Home Process Plans About
From the blog

The 90% Rule: Why Your Biggest AI Opportunity Isn't in Engineering

October 19, 2026 · 5 min read

Here is a number that should change how you think about AI: 90%.

Anthropic disclosed it in July 2026, buried in a product update most people skimmed. Of the 1.2 million weekly Claude Cowork sessions running across 600,000+ organizations, roughly 90% are non-developer work. Finance analysts running reports. Marketing teams drafting campaign briefs. Operations managers building triage workflows. Support leads automating ticket resolution. Executive assistants pulling research briefs.

Subscribe now

Only about 10% of AI agent usage is code-related work.

For the first three years of the generative AI wave, the loudest story was coding. GitHub Copilot, Cursor, Claude Code - the narrative was that AI agents were developer tools first, and everyone else would follow. That story was never quite right, but it was the one that got told. Now the vendor’s own numbers have broken it open.

The business implication for SMB owners is direct: your first AI agent is almost certainly not in your engineering department. It is in whichever function has the most repetitive knowledge work per head - usually finance, operations, support, or marketing. And the ROI path is clearer, faster, and cheaper than the code-first narrative ever suggested.

The Data Behind the 90% Rule

Anthropic’s Cowork usage breakdown is not an outlier. It lines up with every major survey published in 2026.

Upwork’s State of AI in SMBs report found that among small businesses actively deploying AI beyond simple chatbots, the leading use cases are decision support (41%), information retrieval (36%), and workflow automation (34%). None of these are coding. The same study found that 74% of SMBs report AI has improved their productivity, and 62% of SMB leaders are ‘very or extremely confident’ handing high-stakes tasks to AI agents.

IDC’s research adds another layer: 40% of SMBs don’t have a single full-time IT employee. A third cite lack of IT staff as a top challenge. The businesses seeing real results are not bolting on separate AI tools. They are turning on AI capabilities inside platforms they already use - their CRM, their accounting system, their marketing platform. The embedded-AI approach wins because there is no new interface to learn and no separate implementation cycle.

The U.S. Chamber of Commerce’s 2026 Small Business Report found that 66% of small businesses now use AI, up from 23% in 2023. That is the fastest technology adoption curve in modern small business history. And those businesses are not writing code. They are using AI tools built by other companies.

What This Changes

If you are a business owner or operator, the 90% rule changes three things about how you should think about AI.

First, stop treating your AI budget as an IT line item. If your AI spending lives inside the technology budget, it is under-scoped from the start. IT budgets are sized for infrastructure - servers, licenses, security. They are not sized for buying back human time across every department. Move the AI budget to the operating budget of the function that will actually use it. Rebill IT for the infrastructure component. This unlocks 3x to 5x the spend that is actually justified.

Second, your first agent is in operations, not engineering. Even if you have developers, they are the wrong first team to build for. The highest-density ROI is in whatever function produces the most similar-looking outputs every day - the finance team closing the monthly books, the support team triaging the same types of tickets, the operations manager sending the same kind of status updates.

Third, deploy one narrow agent, learn from it, and let strategy follow execution. A 12-week AI strategy engagement without a deployment is not what you are missing. What you are missing is a single working agent that has touched real customer data and produced measurable results. The strategy sharpens against real data, not a whiteboard.

What Real ROI Looks Like

The case studies that matter most for SMBs are not the ones from Fortune 500 companies with AI engineering teams. They are the ones that look like your business.

A home services company was missing 40% of inbound calls during business hours and nearly all of them after hours. Average callback time was 4.2 hours. They deployed an AI call agent for 97 per month. Within 90 days, call answer rate went from 60% to 100%, lead capture increased by 73%, and monthly revenue grew by 2,000. That is a 4,040% ROI.

A dental practice deployed the same kind of agent. A real estate agency followed. The pattern is consistent across every case study: the AI pays for itself within the first month and compounds from there.

The Bluehost 2026 State of Small Business AI Confidence report found that businesses saving 16 or more hours per week using AI are 3.7 times more likely to report revenue growth than those saving only 1 to 3 hours. The threshold where AI shifts from a productivity perk to a revenue driver is real.

How to Find Your First Agent

You do not need a consultant to tell you where AI fits. You need a framework to look at your own operations honestly. Four criteria:

  1. Repetition: Does this workflow run daily or weekly in the same shape?

  2. Volume: Does it consume 4 or more hours per person per week?

  3. Data source stability: Does it live in 2 to 3 systems you already use?

  4. Judgment complexity: Is it mostly rules-based with occasional exceptions?

If a workflow checks all four boxes, you can ship an agent this quarter. If it is close, you can ship after a light two-week audit. If any criterion is a clear no, pick a different workflow first.

The businesses that get ahead on AI are not the ones with the biggest budgets or the best models. They are the ones that looked at their own operations first.

That is what B&TL’s Discovery engagement is built for: a structured two-week audit that maps your operations, identifies the highest-ROI automation opportunity, and leaves you with a working agent - not a deck. No engineering pedigree required. Just a willingness to look honestly at where your team’s time actually goes.

The 90% rule says the opportunity is real, it is now, and it is not about technology. It is about operations. The only question is whether you are looking in the right place.

Enjoyed this? New essays on AI, operations, and building in contrast - delivered to your inbox.

Subscribe on Substack →