Where AI should play: how to pick the bets that move the business

Companies don't have an AI ideas problem. They have a focus problem.

The question isn't where we could use AI, but where to play with it and how to pick the few initiatives that actually move the business.

The investment of time, resources and money in adopting and implementing AI keeps growing. And yet few organizations are seeing any kind of return.

Why? Almost always because of a disconnect between the pilot or initiative being developed and what the business actually needs.

In general, this happens when the AI initiative portfolio is built on:

• Enthusiasm: the most persuasive leader's initiative wins, or the one tied to the trending topic, not the one with the most impact.
• Ease: whatever can start tomorrow wins, which is usually automating something that already works well.
• Accumulation or lack of clarity: many initiatives competing for the same budget and the same people. None gets finished.

When an organization has a dozen initiatives running in parallel, it shows up as several demos, prototypes or tests that create the false sense that a lot is happening. But the outcome, later, shows those initiatives fail when they try to scale and generate real business value.

How is this usually solved?

1. Choosing fewer bets, but more ambitious ones: instead of 50 small experiments, 2 or 3 important business processes.
2. Redesigning the work: defining upfront where and how that initiative will contribute to the company's future, and what it takes to get there.
3. Measuring impact on business objectives: if we understand where an initiative lands, we can quantify what and how much impact we expect (shorter cycle times, customer experience, and so on).

That's why the earlier, more strategic stage matters most: defining where AI is going to play. This is the framework we use at Kintara with leadership teams to decide it, and why the rest has to die in the room.

AI isn't applied in a vacuum. Every worthwhile initiative is tied to a key organizational priority: it has to accelerate it, unblock it or change its scale. The question isn't where we can use AI, but what long-term objective we could reach sooner, or at a scale we don't have today, thanks to AI.

For that, it helps to separate three things:

• Business bet: a strategic priority of the company. It's an input, already predefined.
• Play: a place where AI could accelerate or unblock a business bet. It's the only thing you evaluate.
• AI big bet: the final candidate, the play that already passed the filter and the review, with its own sponsor and budget.

If a play isn't tied to a business bet, it doesn't get in. However good it sounds, however convinced the person proposing it is.

Next, it helps to name the type of play. AI is so associated with automation that almost every initiative ends up trying to make processes more efficient. Naming four categories forces you to think beyond that:

• Capability: something the company can't do today, or does poorly.
• Experience: changes how the customer, external or internal, perceives or relates to the company.
• Decision support: improves the speed or quality of a human decision. The most underrated.
• Automation: executes or compresses a process that already exists.

And the order matters. Automation goes last on purpose, because it's the default answer. Putting it at the end forces you to look at the other three first.

Here's the trap. The easiest play is usually an automation, and it isn't a big bet: it's a quick win. Clean data, clear process, low risk, it takes the highest score in the room. Confusing a quick win with a big bet is how an AI portfolio loses its edge. Both go into the plan, but on separate lists. 15 items where no one distinguishes what changes the business from what saves two hours a week isn't a portfolio, it's a backlog.

• A quick win optimizes something that already exists and is already measured. It runs at the line level, no committee. Do it, but don't call it strategy.
• A big bet changes what the company can offer, or how it competes. It needs a C-level sponsor and its own budget. It's the only thing the committee decides.

Once chosen, big bets are ordered by how ready the organization is, not by how much they're worth. The wave doesn't signal importance, it signals how much preparation is missing before starting.

• Wave 1, now: the organization is already ready. It generates the evidence and infrastructure the next ones lean on.
• Wave 2, short term: a scoped, named piece of work is missing before starting. Weeks, not quarters. It builds on what Wave 1 left behind.
• Wave 3, long term: it's worth a lot and requires building a base. It arrives supported by the previous two, with a review date.

Prioritization is done with intention: Wave 1 feeds Wave 2, and Wave 2 feeds Wave 3. Progressive scaling, not three loose lists.

And here's what no framework solves. No play arrives neutral to the table: it's proposed by the head of an area, for their area, with their area's information. It's not intent, it's bias. We all believe our area's initiative will be the one with the most impact. And no one dares tell that Head their favorite initiative is a quick win.

Without clarity on where to play, you fall into the pilot trap: you use AI everywhere except where it moves the needle. Pilots that look good on paper and change nothing. When the team knows where, why and what's expected, commitment and clarity show up. And only then can you really measure progress and impact.

The method isn't the hard part. The hard part is discarding what's uncomfortable. A portfolio of 3 or 4 initiatives gets executed. One of 10 dilutes, and the next year no one can explain what happened.

In your organization, how many of your AI initiatives are tied to a written business priority?