AI: Endless possibilities, finite budgets

Matthew Day

Staff

September 7, 2026

WISCONSIN – Though AI continues to change the workplace, businesses are still figuring out how to pay for it. 

In June, OpenAI CEO Sam Altman acknowledged a growing challenge for the AI industry: the cost of using AI has become, “all of a sudden, a huge issue.” 

Amid the excitement of the tech’s efficiencies, Altman said a chorus of concerns has arisen, as companies share examples of AI costs causing them to burn through an annual budget allotment in a single quarter. 

But how has this national sticker shock impacted Wisconsin businesses? 

Based on insights from a number of Northeast, Central and West Central Wisconsin experts – plus two Minnesota-based professionals with Wisconsin clientele/territory – such extreme budgetary issues have yet to drastically impact the region’s business community. 

However, whether their organizations are implementing, experimenting with or practically reliant upon AI, these experts agree the evolution of the technology largely outpaces the budgeting practices designed to accommodate it.

There is no definitive playbook for navigating AI – but these professionals are building one as they go, and have offered their work-in-progress wisdom: 

• Tracy Brewer: marketing and events coordinator, Centergy, Inc., Wausau 
• Ann Cappellari: chief medical information officer/system vice president, SSM Health, Fond du Lac 
• Justin Held: founder, Rocket Construction and Property Minds, Eau Claire 
• Andy Miner: senior director of AI, corporate and commercial banking, Associated Bank, Greater Minneapolis-St. Paul 
• Dakota Nyberg: owner and creative producer, DMM Productions, Greater Minneapolis-St. Paul 
• Nikki Peroutka: president, Element, De Pere 
• Amy Seehafer: COO, Orthopedic & Sports Medicine Specialists (OSMS), Green Bay 
• Amanda Van Den Elzen: founder and facilitator, BetterWork Training, Seymour 
• David Zey: owner and CEO of Zeytech, Inc., VP of applied AI systems at Juno, Green Bay 

These nine voices represent a range of AI usage and budgets, with some organizations increasing spending aggressively while others have yet to create an AI budget at all, relying on free options or capabilities bundled into software they already purchase. 

As the experts share, the question is no longer simply what AI costs, but whether employees use it, whether it improves the business and whether companies have budgeted for what it takes to make it work.

Across health care, marketing, banking, creative services, construction and technology, the experiences of these experts point to a common theme: AI is becoming harder to ignore, but its economic impact remains a moving target. 

The AI budget isn’t always an AI budget 

For some businesses, AI has arrived without appearing as a new line item. 

Brewer said her organization, Centergy (Central Wisconsin Alliance for Economic Development), Inc., hasn’t needed to allocate money specifically to AI, because it can accomplish much of what it needs with free tools. 

“We have not done anything for AI regarding budget,” she said. “We use free options.” 

Nyberg said he has taken a similar approach at his video production company, DMM, where he uses AI capabilities built into software it was already purchasing, including Photoshop, Canva and Final Cut Pro. 

“We don’t have an AI budget because AI is already embedded in tools we’re paying for,” he said. 

Nyberg said that means AI can produce a return without necessarily producing a corresponding increase in technology spending. 

Should DMM get to the point where direct investment in AI is imperative, he said he expects the money would likely go toward “business development, CRM or lead generation.” 

Held, meanwhile, said he’s had a similar situation for his commercial construction company – Rocket Construction – as well as his real estate venture, Property Minds.  

Neither business, he said, has needed to devote a particular budget category to AI.  

Both, Held said, have primarily utilized ChatGPT and Grok for enhancing written communications and social media, as well as for gathering data.  

“It’s more used just to make things more professional and make things go faster, really,” he said. 

The efficiency extends into marketing, Held said, where AI can easily turn footage from construction sites into more content, help edit videos and identify clips worth publishing.  

The result, he said, is, essentially, greater output from the same underlying resources. 

“AI in marketing pretty much allows us to double our output,” he said. 

More deliberate spending 

At the other end of the spectrum are organizations that view AI as a strategic technology investment and expect their spending to grow. 

Cappellari said she anticipates spending will increase for SSM, as the technology has become “too valuable to ignore.” 

These increases, she said, will be highly researched and targeted. 

Cappellari said she expects organizations like SSM to increasingly separate useful AI applications from the hype surrounding them, concentrating resources on the technologies that have demonstrated value. 

“We’ll double down and expand on the most valued and proven use cases,” she said. 

Seehafer likewise said she expects technology spending to increase at OSMS, particularly on platforms with AI capabilities built into their core architecture.  

OSMS’ investments in legacy systems, she said, are currently being shifted to AI-native alternatives, including evaluating a new electronic health record system. 

“We will undoubtedly increase our technology spending on platforms with AI native capability,” she said. 

Seehafer said the motivation isn’t AI for its own sake – rather, she sees the investment as part of sustaining growth while automating repetitive, manual work to improve experiences for patients and employees alike. 

“Recently, we implemented a solution to automate our faxes – yes, we still receive far too many fax communications in health care – rather than having humans manually sort and route them,” she said. 

