AI makes it easier to build, but that makes choosing harder
AI is rapidly removing one of the old constraints on business transformation: the difficulty of turning an idea into something that works. Building can now happen much faster, which means organisations have far more opportunities to change how they operate.
But greater capacity to build does not tell us what is worth building.
“Don’t build just because you can,” says Viet Pham. “The question isn’t whether you can build it anymore – it’s what you actually need to build.”
As execution capacity increases, organisations need to become more deliberate about where they use it, what value they expect and how much change their people can absorb.
The underdog may have the advantage
The instinct is understandable: find a high-performing part of the business, add AI and make it even more efficient. Viet, however, thinks there is another, potentially more valuable place to look.
“My view is that you should look at the places where you’re not as strong, where you want to compete and grow.
There is a commercial argument behind this, but also a very human one. Viet recently saw an organisation develop an AI agent for a highly experienced specialist team. Work that had previously taken weeks could suddenly be completed in around 30 minutes, but the people in the team were understandably reluctant to embrace it. They had spent years building their expertise and were concerned about what the technology might mean for their roles.
In another part of the organisation, the situation was almost the reverse. The team was operating in an area where stronger competitors had a clear advantage, and it wanted to grow. Viet sees that kind of setting as potentially fertile ground for AI: people have something they actively want to become better at, rather than an established way of working they feel they need to defend.
This does not mean strong teams should be left untouched. Resistance may simply signal that change management needs more attention. But when organisations are choosing where to make their first serious AI investments, Viet believes ambition and appetite for change should be part of the equation alongside technical feasibility and efficiency.
That gives leaders a different way to approach AI prioritisation. Instead of only asking where AI can make us more efficient, we can ask where becoming radically better could genuinely change our competitive position.
Don’t teach AI your old ways of working
Choosing the right opportunity is only the beginning. Viet sees another problem when organisations take an existing process and simply insert AI into it.
Most processes were designed around humans and the systems available to them. Once agents become part of the work, the same steps, handovers and responsibilities may no longer make sense. Viet recommends starting agents with a limited area of responsibility, monitoring how they perform and correcting them where necessary. As performance becomes more reliable and trust grows, organisations can gradually give them greater autonomy.
“The break happens when organisations try to optimise AI for human work instead of redesigning the work itself.”
His rule of thumb for deciding where humans should remain involved is simple: consider how difficult the outcome would be to undo. The more consequential or irreversible the decision, the stronger the case for human judgement.
Take tendering. AI can help an individual write an RFP or RFQ response faster, which is useful. But the bigger opportunity is to look at the process end to end: how information is found and checked, how the response is created, where human expertise matters and what an agent could handle independently.
This changes the question from “How can AI help someone perform this task faster?” to “How should this work happen now that humans and agents can do it together?”
It also gives us a clearer way to measure the results. Rather than trying to calculate productivity gains scattered across hundreds of individual employees, we can look at outcomes the business already understands, such as how quickly we respond to customers and whether we win more tenders.
Personal productivity still matters, but measuring its wider business impact is difficult.
“Personal productivity gains are much harder to translate – you might have happier employees, but that alone doesn’t show up on the bottom line,” Viet says.
Saving time can make work easier and give people more space for things that require their expertise. The challenge is understanding how those distributed benefits relate to the business case for AI investment.
Welcome to the AI Wild West
This becomes more urgent because AI capability is already spreading quickly inside companies.
Viet describes the current situation in many organisations as a “wild west”. Some employees use AI as a better search engine or writing assistant, while others connect it to platforms and services. At the leading edge, people may already be building their own tools for creating documents, financial models and other business outputs.
“Different people use it differently,” Viet says. Some are still using an extraordinarily fast car for little more than trips to the supermarket, while others are already building increasingly sophisticated capabilities of their own.
That experimentation can be valuable, but it is also beginning to create a new form of shadow IT. People are already building tools that can produce documents, financial models and other business outputs without oversight.
The challenge is to mature the organisation without killing the curiosity that got people experimenting in the first place. As individual tools develop into agents capable of carrying out meaningful work, organisations need a way to bring them into the business deliberately.
Viet calls this “onboarding agents”.
How do you onboard a digital colleague?
Most operating models assume that people work with other people. We have established ways to define roles, responsibilities, access and accountability.
Those assumptions start to shift when agents become active participants in end-to-end processes.
For Viet, onboarding an agent should not be completely different from onboarding a person. When someone joins an organisation, we define their role, decide what they can access, clarify what they are responsible for and establish where they can act independently. Digital colleagues need the same kind of clarity.
“If we were talking about a human, we would decide the role, access, autonomy and accountability. We should do the same thing for agents.”
As agents become more capable, those decisions will matter increasingly. What is this digital colleague responsible for? What can it decide independently? When should it involve a person, and who remains accountable for the outcome?
These questions will become more important as AI continues to reduce the effort required to turn ideas into working solutions. Technology may allow organisations to move faster, but people still have a finite capacity to absorb change, and businesses still have to decide which changes are worth making.
When building was difficult, some choices were made for us. Now we increasingly have to make them ourselves.
There will be no shortage of things AI allows us to build. The advantage will come from knowing which ones are worth building.
Want to explore what happens when execution is no longer the bottleneck?
As AI makes it easier to turn ideas into action, organisations need to rethink how they connect strategy, investment and execution. Explore what this shift means for business transformation in our white paper, Execution is no longer the bottleneck.
Who benefits from AI: your company or your service partner?
AI solved one problem, but it exposed another
