a board asking why the investments are not showing up in the numbers, AI for All has almost nothing to say.

To understand why, it helps to define the three stages most organizations are actually navigating.

Deployment, Adoption, and Execution: Three Different Problems

These three terms get used interchangeably in most AI conversations. They should not. They describe distinct stages with distinct failure modes, and the interventions that help at each stage are completely different.

Deployment is the act of standing something up. You purchased Copilot licenses. You configured a tool. You ran a pilot. You pointed AI at a use case. Deployment is technical and transactional. It is measurable as a binary. Are you using it or not. The Statistics Canada figure of 12 percent measures meaningful deployment. The KPMG figure of 93 percent measures deployment at any level, including a single employee using ChatGPT once a week. Both numbers are accurate. They are measuring different thresholds of the same thing.

Adoption is the organizational behaviour change that follows. Are people actually using it consistently. Is it embedded in how work gets done. Is the organization finding new applications because the first ones worked. Adoption is cultural and operational. It compounds when it works and stalls when the deployment was not grounded in a real business problem, or when the change was not managed. Adoption is not a technology problem. It is a leadership and capability problem.

Execution is whether the investment changed a business outcome. Did deployment lead to adoption. Did adoption translate into measurable results. Revenue protected or grown. Cost reduced. Time recovered. Risk managed. Execution requires a baseline, a target, an owner, and a result. It is not binary. It sits on a spectrum from pilot to partial integration to embedded in core operations with demonstrable return.

AI for All is a deployment intervention. The problem most mid-market organizations face is an adoption and execution problem. Those require different solutions.

Where Most Mid-Market Organizations Actually Are

There is a more precise way to map this. Based on patterns across mid-market technology engagements, AI adoption follows five recognizable stages.

Stage 01: Curiosity — Employees explore AI independently. No strategy. Shadow AI begins.

Stage 02: Systematic Deployment — Tools deployed at scale. Copilot, Claude, and similar platforms. Top-down rollout.

Stage 03: The Wall — Tools plateau. Data gaps surface. Processes break. Foundation exposed.

Stage 04: Integrated Applications — Purpose-built AI in specific workflows. Data cleaned up. ROI measurable.

Stage 05: AI-Driven Transformation — Business model changes. New revenue streams. New competitive position.

Stage 3, The Wall, is where the 93 percent live.

The tools are deployed. The initial enthusiasm has passed. The pilots produced mixed results. The data problems that nobody anticipated are now blocking progress. The processes that AI was supposed to improve turn out to be broken in ways that predate the technology. The governance questions that were deferred are now creating risk. The workforce is uncertain and undertrained. Leadership is losing patience.

The Wall is not a technology failure. It is an organizational readiness failure that deployment exposed. And it is precisely where AI for All stops being useful.

Stage 3 cannot be solved with a training program or an awareness-level diagnostic designed for organizations that have not yet deployed. It requires someone who can identify which part of your specific foundation is exposed and sequence the work to get through it.

The Research Confirms It

Multiple credible Canadian sources published in the last eight months tell a consistent story. The numbers below are not projections. They describe the current state of organizations your size.

93% of Canadian business leaders report using AI in some form, up from 61% the prior year. Only 2% say they are seeing a return on their generative AI investments. (KPMG Canada Generative AI Adoption Index, November 2025)

Only 31% have fully integrated generative AI across core operations and workflows. 20% are still experimenting or piloting. (KPMG Canada, November 2025)

83% of Canadian employees say they want or need to upskill on AI tools. Fewer than half feel their organizations provide adequate support. (KPMG Canada, November 2025)

Among medium-sized firms specifically, skills shortages and integration challenges are the primary barriers, ahead of cost. (Sage Canada Digital and AI Imperative, 2025)

Canadian SMEs could unlock nearly $350 billion in economic growth if more firms reached the digital and AI maturity of the top-performing 8 percent. (BDC Digital Transformation and AI Study, June 2026)

Canada ranks 44th of 47 countries on AI training and literacy, and 42nd of 47 on trust in AI systems. (KPMG University of Melbourne Global AI Trust Study, cited in AI for All)

The pattern across all of this is the same. Organizations your size have deployed. They are investing. They are not getting results. The bottleneck is not awareness or permission. It is the organizational capability to move from deployment through adoption to execution in a way that is specific to their business, their data, and their Canadian regulatory context.

What AI for All Actually Is

To be fair to the document, it was never going to solve an execution problem. Federal policy operates at the ecosystem level. It can fund programs, set standards, create incentives, and build infrastructure. It cannot tell a $75 million logistics company in Mississauga how to prepare its data, which vendor to trust, how to govern AI responsibly under the Consumer Privacy Protection Act, or how to get its operations team to actually change how they work.

That work happens inside organizations. It requires practitioners, not policy documents.

Three specific gaps in AI for All are worth naming directly.

The IP commercialization gap remains structurally unsolved. Canada's own consultation named IP flight as a top concern. AI for All points to ElevateIP as the mechanism. ElevateIP is a program, not a policy framework requiring IP retention as a condition of public investment. Vector, Amii, and Mila represent genuine world-class investment in AI research talent. Nearly 70 percent of Canadian-led startups still end up headquartered outside the country. The mechanism that would change that structural drain is still absent.

Policy stability remains unaddressed. SR&ED has been modified repeatedly. Capital gains treatment has shifted more than once in recent years. A leadership team making a multi-year AI investment decision cannot model their Canadian incentive position with confidence today. AI for All adds programs. It does not add certainty.

