Getting from AI capability to meaningful outcomes

At SWITCH in Singapore last week, the dominant narrative I heard was AI adoption progressing through three waves: capability, scaling and refinement. The framework makes sense. Organisations build capability, extend its use and improve how it performs. The difficult part is making that sequence happen inside a business.

The meetings and discussions I had resonated with the challenge of moving from capability into scale. They didn’t go into the specifics. That may reflect uncertainty about the transition path, commercial sensitivities, or both. I came away wanting to spend more time on that transition, because it’s where having access to powerful tools has to become an ability to deliver meaningful outcomes.

An organisation can have people using AI confidently and still struggle to turn that activity into a dependable way of working. The question is how to connect what individuals can now do with what the organisation needs to achieve.

Define the outcome you want to repeat

A successful demo shows that something is possible. To make it useful operationally, the organisation needs to agree what good looks like when the same work arrives again, under normal conditions, with the information and people available at the time.

Let’s imagine a customer service team using AI to draft customer responses. The drafts may be clear and quick to produce. Resolving the customer’s issue could still depend on checking an order in another system, interpreting a policy and getting approval for an exception. Until those parts work together, faster drafting may leave the overall service largely unchanged.

That gives the team something specific to investigate. It can measure how long a request takes to resolve, how often information needs correcting and where work waits for a decision. Counting generated responses would miss much of what matters to the customer.

One of my more encouraging takeaways from SWITCH was the traction around being “AI bilingual”. Singapore has adopted this at government level: combining expertise in a field with the ability to use AI effectively. In July, IMDA and the Institute of Singapore Chartered Accountants launched AIxAccountancy under the National AI Impact Programme. Singapore government address.

From what I experienced, the approach is doing useful work connecting operators for AI collaboration. That matters because people who know the work can explain the exceptions, recognise an unreliable result and identify the decisions that still need professional judgement. Practical AI fluency helps them participate in shaping how the work changes. A big thank you to Samantha Su for taking the time to run through the initiative in detail with their collaboration partners Mocha Chai Laboratories & Red Airship.

The next step is to bring that understanding into the same conversation as the people responsible for systems, process and delivery. Together, they can work out whether a problem needs a different use of AI, a better connection to existing information or a change in how decisions are made. Training alone won’t settle those questions, but AI bilingual people can help make them much more concrete.

Where capability meets the work

The useful test is whether an improvement holds up through the whole workflow. If a faster first step leaves another team doing more checking or repair work, there is still something to resolve. Following one outcome through to completion can make that visible.

This is the thinking behind our 4D framework. The amount of investigation should fit the uncertainty. Sometimes a short conversation and a practical test are enough; a more complex workflow may need closer examination.

The waves framework gives us a way to talk about progress. The practical work is in understanding what an organisation needs to change between those waves. Leadership can help by making ownership clear and giving people room to question an assumption when the evidence doesn’t support it.

If your organisation has built capability but meaningful outcomes still feel difficult to repeat, that’s a useful place to start a conversation. Where does the work stall between what AI makes possible and what the business can reliably deliver? Getting specific about that gap gives everyone something practical to work on.