Time for Business to Make AI Pay Its Way

After several years of unvarnished excitement and seemingly unquestioned investment in generative AI, technology leaders are being forced to reckon with the value of these costly innovations – especially as organisations look to scale and maintain AI projects.

That’s the gap Apptio is trying to close. Apptio, an IBM company, made its name by providing chief information officers, chief financial officers and technology leaders with a clear view of where the tech pound goes and the value it ultimately delivers to the organisation, so tech budgets become investments that support business priorities.

In practice, that means everyone speaking the same language in terms of cost, consumption and value. It means making choices based on data, not guesswork or estimations.

AI is where this gets real. Implementation is expensive, workloads are prone to spike and the costs are extremely dynamic. One month you’re training a model, the next you’re paying to scale it across business or products with first-of-kind consumption patterns.

New resources and measures appear – token cost, graphics processing unit (GPU) cost and many others – but without full transparency you’re flying blind. Teams need to compare capital expenditure, operating expenses, software as a service and resource costs against each other to understand how much it costs to run and how to achieve value.

The extreme expense of these challenges has made FinOps more essential than ever, giving teams the opportunity to drive strategic decisions for AI investments. A view into the holistic cost and value of AI best positions these teams to track usage, set limits, size resources properly and link spend to results before it costs them.

Companies still want to make big bets on AI, but want to know exactly how they’ll pay off

The need for clarity

Greg Holmes, field chief technical officer for EMEA at IBM Apptio, puts it simply: “AI feels easy until you get the bill. What leaders need is a clear view from cost to value – not a spreadsheet full of assumptions. Our job is to make the numbers so clear that business and technology can decide together what to scale.”

The backdrop doesn’t help. AI has sent data-centre demand through the roof, with costs in key markets rising particularly fast and every forecast pointing to continued heavy capital spend ahead. Translation: rising costs, rising scrutiny. Companies still want to make big bets, but they also want to know exactly how these bets are paying off.

Apptio’s proposition sits squarely in that space. Apptio TBM solutions help break down financial, labour and usage data silos and aggregate them into one dashboard, showing the full AI bill for computing, storage, data pipelines, on-prem kit, people and services.

From there, it’s about doing the work that ensures AI pays its way. That includes deciding what to build, what to buy and which ideas earn their keep – and proving to the board how the cost curve shifts as adoption grows. A full story, from first provision to ongoing return on investment.

The same job can end up with three different price tags depending on who you ask

“Agentic AI is only as smart as the data you feed it,” Holmes says. “If your cost data is distributed across five software systems, your ‘assistant’ will give five different answers.

“Pull it into one financial source of truth and suddenly you can ask sensible questions – say, ‘What would it cost us with AI to get the desired business outcome?’ – and get a sensible answer.”

That sounds like common sense, but it often fails in practice. Tags go missing, costs go “unallocated” and the same job gets three different price tags depending on who you ask. Each team has a piece of the puzzle, but the organisation lacks a complete picture of how AI budgets are being spent – or what value those investments are delivering.

A finance team will see the burn but not the return. And engineers use the largest model available but may not fully understand the cost implications.

Apptio’s view is simple: connect the cost story from end to end. Map what you have, bring discipline to cloud and AI spend, make priorities transparent. Then let your teams work off the same truth. When the entire technology cost story is being told, organisational alignment and smarter spending are possible.

Otherwise, fragmented views will result in redundant tools, inefficient spend and less return on investment. In the AI era, that inefficiency can result in leaders wasting valuable resources on low ROI projects. Transparency enables success for the projects that matter most. IBM, for example, has been weaving that visibility into its wider optimisation stack since the acquisition.

Smarter spending

So what does “doing it properly” mean if you’re a UK tech leader looking at next year’s numbers? First, consider cost management as an essential part of successful AI innovation. Missteps now can hinder future development – short-changing your AI transformation before it has fully begun.

Second, measure where it matters. Knowing the token costs or GPU hour only helps if you can translate it into the things your board understands – customers, revenue and growth.

Third, get your priorities out in the open. When everyone is chasing their own methodology, the question isn’t “Can we afford generative AI?” but “Which projects actually pay their way?” An informed portfolio view turns that debate into a business decision instead of a turf war.

Finally, close the loop. If an initiative isn’t delivering measurable value, reallocate the spend. The lowest-cost AI process is the one you don’t run. IBM Cloudability helps to manage the FinOps layer; IBM Apptio helps to connect cost and value across the IT portfolio.

Measure what matters, fund what pays back and stop polishing prototypes that don’t

Holmes’s parting shot is both blunt and sensible: “Measure what matters, fund what pays back and stop polishing prototypes that don’t show a return on investment. Tie every GPU hour and token to the units the business understands, review the numbers and kill the underperformers. That’s how AI moves from science project to profit centre.”

The truth is not glamorous but it is real: training models is resource intensive and time consuming; production piles on usage fees; data has to be gathered and prepared; licences multiply. And people across the organisation need the tools to manage these processes.

To be truly competitive in the AI era, organisations must shift their mindset around AI from tech innovation to business investment: understanding how costs contribute to value, scaling projects that prove their worth, and investing in the technologies that help unlock AI’s full potential.

Breaking data silos across organisational, operational and technological systems of record is fundamental to achieving this level of transparency and efficiency. With the right data, technology and finance teams can make smarter investment decisions that balance innovation, cost and value for a true competitive advantage.