AI Won't Fix a Broken Business. Here's the Test.

Why are more than half of CEOs getting no return from AI?

Because they're using AI to paper over problems that exist in the business itself. Canva's former CFO Damien Singh, who was part of the leadership team that took Canva from $10 million to $2 billion in annual revenue, said it plainly in a recent Forbes interview: the model you picked is not your problem. Your operations, your clarity, your fundamentals are.

That framing should land hard for any SMB owner who has spent the last two years buying tools hoping something sticks.

What does "broken business" actually mean in this context?

It doesn't mean the business is failing. It means there are foundational issues that AI will not resolve and will often make worse by moving faster.

Think about what AI actually does: it accelerates output, automates repetitive work, and surfaces patterns in data. If your sales process has unclear handoffs, automation makes the chaos happen faster. If your customer data is a mess, AI-generated insights will be wrong at scale. If your team doesn't know who owns what decision, giving them an AI assistant just gives them a faster way to go in circles.

Singh's lens as a finance executive is useful here. CFOs are trained to ask: where is value actually being created, and where is it leaking? Most AI implementations skip that question entirely and jump straight to the tooling.

What does the data say about AI ROI failures?

The numbers are consistent across multiple surveys. More than half of CEOs report no measurable AI return, according to the Forbes reporting on Singh's perspective. This tracks with what we see working directly with SMB clients: the failure mode is almost never "we picked the wrong LLM." It's almost always one of three things:

• No documented process before automation. You can't automate a workflow that lives in someone's head.
• No owner for the AI output. If nobody is accountable for what the AI produces, quality degrades fast and nobody notices until there's a problem.
• No baseline to measure against. If you didn't track how long something took or what it cost before AI, you have no way to prove (or disprove) that anything improved.

Canva at $10 million was not the same company as Canva at $2 billion, but the discipline around measurement and process ownership scaled with the business. That's the part most SMBs skip.

How do you diagnose your business before investing in more AI tools?

Before adding any new AI capability, run this four-question diagnostic:

1. Can you describe the process in writing, step by step, without asking anyone? If the answer is no, the process isn't ready for automation. Document it first. This takes days, not weeks, if you actually sit down and do it.

2. Who owns the output? Every AI-assisted workflow needs a named human who is responsible for the result. Not a team. A person. If you can't name them, you're not ready.

3. What does good look like, and how will you measure it? Set a baseline before you change anything. Time to complete, error rate, cost per unit, customer response time. Pick one number that matters and write it down.

4. Is this process core to how you make money, or is it peripheral? Start AI implementations in areas that are high-volume and repeatable but not the thing your business lives or dies on. Build confidence and process discipline there before you touch anything critical.

"AI doesn't transform a business. Discipline transforms a business. AI accelerates whatever discipline you already have."

What did Canva get right that most SMBs get wrong?

The short answer is sequencing. Canva built operational rigor first, then layered in tools. Most SMBs buy the tools and hope the rigor follows. It doesn't.

At the $10 million stage, the decisions Singh's team made about financial visibility, process ownership, and measurement discipline were what made the $2 billion outcome possible. The tools they used along the way were in service of a clear operating model, not a substitute for one.

For an SMB at $2 million or $5 million or $20 million in revenue, this means the same thing: get your house in order before you automate it. A documented, accountable, measurable process that runs at 70% efficiency will outperform an AI-automated mess every time, and it will actually improve when you do add AI on top of it.

Does this mean SMBs should slow down on AI adoption?

No. It means they should be strategic about where they start.

There are categories of AI application that work well even in businesses with loose processes: first-draft content generation, meeting summarization, research assistance, basic data analysis. These are low-stakes, high-volume tasks where AI produces a useful rough output that a human reviews and refines. The cost of a bad AI output is low, and the time savings are real.

The applications that go wrong are the ones where AI is asked to make or automate consequential decisions in systems that were already unclear. Customer communication workflows, pricing logic, lead scoring, financial reporting. These require clean inputs and accountable ownership before AI touches them.

The framework is simple: use AI to do more of the work you already do well before you use it to fix the work you do poorly.

What we'd actually do

• Run the four-question diagnostic on every process you're considering automating. If you can't pass all four, fix the process first. This is a half-day exercise, not a consulting engagement.
• Pick one high-volume, low-stakes workflow and build a complete AI-assisted process around it, with a named owner and a before/after metric. Get one clean win documented before expanding.
• If you want a structured way to work through AI strategy for your specific business, the AI For Business community at skool.com/aiforbusiness is where SMB operators are doing exactly this work together.