What does enterprise AI actually cost?
Short answer: budget for the system, not the API call. The demo proves the task once; evals, guardrails, data pipelines and monitoring are the other 90 percent, and inference costs compound as usage grows.
A demo proves the task once. Practitioners put that at roughly 10 percent of the work; safety, data, evals and reliability are the other 90 (Thoughtworks, 2026).
The model licence is the cheap part. The cost is everything around it, and it doesn't scale linearly.
// the cost curve
relative cost · illustrative
- Demo + pilot
- The cheap part. A demo proves the model can do the task once. This is the slice everyone shows you.
- Production
- Where cost erupts. Evaluations, guardrails, monitoring, data pipelines, security, human oversight. The part nobody pitches.
- Scale
- Where it compounds. Inference is a recurring cost, and an agent that loops or calls tools burns far more than a single prompt.
›transcript
- What enterprise AI actually costs.
- The demo is about 10% of the work; safety, data and evals are the other 90 (Thoughtworks).
- The licence is cheap. The system is not.
- Run the free, interactive course at heimlandr.io/ai-reality-map.
// the deep dive
A demo proves the model can do the task once. Production means evals, guardrails, monitoring, data pipelines, security and human oversight, the unglamorous 90 percent. Then inference compounds: a multi-step agent resends its whole context on every call, so one task can burn roughly 10 to 100 times the tokens of a single chat turn (Gartner puts it at 5 to 30 times per task). Unit prices keep dropping, but Bain found that as token prices halved, usage grew about 450 percent, so the total bill still climbs. Budget for the system, not the API call. Gartner predicted in 2024 that at least 30 percent of generative AI projects would be abandoned after proof of concept by the end of 2025, naming escalating cost as one reason.
// chapter faq
How much does a custom enterprise AI system cost?
It depends on the workflow, but the shape is constant: the model licence is a small line, and most of the budget goes to integration, evals, data cleanup and running it. That is why we scope to one workflow first, and why any quote without those lines is incomplete.
Why do AI costs rise when token prices keep falling?
Because usage grows faster than prices fall. Bain found usage grew about 450 percent as token prices halved, and multi-step agents resend their whole context on every call, so one task can burn 10 to 100 times the tokens of a single chat turn.
What is the biggest hidden cost in AI projects?
The production cliff: evals, guardrails, monitoring, security and human oversight. It is roughly 90 percent of the work and almost never in the demo or the first quote.
Every figure in this chapter is sourced. The full source list lives on the main map. Open the map
This is one chapter of ten. The whole course is free.
The full map has the interactive tools, the 8 minute audio edition, the live layer and every source. And if you want it run against your own reality, that call is free too.