Todd Bernson is the Chief AI Officer at BSC Analytics, where he leads AI strategy and cloud transformation for enterprise clients.
An enterprise customer hesitated over an AI investment because its leaders didn’t know where the technology would be in 12 months. I understand the concern. I don’t know either.
But their organization will still need accessible data, workable integrations, governance and people who know how to use AI. Those capabilities take time to build. Waiting for a clearer forecast only delays the work that lets them act.
Enterprises can afford to wait on a specific AI model. They cannot afford to wait on building the capabilities or platform to use it. That work needs to start now, and the faster AI changes, the more it costs to wait. A better model next year is worth nothing to a business that still cannot put it to work.
We Have Seen This Before
Cloud adoption played out the same way. Some financial services organizations I worked with moved slowly on cloud and modern application architectures. Now they want AI capabilities while still carrying monolithic applications and manual processes that make integration difficult.
Meanwhile, a startup building its workflows today can design AI into them from the beginning. Established enterprises have valuable data and deep expertise, but getting those assets into a usable workflow takes work. Postponing that work only makes it more expensive.
Cloud taught us that adopting a foundation and adopting its next new capability are different jobs. Migration alone settled very little. Plenty of companies moved badly and rebuilt the same problems on more expensive infrastructure. The ones that benefited had modernized how they built, integrated and operated technology. Companies that waited still had all of that work ahead of them.
AI is exposing the same distinction, only faster. Enterprises have a second chance to learn that lesson. They shouldn’t need a third. Getting it right takes three things:
1. Build What Won’t Become Obsolete
Some investments keep their value no matter which model wins: accessible data, reusable integration patterns, governance teams can work within and people who know how to use AI. Put early effort there.
The hardest of these is organizational. Getting the necessary teams committed to integration and use is where progress stalls. A license does not magically make departments cooperate.
Silos already make enterprise work frustrating and inefficient. AI can help connect information across them, but only if teams agree on access, responsibilities and how the work will change. The business owner must bring in security, technology and the people doing the work from the start.
DORA’s 2025 research (email required) on AI-assisted software development describes AI as amplifying an organization’s existing strengths and weaknesses. Organizational preparation is part of the AI investment. Budget for it that way.
2. Invest Where Value Is Measurable
I’m impatient with waiting, but I’m just as impatient with using AI simply because it can do something. “It can” is a demonstration. It is not a business case.
In an AI development workshop, we had to address a basic gap: Enthusiasm for the tools had outpaced agreement on the starting cost, the desired outcome and how improvement would be measured. Without those answers, no one can prove a return.
Start with the business problem that has the greatest potential impact. That might mean reducing risk, lowering operating costs or retaining customers. Choose the outcome that matters most to the business, then work backward to the role AI can play. Apparently, AI is the one investment category where we occasionally forget to demand a business case.
Measure the current process before changing it. Set a target and a time window. Name the person responsible for certifying the result, and agree up front on what evidence would justify expanding, changing or stopping the effort. Include integration, human review and ongoing operating costs in the calculation. If the benefit is added capacity, explain what that capacity will accomplish.
3. Learn Faster Than The Technology Changes
The technology will keep changing. The advantage goes to the organization that learns fastest from each deployment. Build the foundations needed for a specific use case while putting that use case to work. If your AI strategy requires completing every modernization initiative first, congratulations. You have just created a five-year plan to do nothing. Keep technology choices open to revision, and use measured results to decide what happens next.
What concerns me is the gap that can open between enterprises with access to the same technology. One has already worked through integration and governance, brought teams together and learned where AI produces business value. The other is still debating whether to begin. When the next major capability arrives, those companies will have very different abilities to use it.
Waiting postpones both the potential returns available today and the experience needed to capture tomorrow’s. The enterprise that starts now can use each deployment to improve its next decision. A later purchase may get you the latest software. It will not come with a year of organizational learning conveniently preinstalled. Every deployment should leave the organization with something reusable: an integration pattern, a governance decision, a measurement framework or an operational lesson the next team doesn’t have to learn again.
Twelve months from now, the tools will have changed. Your organization needs to have changed with them. No leader can promise certainty about AI’s future. Every leader can make sure the next year is spent getting better at using it.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?