Kill every AI pilot that can't show ROI in 90 days?
95% of enterprise GenAI organizations see no measurable return, while agentic workflows cost 30 times more than simple chat prompts. Unplanned budget overruns are forcing project cuts and delays, making clear value generation critical.
The question
Our AI pilot portfolio is sprawling and most pilots have no defensible ROI metric. Do we enforce a hard 90-day ROI gate and cancel everything that misses it, or keep funding exploration on faith for another quarter?
Counsel's position
Implement a tiered funding model with phased gates and native ROI measurement, rather than a hard 90-day ROI gate.
Verdict
The verdict: Implement a tiered funding model with phased gates and native ROI measurement, rather than a hard 90-day ROI gate.
How the criteria decide
3 of 3 criteria resolved on cited evidence.
| Criterion | Favours | Evidence |
|---|---|---|
| AI pilot-portfolio ROI gating | Keep funding exploration | 95% of enterprise GenAI organizations see no measurable return 95% of organizations seeing no measurable return, and 5% of integrated pilots extracting millions in value. Artificial Intelligence on Medium Non-deterministic token consumption makes accurate AI budgeting nearly impossible budgeting for tokens and clearly defining when AI is going to help with a problem is a much more indeterminate task than using other kinds of technology. Surviving AI features rely on native measurement, not post-hoc reporting ROI evidence is generated as a byproduct of usage — cycle time, defect rate, release frequency — not collected through a separate reporting process that someone has to remember to run. |
| cost-per-outcome measurement for AI initiatives | Keep funding exploration | Agentic workflows cost 30 times more than simple chat prompts By April 2026, it had spent its entire annual AI coding budget. Every dollar. In four months. Uber’s CTO publicly disclosed the overrun. AI to ROI - By Ray Rike and Peter Buchanan Non-deterministic token consumption makes accurate AI budgeting nearly impossible budgeting for tokens and clearly defining when AI is going to help with a problem is a much more indeterminate task than using other kinds of technology. 55% of APAC companies delayed AI agents due to costs 55% of APAC companies had delayed or scaled back AI agent rollouts because operating costs began to exceed the value generated. |
| AI sunset / kill protocols | Enforce 90-day ROI gate | 95% of enterprise GenAI organizations see no measurable return 95% of organizations seeing no measurable return, and 5% of integrated pilots extracting millions in value. Artificial Intelligence on Medium 55% of APAC companies delayed AI agents due to costs 55% of APAC companies had delayed or scaled back AI agent rollouts because operating costs began to exceed the value generated. |
95% of enterprise GenAI organizations see no measurable return
The vast majority of enterprise AI pilots fail to impact the P&L, making strict operational requirements the dividing line between successful integration and costly rework.
Agentic workflows cost 30 times more than simple chat prompts
Uncapped token consumption from complex AI agents can exhaust annual budgets in months, necessitating immediate usage caps and workflow-level cost attribution.
Non-deterministic token consumption makes accurate AI budgeting nearly impossible
Because AI models consume tokens unpredictably based on architecture and randomness, organizations cannot safely write blank checks for exploratory usage.
Surviving AI features rely on native measurement, not post-hoc reporting
AI projects that treat measurement as an afterthought are being cut, while those that generate ROI evidence as a natural byproduct of usage secure ongoing funding.
55% of APAC companies delayed AI agents due to costs
As organizations move from simple chatbots to expensive agentic systems, operating costs are rapidly exceeding generated value, forcing widespread project rollbacks.
Read another verdict
- Start with a small test, or take on the whole process at once?
- Our competitors advertise AI and we don't — match them, or hold the line?
- Our people already put client files into ChatGPT — ban it, frame it, or supply a tool?
- Our most experienced person retires in two years — how do we keep what they know?
- We can't hire the experienced people we need — automate, train up, or outsource?
- Slow our EU AI Act prep now the deadline's moved to 2027?
- Use AI to flatten middle management this year?
- Let an AI agent act on its own — or keep a human in the loop?