Seehafer said OSMS is also considering AI voice automation for inbound and outbound calls, although she stressed that technology cannot come at the expense of patients. 

At integrated marketing agency Element, Peroutka said she also expects continued investment. 

In fact, she said Element is currently developing proprietary AI tools intended to improve client service and provide faster access to information. 

However, Peroutka said AI spending is part of a larger evolution of the business, rather than a replacement for its existing marketing investment. 

“The opportunity isn’t to spend less on marketing,” she said. “It’s to make every marketing dollar work harder.” 

Difficult predictions 

Each leader interviewed shared a baseline agreement: AI is valuable. 

The challenge, they agreed, lies in determining what that value – and the cost of achieving it – will look like six months or a year from now. 

Cappellari said AI’s consumption-based pricing makes budgeting particularly difficult. 

“All AI has a token cost or a consumption model,” she said. “The more you use, the more you’re charged.” 

Usage can also grow unexpectedly, Cappellari said, as employees discover new applications for a technology. 

“This starts as a minimal cost intervention, but then grows sometimes very rapidly over time,” she said. 

There is a countervailing benefit, however, as Cappellari said the cost of generative AI has been falling on the supplier side, helping keep expenses under control. 

Van Den Elzen – who, with BetterWork, works with mid- to enterprise-size organizations on AI adoption and value realization – said she sees budgeting troubles across a broad range of companies. 

Today’s AI-pricing environment, she said, is “ambiguous and disjointed,” with costs ranging from bundled capabilities and monthly licenses to token-based fees. 

Van Den Elzen said that makes forecasting particularly difficult for organizations whose employees don’t know how much computational work a task will require. 

Still, she said pricing is only half the problem – the other half is return on investment. 

“Most organizations do not have a clear tether between the cost of their tools and the actual return they’re providing,” she said. 

Tracking “hours saved” KPIs, Van Den Elzen said, is an easy starting point, but doesn’t necessarily demonstrate bottom-line financial impact. 

“In my experience, this comes from applying AI tools to parts of work that don’t actually save or make money,” she said. “Summarizing documents, writing emails, generating images, etc., are great starting points to learn how the technologies work, but they don’t actually make movement on strategic initiatives or organizational goals aimed at real return.” 

Van Den Elzen said her approach is to connect AI use directly to measurable business goals. 

She said she was able to teach the employees of one BetterWork client – a sales organization whose objective was to double regional market share – not only the basics of Copilot, but also how to use it for tasks such as planning sales routes, anticipating objections and identifying promising accounts. 

The distinction, Van Den Elzen said, is important: AI use is not necessarily the same thing as AI value. 

The hidden budget: Getting people to use AI 

Van Den Elzen said she sees another cost businesses frequently overlook: AI adoption. 

“Organizations are budgeting for the tool,” she said, “but not for the work of getting people to actually use it and use it well.” 

Though buying licenses is relatively straightforward, Van Den Elzen said changing how employees perform their jobs is not. 

Too often, she said she’s seen organizations roll out premium AI licenses but neglect to provide adequate training or follow-up.  

Van Den Elzen said one employee she interviewed didn’t even know he had access to the paid tool, yet the company was still paying for the license every month. 

In that company’s case, she said she calculated “there was at least $27,000 in licensing fees that weren’t being utilized at all.” 

Her recommendation, Van Den Elzen said, is straightforward: “Treat adoption as its own line item, not an assumed byproduct of the purchase.” 

Broadening what constitutes an AI budget, she said the expense considerations should go beyond mere software. 

“The software is the obvious cost, but the work of making sure it’s actually used – and used on the right things – is where the real effort needs to go,” she said. “It’s the piece almost no one is currently accounting for.” 

Using more than one AI system 

Another emerging source of spending, Van Den Elzen said, is organizations purchasing multiple AI platforms because different systems excel at different tasks. 

She said she increasingly sees businesses using combinations of Copilot, Gemini, Claude, ChatGPT, Perplexity and other systems rather than standardizing on one provider. 

The rationale, Van Den Elzen said, is that the tools aren’t interchangeable – some integrate particularly well with workplace applications, while others may be stronger at research, coding, image generation, strategic work or autonomous task execution. 

Fiscally, she said this can amount to companies paying for multiple subscriptions for the same employee. 

Van Den Elzen said a new budgetary question is thereby posed: Is paying for several AI systems wasteful duplication, or is it the cost of giving employees the right tool(s) for different jobs? 

Zey said his experience illustrates why some organizations may be willing to accept the expense. 

Affording advanced AI use 

Zey – who works across technology and leadership environments at Juno and Zeytech, and previously worked with the Microsoft AI Co-Innovation Lab in Milwaukee – said his use of AI is much more intensive than the occasional chatbot query. 

For example, he said he recently used AI agents and Claude to complete what he estimated would have been roughly four weeks’ worth of work – including building a 15-week executive roadmap, creating/revising presentations, developing a customer beta program and preparing for and conducting eight job interviews – in a single week. 

The financial calculation of such productivity, Zey said, can be compelling. 