The funding landscape is real but largely inaccessible at scale. AI for All names the BDC LIFT program, the Regional AI Initiative, the Compute Access Fund, the Canadian Tech Growth Fund, SR&ED, the Productivity Super-Deduction, ElevateIP, IP Assist, the AI Missions Program, Skills for Success, Mitacs ADOPT, and more. Each has its own eligibility criteria, application window, delivery agency, and approval process. For a mid-market organization where the leadership team is already at capacity, navigating that funding landscape is itself a full-time project. Build your AI investment case on business fundamentals. If a program fits, treat it as a bonus, not a plan.

What This Means for Your Seat at the Table

The practical implications differ by role. Here is the honest read.

CEO

The risk calculation has shifted, but not in the direction AI for All implies. The real risk for your organization is not failing to start. Most of your peers have already started. The risk is continuing to invest in AI activity without a path to results. The question your board should be asking is not whether you are using AI. It is whether you have the organizational capability to move through The Wall and get to measurable outcomes, and what it would actually take.

The BDC research published this month puts the commercial scale of the gap in terms worth taking to your next strategy session. Canadian SMEs could unlock nearly $350 billion in economic growth if more firms reached the maturity of today's top performers. The difference between a top performer and the average is not the tools they bought. It is how deliberately they moved from deployment through adoption to execution.

CIO / VP Technology

The KPMG data gives you specific language for conversations that may have stalled internally. 83 percent of your employees already want to upskill on AI. Fewer than half feel the organization is giving them adequate support. That is not a technology gap. It is a leadership and investment gap that sits above your function but that you can now frame precisely.

The sovereign infrastructure argument in AI for All also gives you new leverage. If your AI workloads are running on US-owned cloud infrastructure and your organization handles regulated data, that is a board-level governance exposure the federal government has now explicitly named. Use it.

CFO

Two things deserve attention before the next budget cycle. First, the incentive environment is unstable enough that AI investment decisions should be modeled on business fundamentals, not assumptions about government support. The programs exist, but accessing them takes time and resources most teams do not have spare.

Second, the 2 percent ROI figure is your most important planning input right now. It tells you that most organizations are spending on AI without a return path. The CFO question worth pressure-testing is whether your current AI investments have a defined business outcome attached to them, a measurable baseline to improve against, and an owner accountable for the result. If the answer to any of those is no, the investment is activity, not strategy.

Chief Risk Officer / VP Risk

AI for All gives the risk function new standing on two fronts. Data residency is the first. If AI workloads are processing sensitive customer or financial data on foreign infrastructure, the strategy has explicitly named that as a national sovereignty concern. The organizational exposure is not theoretical.

The second is governance maturity. The KPMG research found that organizations moving AI into core operations without corresponding investment in people, process, and governance are creating compounding risk. The Consumer Privacy Protection Act is moving toward implementation. The time to establish governance architecture is before a production AI system creates an incident, not after.

Two Things Worth Doing Before AI for All Matures

The programs in AI for All are not fully operational. The incentive mechanisms are still being designed. Waiting for clarity before moving is understandable. These two actions cost nothing, and their value does not depend on anything Ottawa does next.

Name the specific business problem before the technology conversation starts. The research is consistent on this point. Organizations that deployed AI without first defining a specific, measurable business problem are the ones that hit The Wall hardest. AI for All's own data says 78 percent of non-adopting Canadian firms cannot see how AI applies to their business. That is a translation failure, not a technology failure. The question worth asking is concrete: where is your business losing time, margin, or competitive position that better information or faster decisions would fix. Write it down in a language your CFO recognizes as a cost, and your CIO recognizes as a system. That conversation takes 90 minutes with the right people in the room. Most organizations running stalled pilots have never had it.

Assess your data before you assess your tools. Every credible research source on AI adoption barriers names data readiness as a foundational constraint. The G7 SME AI Blueprint, prepared by Canada's own national AI institutes, Amii, Mila, and Vector, identified it explicitly: many organizations lack data that is AI-ready or face significant challenges preparing it for use. For your organization, the question is specific. For the business problem you just named, do you have clean, accessible, trustworthy data that describes it well enough to improve on? Most organizations discover at this point that the answer is more complicated than expected. Knowing that now is free. Discovering it six months into a pilot is not.

The Bottom Line

The pattern is consistent. An organization deploys AI across three or four use cases. One delivers partial results. The others stall. When you trace the stalls back, they almost always lead to the same foundational gaps: data that was never clean enough, governance that was never defined, or a workforce that was never brought along.

But the deepest gap is rarely any of those. It is that the process AI was applied to was already broken before the technology arrived. Not broken in an obvious way. Broken in the way that years of workarounds, undocumented exceptions, and institutional memory become invisible until something tries to systematize them.

Organizations that treat AI as an opportunity to rethink how work actually gets done reach Stage 4. Organizations that use AI to automate what they already do, without asking whether they should be doing it that way at all, hit The Wall and stay there.

AI for All solves a real problem for organizations that have not yet started. Canada needs those organizations moving. That matters. But the strategy has almost nothing to say to the majority of mid-market organizations already stuck at The Wall. The execution gap is not a policy problem. It is an organizational capability problem that requires someone who can diagnose exactly which part of the foundation is exposed, sequence the work to get through it, and build the internal capability to sustain results on the other side.

I am currently working with a select number of Canadian mid-market organizations on a structured diagnostic built specifically for Stage 3. It examines seven dimensions of organizational readiness, identifies precisely which Wall failure mode is blocking progress, and maps a clear path to Stage 4. A facilitated workshop format is available now for leadership teams who want to move quickly. A self-serve online version is in development.

When you hit The Wall, call Paul.

Reach out directly at paul@nextambitions.ca or visit nextambitions.ca

Sources

Share this post