The AI divide is growing for SMBs — Here’s how they can still win

Keith Shaw: AI may be dominating boardrooms at the enterprise level, but what’s going on at small and mid-sized businesses? On this episode of Today in Tech, we dig into the biggest AI hurdles and opportunities for smaller companies. Hi everybody, welcome to Today in Tech. I'm Keith Shaw.

Joining me on the show today is Ed Keisling. He is the Chief AI Officer at Progress Software. Welcome to the show, Ed.

Ed Keisling: Thanks, Keith. Great to be here. Keith: So let’s set the stage a bit by talking about the differences between large enterprises and small or mid-sized companies. How do you see the approaches to AI differing — whether in terms of budgets, talent, or risk appetite?

What stands out to you? Ed: There's definitely a “haves and have-nots” situation. I remember an article a couple months ago that mentioned how if you want to spend $10 million with OpenAI, they'll graciously assign forward-deployed engineers to help you.

But the reality is, very few organizations can afford that, or are organizationally ready to take that step.

Some industries like finance, healthcare, and automotive have a legacy with machine learning and were able to make the leap and benefit from these services — but that’s still a fairly limited group.

Keith: So for those that don’t have that kind of budget, are they giving up? Or are they looking for alternatives? Ed: Absolutely, they’re looking for alternatives. It’s often a “build versus buy” decision.

But it’s also very spiky — some organizations, like one I spoke to at a conference in France, had partnered with Mistral and were doing all the cutting-edge agentic development. That speaks to how complex it still is to roll out these applications. The level of expertise required is high.

But there’s still a lot of white space out there. You can absolutely start small and scale.

Keith: Over the past three years of GenAI’s rise, have you seen the gap close a little bit? Are more SMBs getting into the game? Ed: There’s definitely appetite from organizations of all sizes, but that brings new problems.

Many are rushing in to “AI-ify” everything without understanding the problem they’re trying to solve. For smaller businesses especially, the perception is, “This looks easy — just plug in some AI and go.” But getting these tools into production is still incredibly difficult. That’s the first major pain point.

Keith: Are SMBs making the same mistakes as large enterprises, or different ones? Ed: I think we’re all in the same boat, figuring this out together. One big mistake across the board was the hype around agents.

Unless you were one of the chosen few working with foundation models, it was really hard to put those agents into production because the frameworks weren’t mature yet. That became a huge distraction.

Companies were chasing agents instead of focusing on easy wins — like using GenAI to save 10 minutes a day through automation.

Keith: You mentioned build vs. buy earlier. Since many SMBs don’t have in-house AI talent, are they leaning more on external partners or off-the-shelf tools? Ed: Definitely. Every business has to ask, “What do we do better than anyone else?” That’s where you want to enhance with AI.

But the surface area of AI is so broad, no one can keep up alone. We recommend forming close-knit relationships with trusted vendors and advisors — people you can ask transparent questions, who are invested in your success, and won’t try to oversell you.

Keith: There were some high-profile AI flops last year. So would you say: don’t jump into the pool until you’re ready? Ed: Exactly. Jumping in too fast can be disastrous. Many companies started building agents without understanding foundational building blocks. That’s risky. Instead, start small.

Understand what AI is capable of. That hands-on experience is critical. Keith: Right — and smaller companies probably don’t have their data ready, either. Ed: That’s another big challenge. Often, people try to automate processes that aren’t even documented — they’re just in someone’s head.

Before you automate, you need to understand the process. Once you frame the problem properly, the data issues become easier to address.

Keith: So how should SMBs prioritize when selecting AI tools? Should they focus on cost, usability, flexibility? Or just outcomes? Ed: Vendor selection is tricky. Larger companies benefit from bundling and consolidated support. But that can also create price pressure and single points of failure.

We had a client tell us, “Working with you is like going to a local hardware store — you’ve got just what we need, and someone to answer questions.” That’s what SMBs need: a partner who understands their problems and isn’t overselling solutions.

