Your AI can see your books. It can't see your business.

Why AI stalls in $5–50M companies — and what the data actually says about it

---

Every finance tool you use is suddenly “AI-powered.” Your accounting software will summarize a report. Your CRM will draft an email. ChatGPT will explain your P&L back to you if you paste it in.

And yet the question you actually want answered — which location is losing money on labor this month? — still takes someone half a day and a spreadsheet.

That’s not an AI problem. It’s a reach problem. AI can only answer with the data it can get to, and in most growing businesses, most of the data is out of reach.

---

The gap between wanting AI and being able to use it

Almost every business says it’s doing something with AI. Very few are getting much out of it, and the reason keeps coming up the same.

The most direct evidence comes from a 2026 global study of small and mid-sized business leaders, commissioned by AWS and conducted by Techaisle. 96% of SMBs are acting on AI, but 54% say they’re held back by data. The top barriers are cybersecurity (53%), data quality (47%), and integration complexity (47%). The study’s own summary: SMBs are “AI ready in mindset but data blocked in practice.”1

Read that as a sequence, not a list. Businesses tried AI, hit the wall, and figured out the wall is their data.

---

It isn’t a big-company problem

You might assume the largest companies have this solved. They don’t — they’ve just named it.

Deloitte’s Q4 2025 CFO Signals survey asked 200 CFOs at North American companies with at least $1 billion in revenue about their 2026 priorities. 54% named integrating AI agents into finance a top transformation priority. Right behind it, at 52%, was improving data quality, access, and usability.2

Those CFOs have IT departments, data teams, and budgets. They still put “make the data reachable” essentially tied with “deploy the AI,” because one doesn’t work without the other.

Now take away the IT department, the data team, and most of the budget. That’s a $15M business.

---

The smaller the company, the bigger the gap

The Techaisle study draws a line at $30 million in revenue: below it, AI decisions are CEO-led and cost-driven; above it, they become CISO-influenced and productivity-focused.1 In plain terms — under $30M, the owner is making the call, and nobody on staff owns the plumbing.

The companies that can afford someone to connect the systems are the ones actually using AI in the business.3 Bluevine’s 2026 survey of 942 U.S. small business owners found that 82% have hit at least one barrier when trying to use AI more deeply.4

None of these surveys isolate the $5–50M band exactly — “SMB” usually means under 500 employees — but they bracket it, and the pattern holds at every size. The willingness is there. The wiring isn’t.

---

What “out of reach” actually looks like

We work with a 21-store Crumbl franchise group. Before we started, the numbers the owners needed lived in nine places: timesheets in When I Work, payroll in ADP, the books in QuickBooks, sales and labor in Crumbl’s internal system, a set of Google Sheets, and expenses split across Divvy, Bill.com, and Sysco invoices.

Any single AI tool could see one of those. None could see all of them. So “what’s our labor-to-cookie ratio by store this week?” — the question that actually runs the business — wasn’t answerable by anything except a person with a spreadsheet and a free afternoon.

That’s the normal state of a growing company. Accounting in one system. Sales in a POS, a Shopify store, or a CRM. Operations in payroll, scheduling, inventory, or a web app’s database. Reports assembled by hand, monthly, by someone who’d rather be doing their real job. Nobody set it up this way on purpose. It’s just what happens when a business grows faster than its systems.

---

The fix isn’t more AI

The instinct, when the AI doesn’t answer well, is to try a better AI. That’s backwards. A better model with the same one-ninth of your business gives you a more articulate wrong answer.

The fix is to make the whole business reachable: connect every system into one place you own, define each metric once so finance and operations agree on it, and keep it current automatically. Once that exists, the AI part is almost trivial — ask a question in plain English, get an answer from live data, with the query shown so you can check it.

That’s what we built for the Crumbl group. Consolidated P&Ls across every store, labor-to-cookie ratios, and their core KPIs — delivered weekly, without anyone building a report. The owners ask questions of their own data now instead of waiting for someone to answer them.

---

Greg Armstrong is the founder of StatLogic, a managed data engineering firm in Highland, Utah. StatLogic connects a company’s systems into a warehouse it owns and delivers the numbers — and an AI that can reach all of them.

Sources

1. AWS Global SMB Quantitative Study 2026, conducted by Techaisle. Survey of IT and business decision-makers at small and mid-sized businesses worldwide. Summary PDF · AWS analysis
2. Deloitte Q4 2025 CFO Signals Survey. 200 CFOs at North American companies with $1B+ in annual revenue; fielded Q4 2025, published January 2026. Deloitte press release
3. U.S. Census Bureau, Business Trends and Outlook Survey (BTOS), December 2025 – May 2026. Biweekly, nationally representative sample of ~1.2 million businesses. Census analysis
4. Bluevine 2026 Small Business AI Trends Report. 942 U.S. small business owners and financial decision-makers. Report