TL;DR:

The defining resource of the internet was page views. The defining resource of AI is tokens. But while companies can measure exactly what their tokens cost, they still can’t easily tell what those tokens produced.
Tenor connects the two. Every piece of AI work starts with an outcome and a budget. Tenor tracks the resources spent against the result produced, learns from that result, and helps determine where the next tokens should go.
That’s the bigger opportunity: not just attributing AI spend, but allocating machine intelligence toward what creates value. Once a company knows where its tokens make money, it can do more of what works, less of what doesn’t, and continuously reshape itself around where AI creates value.

https://youtu.be/rRaLDa9yvC8

🤯 The Problem

While companies have become good at measuring AI consumption, they still can’t tell what that intelligence produced in business value.

The internet went through a similar transition.

Its defining resource was page views. Early on, advertisers could buy impressions and measure exposure. Search advertising changed the model by connecting spend to intent and outcomes. Google then turned that into an engine for allocating dollars toward what performed.

AI is still early in that transition. Companies deploy copilots, agents, and automations across the organization. Each produces activity and consumes tokens. But the economic context usually lives somewhere else.

What was this intelligence supposed to accomplish? What changed because it ran? Was the outcome worth what it cost? Should the next token go here or somewhere else?

Without those answers, AI remains largely a software expense: something companies buy and monitor rather than a resource they can economically allocate.

And the bigger loss isn’t wasted tokens. It’s lost learning.

Millions of pieces of work happen, outcomes disappear into different systems, and very little of that experience determines how the next unit of intelligence gets deployed.

⚙️ The Solution

Tenor starts with the outcome.

Before intelligence is deployed, the work gets an economic identity: what it owns, what success means, what it can spend, which systems it can act on, and when humans need to intervene.

Now every token is spent in service of something measurable.

So instead of: This agent consumed X.

You can ask: What did X produce?

That closes the loop: allocate → work → outcome → attribute → learn → reallocate

Attribution tells you what worked. Allocation is what makes that knowledge valuable.

Work that produces value can receive more intelligence. Work that doesn’t can be changed, rebuilt, or stopped. Every outcome improves the next allocation decision.

Over time, that feedback loop reaches beyond individual workflows.

Responsibilities move. Processes disappear. New ones emerge. Software changes around the work. The organization gradually restructures itself around where machine intelligence produces the greatest value.

That’s what we’re building Tenor for: not just to measure the token economy, but to give companies the machinery to operate inside it.

👋 Team

Muhtasham Oblokulov — AI researcher, 3,000+ citations, ICML-published. Former Senior ML Engineer at Munich Re, where his models drove $1B in revenue.

Hamze Al-Zamkan — President of TUM.ai, which he grew into Europe's largest student-led AI organization and a talent pipeline for Munich's AI ecosystem. Deployed AI systems at appliedAI, and worked on growth at Voggt and Terra. Studied CS at TUM and CDTM.

Amgad Al-Zamkan — Studied CS at TUM and UC Berkeley. Researched adaptive decision systems. Automated judgment-heavy work at Amazon.

🙏 Our Ask

If your AI spend is growing and your company isn't visibly getting better at its work every week, that gap is our market.

Want to see what the work produces? email hamze@heytenor.com