Keith: Let’s say they’re ready: data is clean, vendor selected — now what about risk and governance? Can SMBs experiment safely? Ed: Larger companies often have so many policies and governance layers that they do nothing. SMBs can move faster.

It’s easier to communicate risk, to explain what tools to use or avoid. That makes it easier to run safe experiments.

Keith: What essential governance steps should companies — large or small — take before going to production? Ed: Like any software, AI needs to go through security review, SDLC processes, observability, metrics — you need to check all the boxes. But AI adds extra risk.

You must monitor for bias, inference misuse, unauthorized access via chatbots, and so on. Think about the worst-case scenario: what’s the blast radius? Can you pull the plug quickly if needed? Keith: And sometimes the pressure to “just ship it” comes from the business side, right? Ed: Definitely.

Influencers make it look easy — but MVPs are only 80% of the work. That last 20% — security, observability, scaling — is real work. You can't skip it. Keith: So would you recommend that SMBs go for smaller wins first? Ed: Absolutely. I have strong opinions here.

Most workers aren’t using AI beyond summaries or writing help. But tools like NotebookLM or DeepResearch can give you weeks of research in minutes — huge productivity gains. Every employee should find their own AI win. That builds confidence and literacy.

Keith: So the goal is to get humans comfortable with AI, and they’ll find the wins organically? Ed: Exactly. It's a “crawl-walk-run” model. You need to play with the tools, hit the edges, and learn limitations. That’s how you build toward more complex workflows.

Keith: Should SMBs focus on internal use cases first, instead of external ones like customer service chatbots? Ed: Yes. One great place to start is agentic RAG (retrieval-augmented generation). Many companies have tons of unstructured data — documents, videos, etc. — that’s hard to access.

RAG forces you to organize that data, and it can power both internal knowledge bases and external support tools. It’s a proven entry point.

Keith: Let’s talk about workplace culture. At the SMB level, are employees still fearful of AI? Ed: I’d argue that if you’re only using AI for efficiency, you’re missing the bigger opportunity — innovation. At Dell, Michael Dell said other companies would eventually offer Dell’s services faster, better, cheaper.

That was their burning platform. That’s the mindset all companies need. If your only reason for AI is “make everyone work faster,” that’s not a good narrative. You need to articulate the why clearly — and empower employees to grow and gain new skills.

Keith: You opened by noting a gap between the “AI haves” and “have-nots.” If that gap persists, is that your biggest fear? Ed: It is. Too many people still don’t understand what AI is actually capable of. You need to use it regularly to gain that understanding.

And yes, there’s still time for SMBs to catch up. Their personal customer relationships and localized knowledge are huge differentiators. But they have to start now.

Keith: We’ve had other guests say the same — SMBs can be more agile and compete faster if they embrace the right tools and mindset. Ed: Absolutely. I’ve worked at companies like Pegasystems when it was small. Everyone wore every hat. AI fits naturally into that kind of environment.

Keith: Let’s end with a look at the future. What excites you most — and what concerns you? Ed: Agentic tech might finally become real. Gartner’s hype cycle has it at the peak of inflated expectations, but I think we’ll see value in the next 2–4 years.

I just got back from GitHub Universe, and I believe agentic coding will be a major shift. The same frameworks powering developer tools will fuel agentic automation for business, too. Keith: Will agents take off more in business than on the consumer side? Ed: I think so.

Business applications are more structured, and frameworks like A2A and MCP are maturing. I think we’ll all feel the impact in the next 6–9 months — especially in business. Keith: Very cool.

Ed, thanks so much for joining us — this was packed with insights. Appreciate your time today. Ed: Thanks, Keith. Great to be here. Keith: That’s going to do it for this week’s show. Be sure to like the video, subscribe to the channel, and check out our other episodes.

I’m Keith Shaw — thanks for watching.