For twenty years, software has been predictable. You paid $15 per seat for Slack. $30 for Salesforce. Fixed monthly fees. The vendors loved it because once they built the code, serving you cost almost nothing. Those 80-90% gross margins weren't an accident.
That model just ended.
The Rise of Agents
Microsoft, Google, Anthropic, and OpenAI are all building agent platforms right now. They're letting you build workflows that pull from your Drive, write to your CRM, analyze transcripts, and generate reports. But here's the part nobody's talking about: none of this runs on subscription pricing.
It runs on tokens. And tokens cost money every single time they execute.
Your Zoom meeting agent that grabs the transcript, extracts action items, writes follow-up emails, and updates your project tracker? That's not included in your $20 Google Workspace seat. That's burning tokens through the API. Every meeting. Every month. Forever.
Scale that across every workflow you're building and the math gets uncomfortable fast.
I’m launching round two of my AI Cohort training for small business owners next month. Sign up below if you are interested in getting an AI kickstart.
The Token Efficiency Ratio (TER)
I’m a big fan of using Greg Crabtree's Labor Efficiency Ratio that he wrote about in his first book Simple Numbers, Straight Talk, Big Profits. He showed that the best companies measure revenue per dollar of labor spent because labor is the biggest cost.
You're about to need a Token Efficiency Ratio for the same reason. Which customers consume more tokens to service? Which workflows justify expensive models versus cheap ones? Where are you burning money on AI that isn't generating proportional revenue?
Could I Be Wrong?
The obvious counterargument is that token costs will drop to near zero as compute scales. More data centers, cheaper models, efficiency gains. Maybe.
But I'd bet on Jevon’s Paradox instead. When something becomes cheaper and more abundant, consumption explodes. We won't just run simple tasks on cheap models. We'll keep finding ways to leverage the cutting-edge capabilities because they create competitive advantage. Your overall token budget grows even as per-unit costs fall.
The vendors know this. They're betting on it. Usage-based pricing means they capture the upside as you scale, not just a fixed seat fee.
What Does It Mean For You?
So while you're excited about building agents, start tracking the economics now. Measure token consumption by workflow. Calculate revenue per token dollar spent. Figure out which processes justify premium models and which don't.
Because the companies optimizing for token efficiency in 2025 will have a cost structure advantage that's almost impossible for competitors to close by 2027.
The game changed. Most owners just haven't noticed yet.