For one of his organizations, he said he estimates AI spending at roughly $250 per person per month at the base level, with additional credits potentially adding hundreds of dollars.  

Another of his teams, Zey said, operates at a lower spending level because its workload is different. 

He said for a start-up operating at a fast pace, however, he considers premium AI spending worthwhile. 

“If, [for example], for $1,000 they can get four weeks of work from me in one week – totally worth it,” he said. 

The magnitudes of capacity the tech can facilitate, Zey said, therefore deepen questions of budgeting philosophies beyond simply asking whether an AI subscription is expensive. 

AI costs primed to increase? 

As companies grapple with AI budgets, Zey said a further complication looms: rising costs for AI access. 

He said he believes current AI prices do not reflect the technology’s eventual economics, and a drastic adjustment is imminent. 

“I think it’s going to be 10x what it costs now,” he said. 

Today’s prices are being suppressed, Zey said, by intense competition and investment among AI companies. 

He said anticipation of this increase, with only an educated guess as to when it will take place, makes long-term budgeting difficult. 

Zey said cost estimates are all the trickier for heavy users, who often run up against usage limits and must purchase additional credits, thus creating a variable-cost component traditional software subscriptions don’t have. 

This present time of relative affordability, he said, offers a prime opportunity for companies to test out the tech. 

“While the prices are low, take advantage of maximizing your capability and experimentation now,” he said. 

Equating AI to labor capacity 

Zey said he also expects businesses to become much more rigorous about calculating AI ROI in terms of individuals’ productivity.  

“Depending on who the engineer is, it could be 50% to five-times,” he said. 

Zey said his current approach to AI budgeting is to treat it as “more of an R&D cost at the moment.” 

As applications mature, however, he said he expects some of those experiments to become operational budgets – particularly when companies build AI-powered applications that require IT support, maintenance and ongoing computing costs. 

That distinction, Zey said, mirrors what the other regional respondents are seeing: Experimentation can begin cheaply, but successful applications eventually become part of the business infrastructure. 

Calculating better decisions 

Miner said for Associated Bank, the mindset has been to measure AI by the quality of the result rather than the number of minutes saved. 

At Associated, he said AI is helping employees research clients, understand industries, prepare for conversations and improve credit decisions. 

“The most meaningful contribution has been to the quality of our bankers’ decisions, not simply how quickly they work,” he said. 

Miner said AI is enabling the bank to bring some research capabilities in-house that it previously purchased from outside providers. 

“The goal there is not cheaper research – it is research we can stand behind, and that our bankers can rely on,” he said. 

Under Associated’s philosophy, Miner said a technology investment can be worthwhile even if it doesn’t directly reduce an existing expense, provided it improves the quality of a consequential business decision. 

When it comes to ROI and efficacy, he said the bank is “careful not to overclaim.” 

“We measure against our own baselines, because a claim you cannot stand behind is not worth making,” he said. 

Human judgment remains part of equation 

Despite the enthusiasm surrounding AI, none of the experts described a future in which the technology solves budgetary conundrums by replacing paid employees. 

For example, Cappellari said AI documentation still struggles with nuances such as sarcasm and nonverbal communication. 

“It can’t ‘see’ an eye roll, a raised eyebrow of mirth or a deadpan exaggeration,” she said. 

In creative work, Nyberg said AI can accelerate pre-production, scripting, editing and other routine tasks – but it cannot reproduce the wisdom, curiosity or discernment of a creative team. 

“Those experiences and perspectives are hard to replicate with AI,” he said. 

Similarly, for marketing, Peroutka said AI can produce content quickly, but it comes up short in originality, among other qualities. 

“AI struggles with context, nuance, strategic judgment and the human insights that drive successful marketing,” she said. 

Held said he’s reached the same conclusion: “You need a human aspect.” 

Human accountability, Miner said, is especially important in a regulated industry such as banking. 

“We treat AI output as a draft to be validated, not as an answer to be trusted blindly,” he said. 

The effort expended to prompt AI to more closely mimic humanity, Brewer said, can sometimes consume more time than it saves. 

“It seems, at times, we can spend time we don’t have trying to use AI,” she said, “and it may be best to use our own brain power.” 

Consensus: Spend strategically, not automatically 

These nine perspectives reveal a spectrum of AI adoption, with organizations at different stages of an AI spending curve.  

However, a common thread suggests AI is increasingly being judged as an investment rather than simply a technology purchase, bearing greater significance than any run-of-the-mill budget line item. 

Businesses are asking whether AI saves labor, increases output, opens new capabilities and/or improves quality, decision-making and customer and staff experiences.  

With a combination of technology, training, experimentation and business transformation, formulating AI budgets will undoubtedly be a unique process for any given company. 

But the more a company comes to rely on AI, the more willing it will be to foot the bill at scale. 

Take, for example, one of Zey’s managers, whom he said is “very operationally focused.” 

“We talk about this on a regular basis…,” he said. “She literally says, ‘I have nightmares about every other week that we don’t have AI anymore.’”

For businesses, then, the question may no longer be whether AI belongs in the budget, but how much they are willing to invest in it.