(This column originally appeared in Forbes)
There were a number of surveys published this year about the increased use of AI amongst small businesses this year. Most of this wasn’t really news. It was simply using popular AI chatbots like ChatGPT and Claude to answer questions, clean up emails and perform advanced searches.
However, as I’ve been travelling around the country and meeting with companies both at conferences in their offices I’m noticing a growing number of business owners starting to step up their AI usage by connecting these chatbots to their data using MCP and turning these interfaces into natural language reporting engines.
AI Using Model Context Protocol
I liken MCP, or Model Context Protocol, to ODBC (Open Database Connectivity) that, for many years has used by countless organizations to connect interfaces and other applications to their databases. In the AI world, MCP can do just about the same thing by providing a conduit between a chatbot — licensed or created — to a company’s accounting, customer relationship, human resources, order entry and inventory systems as well as to their files, folders and other data.
It’s not that hard. In fact, most of the mainstream AI platforms already come with out-of-the-box connectors to popular software applications like QuickBooks, PayPal, Docusign, Gmail, Outlook and Dropbox. MCP — or a form of it — is used by these vendors to do this. Companies that rely on applications, or proprietary systems where an out-of-the-box connector may not be available, can hire an AI developer to use MCP or similar connector to accomplish the same thing.
AI Agents? Not Yet.
How are small and mid-sized business owners using AI once these connections are established? Well, for one thing, they’re certainly not letting AI run their businesses. Bigger companies and software vendors are touting agents to perform a number of tasks and make decisions for business managers in lieu of employees. They’re leaning on these agents to create invoices, reconcile accounts, generate service tickets, schedule service visits and a number of other chores.
However, at least for now, I’m not seeing many small businesses demonstrate the same kind of faith in these agents as their larger counterparts are doing. That’s not to say this won’t happen in the very not-too-distant future. But, for now, agents are not being used much, if at all, by small businesses.
AI Reporting
What they’re doing instead is leaning on their newly connected AI chatbots for natural language reporting. Because, after a few weeks of learning, the chatbots get better at their jobs. Which means that they can better understand a company’s underlying data and generate answers about it.
For example what customers haven’t ordered from you in over 6 months? Which customers are your most and least profitable? What are your projected sales for the next 90 days taking into consideration open quotes, pipeline data, seasonal activity and historical sales?
How about reviewing your health insurance contract and bills from your provider and suggesting ways to reduce these costs? Or looking at the past 6 months of uploaded utility bills to find discrepancies?
Please create a rolling forecast for my stuffed bears product line. Please generate a monthly income statement for my peanuts division. What employees have the most overtime? What machines are running the most productive? Show me a history of material prices from my Alabama supplier.
I’m clearly scratching the surface. But my clients aren’t. Some of them — and surprisingly to me they skew older in age because it seems like the older you are the more wondrous you are about this new technology — are spending hours prompting their chatbot about their company, asking for advice, analyzing the numbers, spotting opportunities. They’re doing this on their desktop, laptop, tablets and phones. They’re getting numbers from these chatbots that they never got from their accountant or their finance team…or were simply just too afraid to ask. There’s no sorting, filtering or formatting.
Business people ask me all the time how they can be using AI in their business. This is what I suggest. Connect it to your data. Then ask and ask and ask. Just don’t blame me for the answers.
