(This column originally appeared in Forbes)
Corporate America is beginning to buckle under the costs of AI. Many companies, from Uber to Starbucks to Microsoft to Coinbase and Walmart are either ditching some of their AI projects or reigning back their AI spending as it becomes increasingly difficult to justify the ROI of these efforts when compared to what the rising costs of subscriptions and tokens.
“If you’re not actually able to draw a direct line to how [many] useful features and functionality you’re shipping to your users, that trade becomes harder to justify,” Uber President and COO Andrew Macdonald said.
But something’s happen on the other end of the food chain. Quietly and without fanfare I’m seeing a growing number of smaller companies dipping their toes in the AI water…and with success. No, I’m not talking about just getting help drafting emails or researching a topic. I’m talking about real life implementation of AI that’s showing significant ROI promise.
For the past few years the AI story has been around big business. In 2026, it’s becoming a small business play.
The reason starts with Anthropic. Their release of Claude Cowork earlier this year has been a game changer. No longer do I hear ChatGPT, ChatGPT, ChatGPT. Now, everywhere I go, I hear Claude, Claude, Claude. How are they using AI to increase productivity and profits? Here are few things I’m seeing, out in the field.
A Weekly Executive Summary
A number of business owners are now getting information from multiple systems in ways they’ve never had before. A 62-year old owner of a foundry company in Georgia has connected Claude to his ERP system and to certain sites on the Internet and is using it to track the pricing of steel, iron and other core materials to spot trends or identify issues.
A younger operations manager hired a developer and is using MCP (Model Context Protocol) to connect to his AI provider to his accounting, CRM and order entry systems and is now able to track pipeline, orders and activities all on one screen. A client of mine had us create a weekly “executive report” that pulled relevant activities from their Gmail, Dropbox, CRM and accounting systems into a readable narrative delivered to her inbox every Friday that provides her information that goes well beyond just a financial summary.
An AI Tool For The Showroom
The owner of a windows and doors company I recently met spent $10K developing a mobile AI application for his salespeople in the showroom. Thanks to this application, the salespeople have conversations with their prospective customers without worrying about taking notes. The application listens, transcribes, summarizes and — ready? — automatically creates a quote based on the conversation. He told me it’s saving significant time and improving the accuracy of what they’re quoting. His next iteration, he says, will be for the AI application to make product suggestions and search for better pricing as the conversations are happening.
Quick Communications Capture
A client of ours is having us develop an AI system for capturing all phone calls, emails and correspondence into a model that’s shared by teams company-wide. This way anyone can ask anything about a customer in natural language and be completely up to date on their relationship. The system is also being configured to generate proposals, project plans and timelines based on conversations. The cost? About $15K. For a small business, this type of up-to-date shared data is priceless.
An AI Specifications Resource
A friend of mine runs a family-owned business that sells and services farm equipment. He has reams of product literature, warranty information, technical specifications, instructions, guidebooks and manuals for the equipment and the countless parts that go into their operation. All of this documentation is stored in a few online folders. Using Claude (or any other AI provider for that matter) we’re connecting to these storage locations so that her technicians, customer service agents and eventually her customers can query all of this data for quick, accurate information on the equipment they own (or are thinking of buying).
A Knowledgebase To Fix Problems
Remember “knowledge bases?” AI is replacing them.
For this farm equipment reseller the next step planned is an interactive, AI-leveraged knowledge base. They’ve been recording issues and resolutions in a cloud-based service and ticketing system. Now we can connect an AI provider to that system so that technicians and customers can get step by step instructions for fixing issues that have happened elsewhere in the past.
Forecasting
Most of my smaller clients are not mature enough to accurately forecast their cash flow. Yet doing so would have an enormous impact on how they run their businesses and control their costs. Projecting overhead and estimating margins isn’t the issue. The issue is forecasting sales. Which is why we’re now working with a client — a 150-person manufacturer of coated paper and film — to connect their AI provider (in this case Gemini) to their ERP and CRM systems, as well as their folders which store all historical quotes and, based on prior activity and current data, project revenues going forward for 90 days. And you know what? It’s pretty good! The key is continuously asking Gemini to re-evaluate its projections compared to actual invoicing and cash receipts so it can learn and improve.
SMBs Are Using AI. I’m Seeing It.
This is real stuff happening with AI in the real world. It’s not out-of-the box. But the tools are there. MCP — which allows the connection between AI providers and disparate databases — is a game changer. Many AI platform providers like Anthropic, Microsoft, Google and Grok are rolling out built-in connectors to popular business software applications like QuickBooks, Slack, Salesforce and the like. Some of my more DIY-clients are doing these connections and setting up systems on their own. Most are hiring experts and developers to do this for them, like they do for any other business service that they want to leverage without having to learn how the sausage is made.
And because these businesses are small and are generating far fewer tokenization requests, their costs for leveraging these platforms have been minimal, particularly compared with the tens of millions of requests being made by users at larger corporate firms that are now feeling the pain of these costs.
We’re just getting started. I have many more use cases that I’ll share in the coming months. And once agentic AI becomes more reliable, my clients and other SMBs will likely be taking their AI systems to the next level and relying on them to start performing actual work functions, which will free up their over-worked employees to do more productive things with their day.
