Agentic math includes shifting costs of service (expenses) as users migrate from chat interfaces to agentic workflows. Tokens are the unit of work and the related expense shifts (in part or whole) to the owner/user of the agent consumption, subsidizing the cost of compute.
We'll see how the users/customers feel about that. Some will accept it. Others won't. What we're describing is significantly more expensive software than anyone has budgeted for.
Truth. Ask any new owner of a workflow who needed to upgrade from $20/mo to $200/mo in token spend. It is a reality. That forces the right questions IMO.
Great article Sam. In the AI business, it is hugely important to know what you don’t know. In that sense the piece was extremely informative highlighting the real constraints in achieving sustained “SaaS” like margins.
I am curious to understand the evolving relationship between FTE reductions and the increase in AI infra spend.
Can you create great businesses by shrinking them?
Sam, great piece. You’ve shared some really deep insights here with plenty of nuances to think about.
From my own experience, every SaaS company I interact with is constantly pivoting toward AI integrations. However, they seem solely focused on the new features, while the financial downsides and real infrastructure costs are almost never addressed. Everyone is chasing the hype, but the "Agentic Math" you described is exactly the reality check the industry needs right now.
Thanks for the comment, Alexander. Yes, that's really the purpose of the piece. Because, speaking from experience, sometimes you build up organizational infrastructure (in the form of costs) that becomes very hard to unwind.
This was a great summary. Sam. Going to comment/ share on LI
Agentic math includes shifting costs of service (expenses) as users migrate from chat interfaces to agentic workflows. Tokens are the unit of work and the related expense shifts (in part or whole) to the owner/user of the agent consumption, subsidizing the cost of compute.
We'll see how the users/customers feel about that. Some will accept it. Others won't. What we're describing is significantly more expensive software than anyone has budgeted for.
Truth. Ask any new owner of a workflow who needed to upgrade from $20/mo to $200/mo in token spend. It is a reality. That forces the right questions IMO.
The AI Shift
AI isn’t replacing every job. But it is quietly changing how many jobs work.
Question….
What part of your job could AI assist or replace first?
Great article Sam. In the AI business, it is hugely important to know what you don’t know. In that sense the piece was extremely informative highlighting the real constraints in achieving sustained “SaaS” like margins.
I am curious to understand the evolving relationship between FTE reductions and the increase in AI infra spend.
Can you create great businesses by shrinking them?
Thanks again for a great article!
You probably can create great businesses by shrinking certain parts very skillfully.
Sam, great piece. You’ve shared some really deep insights here with plenty of nuances to think about.
From my own experience, every SaaS company I interact with is constantly pivoting toward AI integrations. However, they seem solely focused on the new features, while the financial downsides and real infrastructure costs are almost never addressed. Everyone is chasing the hype, but the "Agentic Math" you described is exactly the reality check the industry needs right now.
Thanks for the write-up!
Thanks for the comment, Alexander. Yes, that's really the purpose of the piece. Because, speaking from experience, sometimes you build up organizational infrastructure (in the form of costs) that becomes very hard to unwind.