I spoke to businesses about AI. This is what I learned.

Hi everybody, 

I spent the past weeks talking to dozens of small businesses. (I’ll tell you why below.) 

I wanted to know their real experience of AI. What’s actually working? What are they trying next? What’s just a pipe dream. 

Here’s what I heard - and where I think it’s going.

Based on what I'm seeing now, the use cases for AI in regular Main St businesses fall into four different stages. Each one looks further into the future.

Stage 1: Single-Player Mode

Most businesses are here already.

A typical white-collar worker on a computer spends 95% of their time in just a few different applications:

In the past year, though, AI has joined this shortlist. It’s ChatGPT, Microsoft Copilot, or something similar. It’s become a reliable and consistent part of a knowledge worker’s workflow. 

Personally, this bucket is 95% of my use case for AI. It helps with my research for studying companies, pulling in sources and finding data points so I can be faster and more effective than if I had a team. It's totally worth the $200/month that I spend on it.

One interesting thing I heard: the people using these types of AI tools are the employees who were outperforming before. 

They were already better at Excel, PowerPoint, or email. Now that an even more powerful tool has shown up, it's made them even stronger compared to typical employees.

For 95% of businesses, this is the extent of their AI use.

—

Stage 2: Customer-Facing AI

This is where I'm seeing the biggest growth right now. Stuff like customer service, email responses, automated processing of a shared inbox.

Tons of people are building (or want to build) agents for this stuff based on automation tools.

The most interesting stuff is around voice and text communication with clients. Stuff that once required a whole call center can now be done with a much smaller team of humans augmented by strong voice or text agents.

One person had a single individual and a set of cloud AI tools handling the workload of six accounts receivable personnel. Instead of a human sending texts, running through scripts, doing follow-ups, and answering questions, it took one person to supervise the bots.

Another wants his property management business voice AI to handle and triage calls after hours. Next he’s looking at having it send out automatic repair requests with no human intervention.

The key here is ensuring that these bots are sufficiently reliable to be actually helpful, rather than just frustrating potential customers.

The other use cases here are addressing “internal customers” — so stuff like an HR chatbot, or an interactive internal knowledge base. 

It makes sense: scaling up customer support or selling with less staff has a big upside, and if the bots have a brain fart, the damage is relatively low risk. 

Only a tiny percentage of businesses are doing this at scale today. But it’s the next frontier. 

AI usage will grow here for the foreseeable future. 

Stage 3: Internal Process and Workflow Automation

At this stage (which I think is still a ways in the future), AI systems will be baked into internal processes for a company. 

Stuff like bank reconciliations, invoicing, verifying that a given vendor meets qualifications.

Today, this isn’t working very well.

Sure, there are tons of internal processes that are perfect for automation… in theory. But when it’s mission-critical stuff, you can’t take a risk on AI hallucinating and screwing it up.

So for now, a ton of paper-shuffling jobs are safe. These are key people, running processes with old-school software and elbow grease. 

You might have seen MIT’s sobering report that 95% of organizations are getting zero return from AI. 

Two thoughts: 

My sincere advice is, don't try to add this stage 3 stuff to your company at this point. It's just doomed to fail.

—

Stage 4: Digital Employees

This is where we’re headed eventually. Lots of people will hate this, but at this stage every employee is basically a manager for AI agents. 

You'll talk to them just like a remote employee: lots of Slack and occasional live conversations. They’ll have long-term memory and learn their job over time.

Enterprises are ready for this. Managers have been learning how to hire, train, motivate, and compensate employees for centuries. Then in 2020, we all learned to do it remotely. 

But you, me, and everybody else making a salary these days says that the technology is nowhere near ready for this.

And like most things that come out of Silicon Valley, expect that people are pretty good at predicting the eventual future, but wildly wrong about the timeline. 

I bet we’re cool for decades.

—

There’s an opportunity here.

My thesis is that a business can develop a real, sustainable competitive advantage by adopting some of these Stage 2 technologies. 

Can you automate and enhance your customer communications via voice and text? 

The tools are out there, and the world is pretty much ready. The challenge is that they’re still a bit of a mess for the average person to deploy… especially if you don’t want them to suck.

So, AN ASK: I’m mentoring a team who help people deploy AI voice and text agents for customer service, support and sales functions. 

If you’re a business owner, and you’d be interested in talking with us (and potentially getting some help setting this stuff up), reply to this email and we’d love to talk to you.

That’s where I’m at these days on AI. 

What do I have wrong? Know someone we should talk to? Hit reply and let me know. I read all your messages!

—

Thanks for reading!

Michael

P.S. Quick reminder: if you’re interested in deploying AI voice and text agents in your business for CS, support, and sales — we’d love to talk to you. Reply to this email!