You Don't Have an AI Strategy. You Have an AI Pile.

The bill for two years of uncoordinated AI adoption is coming due - and most businesses don't even know what they're paying for.

Subscribe now

Last week, KPMG published a survey of 2,145 business leaders. The headline finding: 49% of executives who deployed AI agents have already scaled them back because operating costs outpaced the benefits. Not because AI doesn’t work. Because they deployed it without a plan.

Here’s the part that didn’t make the headline: those same organizations kept AI as their number-one investment priority, with average spending holding steady at $188 million annually. They’re not retreating from AI. They’re cleaning up after themselves.

If you run a small or mid-size business, that number might feel distant from your reality. You’re not spending $188 million. You might be spending $500 a month, maybe $2,000. But the underlying problem is exactly the same - just scaled down to your budget. You have tools nobody fully uses, automations someone built and nobody owns, and AI subscriptions that renew quietly while the team has already moved on to the next thing.

You don’t have an AI strategy. You have an AI pile.

How It Happens

Nobody plans to build a pile. It grows one reasonable decision at a time.

Marketing signs up for an AI writing tool in January. Sales plugs in an AI email sequencer in March. Operations builds a few Zapier automations that use GPT in the background. The CEO starts using an AI research assistant. IT connects Microsoft Copilot to the calendar. Someone in customer service deploys a chatbot.

By August, you have seven different AI tools, four of which overlap, three of which are actively running automations against your company data, and exactly zero of which anyone has a complete picture of.

Gartner calls this “agent sprawl” - and their research shows it’s not just an enterprise problem. The average organization now runs 12 or more AI agents, with roughly half of those operating without security oversight, ownership documentation, or any kind of governance. SAP flagged this as a board-level risk issue just three weeks ago. The EU AI Act, which hit full enforcement on August 2nd, is making it a compliance issue too.

For a 50-person company, the risks look different than for a Fortune 500 - but they’re still real. Ungoverned AI tools mean:

• Data you didn’t mean to share is being processed by vendors you didn’t vet

• You’re paying for duplicate capabilities across multiple tools

• When something breaks, nobody knows who’s responsible

• You can’t tell a client, an auditor, or yourself what AI is touching your business

The Cost Problem Is Happening Right Now

This isn’t a future risk. The KPMG data, published August 9th, describes what’s already happening. Companies deployed AI fast - and the bills caught up.

The Wall Street Journal reported earlier this year that companies are seeing independently developed AI bots duplicate functions, strain IT budgets, and create governance headaches nobody budgeted for. One analysis found that companies with no central AI inventory are essentially paying for the same capability two or three times because different departments solved the same problem independently.

And there’s a subtler cost: the employee time sunk into tools that don’t get used. Research from Thryv shows SMBs that actually use AI tools report saving 20+ hours a month and between $500 to $2,000. But you only realize those savings if the tools are working. A subscription you’re paying for but nobody trained anyone on is just overhead.

What a Real AI Strategy Actually Is

Here’s what I’ve noticed after working with SMBs on AI implementation: the ones who get real ROI almost always started with an audit before they started adding tools. They knew what they had before they bought more.

A basic AI inventory for a small business isn’t complicated. You just need to answer four questions for every AI tool you’re using:

1. What does it do, and what data does it touch?

2. Who owns it - meaning who is responsible if it breaks or causes a problem?

3. What are we actually paying for it, and is that cost justified by the output?

4. Does it overlap with anything else we’re running?

That’s not a technology exercise. That’s a management exercise. You don’t need a CTO to do it. You need someone to spend a few hours tracking it down.

Once you have the inventory, the decisions usually get obvious fast. You find subscriptions nobody’s using. You find two tools doing the same thing. You find automations running in the background that nobody remembered setting up. And sometimes you find genuine gaps - places where the right AI tool would actually save your team significant time - that you weren’t filling because you were too busy managing the pile.

The Governance Question SMBs Are Ignoring

The enterprise world is having loud conversations about AI governance right now - and most SMB owners are tuning it out because it sounds like big-company overhead.

But the governance conversation isn’t really about compliance frameworks. It’s about basic operational hygiene. Who can spin up a new AI tool? Who has to approve it? What happens when an employee who built an automation leaves? What data are you feeding into which systems, and what are those systems’ terms of service saying about what they do with it?

These are questions every business owner should be able to answer - not because a regulator is asking, but because they affect your costs, your data security, and your ability to actually scale AI across your organization instead of just accumulating more tools.

The companies that are pulling back on AI right now - the ones in the KPMG survey - aren’t pulling back because AI is bad. They’re pulling back because they built without a foundation. The good news is that the foundation isn’t complicated. It’s just work nobody made time for.

What to Do This Week

If you’re running a small or mid-size business and you’ve been adding AI tools for the past year or two, take an hour and build your AI inventory. It doesn’t have to be fancy - a spreadsheet is fine. Get every AI tool your company is using in one place: what it is, what it costs, who owns it, what data it accesses.

Then look at what you’re spending and what you’re actually getting. If you can’t tie a tool to a specific business outcome, you have your first candidate for cancellation.

If what you find suggests deeper work - overlapping tools, ungoverned automations, gaps between what you have and what you actually need - that’s exactly the kind of thing we help with at Black & Tan Labs. Our AI Discovery engagement is designed to get SMBs from pile to strategy: a clear picture of what you have, what you should keep, what you should cut, and where the real opportunities are.

You’ve probably been adding AI tools. It’s time to take stock of what you’ve got.