(This column originally appeared in Forbes)
Over the past few weeks there have been a number of news reports about the escalating costs of AI implementations, particularly for big brands.
Canva recently cut its forecast after an AI costs “blow out.” Website developer Wix saw its earnings and valuation negatively impacted partly because of “high AI related expenses.” Uber has capped its AI spending due to high internal use.
“In some cases people are seeing the cost of tokens exceed the cost of the employee within a month or two of use, just because they’re using it too much,” analyst Jack Gold of J.Gold Associates said in a separate report.
SMBs Not Impacted By Tokens…Yet
The good news is that smaller businesses aren’t having this issue. Unfortunately, that’s not going to last.
Most small businesses aren’t really using AI the way big companies are. For sure, SMBs are leaning into platforms like Claude and ChatGPT to help them write better emails, perform research and offer advice. They’re using (somewhat) the capabilities of Copilot and Gemini in Microsoft Office and Google Workspace to increase their productivity with presentations, spreadsheets and file management. But most are not developing AI solutions for their businesses.
SMB AI Projects Are On The Rise
However, some of my clients are starting to do this. They’re connecting their AI platforms like Claude and ChatGPT to their accounting, CRM, HR, logistics, order management and other systems. They are hiring developers and using technology firms (like mine) to build models trained to read and interpret their company’s call transcripts, emails, files, manuals, documents and other technical materials.
Then they’re building AI-based applications to generate weekly “executive summaries” or produce quotes based on conversations or portals where both customer service representatives and actual customers can search for answers to their problems using natural language prompts. The cost of these projects is not astronomical. But, depending on the size of your business, they’re not insignificant.
For example, to do most of the above you’ll have to subscribe to an AI platform which is generally $20-$30 per month per user and then pay for implementation and programming services from a developer, connections such as Model Context Protocol (a popular connector between data and an AI platform) and ongoing maintenance.
For a small business with 25 users you can expect a first-year cost to be about $6,000 for the subscription, about $10,000-$20,000 for the implementation and another $1,000-$2,000 for the MCP connector. After that you can expect to spend a few thousand dollars a year for support because things break and people will always need help.
It’s not a trivial cost. But a small business owner can justify the return-on-investment by quantifying productivity gains and customer retention or increased sales and if those profits are substantial enough the effort may be worth it.
SMB’s AI Projects Will Soon Use More Agents
But this is just prompting and reporting and assistance. It’s still not “doing.” Other than the largest of companies, most businesses aren’t really using AI to actually perform actions on their data, like run e-mail campaigns, converse with customers, reconcile accounts, create service tickets or generate invoices.
Those activities require agents and agents will require using Application Programming Interfaces (APIs) and when that happens tokens are required. And it’s the consumption of tokens that’s driving up the costs for big companies. But really, how many tokens will an SMB use to really have an impact on cost? No where near as many. But get ready, because the costs for SMBs are still going to increase. A lot.
Many experts believe that tokenization and subscription costs are going to rise substantially in the next few years — even by as much as five to ten times. Why? Take OpenAI. Ahead of their planned IPO, the AI leader is already draining vast amounts of cash in compute costs and spend, all to build infrastructure, advance their product and gain market share. This cannot keep up forever.
Stacia Garr recently wrote in the Harvard Business Review that the corporate AI era is on the verge of a massive shift.
She argues that organizations are “effectively being paid to adopt AI.”
“This customer-acquisition playbook created a false sense of budgetary and operational security for executive buyers, characterized by “unmetered,” “complimentary,” or “included” premium AI agent features,”” she says.
Garr and other experts predict that these subsidies are going to disappear — in fact they already are. She says that the “the era of artificial pricing is coming to an end.” The big companies — OpenAI, Anthropic and others — that are gearing up for IPOs are facing reality: Wall Street.
We’ve all been to this rodeo before. It’s the freemium model. Netflix, once the cheap provider of online content, has doubled its monthly fees over the past decade. Uber generated huge losses while spending hundreds of millions to gain market share for its services, and now has significantly increased its fees over the past few years. Other online services — eBay, Etsy, Booking.com, Expedia, etc. — have hiked their “take rates” as well.
So what does this mean for SMBs? It means that it’s going to become even more costly to implement, and maintain, an AI project in the not-too-distant future. Will it be 5 times more? 10 times? That’s up for debate. Regardless, the age of affordable AI for smaller firms is about to end.
