Kill every AI pilot that can't show ROI in 90 days?
88% of AI pilots fail to reach production per CIO May 2026. MIT NANDA: 95% see no P&L impact. 71% of CIOs expect budget cuts if mid-2026 targets are missed.
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?
The premise
- Team
- ~50 engineers, ~10 actively building AI features, single MLOps engineer. AI work pulls from feature-shipping capacity — any new commitment has to trade against the roadmap. ~6 active AI pilots split across product, ops, sales-enablement. Two pilots each consume ~30% of one engineer's time.
- Compliance
- SOC2 Type II in scope. EU customer data subjects us to GDPR plus the EU AI Act's August 2026 GPAI-deployer obligations.
- Stack
- Active pilots: (1) sales-prep AI assistant, (2) AI customer-support deflection, (3) AI-generated marketing copy, (4) AI-assisted code review for security, (5) internal HR-policy Q&A bot, (6) AI-driven anomaly detection in product logs. (2) and (6) have measurable outcomes; (1), (3), (5) don't yet. (4) is mid-pilot.
- Budget
- Monthly AI spend ~$30K with quarterly board visibility. Approvals required for sustained jumps >20%. Cost-per-outcome metrics in place; finance asks for unit economics by use case. ~$8K/mo of total AI spend goes to pilot infrastructure (not yet attributed to specific business outcomes).
What concretely counts as 'showing ROI' inside 90 days?
Either a measurable downstream metric we already track moving in the right direction (deflection rate, MTTR, win rate, etc.), or a user-cost reduction the finance team will accept (hours saved × loaded cost, validated). Vibes and anecdotes don't count. New custom metrics invented to justify the pilot don't count either.
Which pilots get a longer leash regardless of ROI status?
Only those with a regulatory or unavoidable obligation gate — e.g., AI-Act tooling, compliance evidence generation. Everything else, even 'strategically interesting,' is in the 90-day gate. Strategic exceptions tend to multiply if we allow them; one is the maximum.
How do we actually kill a pilot without it becoming a political fight?
Pre-commit to the kill criteria in writing at pilot start (we should have done this; we mostly didn't). For ongoing pilots: gate review at day 60 with named criteria; missed-gate = sunset plan due in 2 weeks, no surprise. The framing is 'this didn't earn more time' not 'this failed' — same outcome, less collateral.
Counsel's position
Reject a universal 90-day ROI gate in favor of timeline-specific targets based on each pilot's value mechanism, but immediately decommission any initiative that cannot define a measurable business output.
Verdict
The verdict: Adopt the AI Value Matrix to set timeline-specific ROI gates.
Adopt the AI Value Matrix to set timeline-specific ROI gates
Given your sprawling portfolio of six active pilots, abandon the one-size-fits-all 90-day gate and categorize each pilot by its value mechanism to set realistic payback expectations.
Map every AI pilot to a specific business output metric
To justify your $30K monthly AI spend to the board, stop measuring input activities and rigorously track how each pilot moves a core business metric.
Read another verdict
- Use AI to flatten middle management this year?
- Stand up a FinOps practice for tokens and GPUs now?
- Replace customer support with AI — or avoid the Klarna outcome?
- Adopt MCP as our default agent-integration standard?
- Crack down on shadow AI, or sanction it with guardrails?
- Red-team our own AI agents before shipping them?
- Give every AI agent its own scoped identity before scaling?
- Adopt Microsoft Agent 365 as our agent control plane?