The AI Brief #37 AI ROI SMB automation AI agents AI infrastructure generative AI adoption

AI costs 2.5 trillion, but 95% of projects deliver nothing

Rodrigue Le Gall | | 3 min read

Gartner just released its 2026 forecast: 2.5 trillion dollars in global AI spending. Impressive, except this same week, MIT’s NANDA initiative drops a report that hits different: 95% of generative AI projects in enterprises produce zero measurable return. Not weak returns. Zero.

This isn’t a stat pulled out of thin air. It’s based on hundreds of real projects, from chatbot implementations to complex automation systems. The common denominator? These projects didn’t address a specific business problem before launch.

Here’s what happens: SMBs see everyone investing in AI, read use cases promoted by vendors, and launch initiatives without clear ROI frameworks. They buy the dream, not the solution. Result: piles of unused data, employees who won’t use the tools, or worse—tools that create work instead of eliminating it.

The real red flag? MIT isn’t saying “AI doesn’t work.” MIT is saying “5% of projects deliver returns.” These 5% share one trait: they started by identifying an expensive, measurable, painful workflow to automate—before choosing a solution.

What this means for your business

For you: stop thinking “AI” and start thinking “process that’s costing me time and money.” An SMB that launches an AI project without first defining the cost of its inefficiency is headed straight for the 95%. Before buying a tool, answer this: which process is eating up your team’s time? What’s it costing monthly (salaries + errors)? How much will I save if it’s cut by 80%? If you can’t put a number on it, you’re not ready. And honestly, that’s a healthy filter.


In brief

SMBs can finally automate admin tasks without coding

MIT Technology Review explores how AI tools can handle accounting, HR, and client research directly. No custom development needed. The catch: these tools work best when the process is already documented and repetitive. Before buying, map out your actual admin tasks.

Read source

Apple launches AI agent platform for SMBs

Poke becomes the first AI agent app approved on Apple’s Messages for Business. What this means: AI agents are starting to integrate into existing channels rather than forcing users to new interfaces. For SMBs already using iMessage Business, this is a frictionless entry point.

Read source

When everyone has access to AI, what actually matters?

Reddit thread that asks the real question: if AI becomes a commodity, the differentiating skills become critical results evaluation, workflow design, and domain expertise. For SMBs, this means your competitive edge won’t be “we have AI” but “we know how to use it better than everyone else.”

Read source

Amazon deploys robots you can command by voice

Amazon’s new Proteus now responds to natural language commands instead of code. Signal: logistics automation (warehousing, supply chain) is finally becoming accessible to SMBs without technical expertise. Worth watching if you manage a small warehouse or distribution center.

Read source

AI chip shortage: TSMC crying for help

The world’s leading semiconductor maker can’t keep up with AI demand, even while building factories in the USA. The implicit message: GPU and TPU prices are staying high. For SMBs considering local AI infrastructure, hardware costs remain prohibitive in the near term.

Read source

Get The AI Brief in your inbox

3x per week, the essentials of AI decoded for business leaders.

Subscribe

Take action

Ready to automate your repetitive tasks?

Discover what AI can concretely change in your business. In 2 hours, we identify your automation opportunities.

Free AI Checklist

10 processes to automate in your business

Download PDF