SMB owners, you don’t have an AI problem
After 20 years in Silicon Valley — from founding and selling a company to working at Netflix — I moved home and have spent the last few weeks meeting with a few business-owner friends. The question I kept getting was some version of:
“How do we implement AI?”
After digging into several companies over the last few weeks, I kept seeing the same patterns. Now for context, they were all established, 20+ year-old companies:
- Multiple-location retailer
- A convenience store chain with 10+ locations
- A local manufacturer
Different industries. Same underlying problems.
The low-hanging fruit is not AI. It is something much more basic: SMB tech debt.
If you are a software engineer, you know the term “tech debt.” Most business owners probably do not use that phrase. So let me describe it in the way I have been hearing it lately:
- “Can you take a look at our marketing and tell us if it is working?”
- “This team is slow, but we cannot figure out why.”
- “We use this tool, but it is not set up for how our process actually works.”
- “The data has to be manually entered into multiple systems.”
- “People manually copy and paste which creates errors.”
I could go on, but you get the idea. Increased complexity introduced by software and data.
If you have said some version of these things, you probably have tech debt that is making your business harder to operate.
IMO bolting AI onto messy systems makes the mess worse (and much more expensive).
These are the problems I keep seeing.
1. No one owns the applications, data, or systems.
Inside software companies, someone owns the systems. Someone owns the data. Someone is responsible for how tools connect and how information flows.
But in many SMBs, this falls through the cracks.
The owner handles some of it.
A manager handles some of it.
The employee who is “good with tech” handles some of it.
The MSP keeps the network and devices running.
But no one truly owns the application layer: the CRM, ERP, accounting system, website, marketing tools, reporting, integrations, and data flow across the business.
So the systems & data between applications.
2. Software spend becomes an afterthought.
Thousands, and sometimes tens of thousands, of dollars a month becomes “just the cost of doing business.” But basic questions often are not being asked:
Do we still need this tool?
Are we paying for duplicate functionality?
Is the software set up around how the business actually operates?
Are we getting useful reporting out of it?
Often, no one really knows because no one has looked at the stack end to end.
3. Processes are not documented or thought through.
Because no one owns the full workflow, one tool gets connected to another, and the gaps are filled with manual work.
Someone exports a file.
Someone cleans up a spreadsheet.
Someone uploads it into another system.
Someone copies and pastes data from one place to another.
An employee updates a field incorrectly which is never caught.
What was originally thought of as "not a big deal," later become errors. Errors create delays. Delays create poor customer experiences.
And because the process is not documented, the business usually cannot see where the real bottleneck is.
4. There is no central source of truth.
- Sales puts data into the CRM.
- Operations puts data into the ERP.
- Finance has its own numbers.
- Marketing has its own reports.
- The MSP keeps the infrastructure running.
But no one is mapping how data should flow through the business. So the company ends up with multiple versions of the truth.
- Customer records do not match.
- Jobs do not match.
- Orders do not match.
- Reports do not match.
At that point, messy data without an easy way to resolve the solution.
5. Analytics and accountability are missing.
Most owners I spoke to can feel something is broken.
They know when a team is slow.
They know when marketing feels inefficient.
They know when operations are creating too much friction.
But they often cannot prove where the issue is or what is causing it. Questions like -
- Is the marketing spend actually converting?
- Which team is creating the most rework?
- Where are jobs getting stuck?
Because the data is scattered or unreliable, decisions get made based on gut feel instead of measurable data.
So what am I proposing?
Instead of rushing to AI -
1. Put someone in charge of the systems and data flow.
This does not mean every SMB needs a CTO.
But someone needs to own the software stack, the data, and the way systems connect.
2. Document the workflow.
How does work actually move through the company?
Where does data start?
Where should it live?
Who owns each step?
Where are the handoffs?
3. Create a central source of truth.
Once the workflow is clear, the data can be cleaned up and organized into a central source of truth.
4. Analytics, analytics, analytics.
Whether you use Power BI, Google’s Data studio, Tableau, you name it. You need to have data flow to a business intelligence / data visualization tool. Owners need visibility into what is actually happening across the business.
5. Track the impact of every change.
Analytics are only useful if they change how you operate.
Before you change a process, implement a new tool, or automate something with AI, establish the baseline. How long does it take today? How much does it cost? How many errors happen? What does the customer experience look like? Then make the change and measure the result.
- Is it saving time?
- Is it reducing cost?
- Is it improving accuracy?
- Is it improving the customer experience?
- Is the juice worth the squeeze?
- Or is there a simpler, cheaper, more reliable fix?
This is the feedback loop we rely on in software: instrument, measure, improve. I think more SMBs would benefit from adopting the same mindset.
Only then can you begin thinking about where AI can be used to improve the business. Once implemented, then you can actually measure and empirically state whether it is working or not.
Hopefully this helps some other business owners out there. Happy to answer any questions.