"We have several AI pilots and no results"
Lemur Media · author: Dominik Dębski · updated 28.10.2026
AI pilots rarely reach production if no success measure and no person deciding on scaling were set before they began. The solution is to choose 2–3 use cases with an estimated value and a decision gate after 6–8 weeks.
What are the symptoms?
- The pilot runs longer than planned, and no end date is set.
- Pilot results are presented as demos, with no data on savings or revenue.
- The use cases were chosen by IT or a vendor, without the business process owner.
- Integration with production data was postponed to a later stage.
What are the causes?
- 01No success measureWithout a value set before the start, any result can be called promising.
- 02No business ownerThe pilot belongs to IT or a vendor, and no business unit has a budget for roll-out.
- 03Data out of scopeThe pilot runs on a slice of data, and integration with production systems was not part of the plan.
| Headcount | Typical picture |
|---|---|
| 50–249 employees | The pilot is run by one person alongside their core duties, with no roll-out budget. |
| 250–999 employees | Several departments run similar pilots independently of each other. |
| 1,000+ employees | The pilot stalls on security, legal and procurement requirements. |
What questions should the board ask?
- What success measure was recorded before each pilot started?
- Who holds the roll-out budget once the pilot ends?
- How many pilots were started, and how many closed with a decision, in the last 12 months?
Which metrics should you check?
- Number of pilots in progress and their duration
- Share of pilots with a recorded success measure
- Cost of pilots relative to documented impact
Which mistakes should you avoid?
- Launching new pilots before deciding on the ongoing ones.
- Judging a pilot on user feedback with no impact data.
How does the solution work?
Methods: AI Use Case Portfolio + From Pilot to Scale. Lemur Compass sets the sequence of actions based on the company's development phase, scale, business model and profitability.
- 01Pilot inventoryScope, cost and owner of every pilot in progress.
- 02Value × feasibility assessment2–3 use cases remain; the rest are closed.
- 03Pilot charterSuccess measure, business owner, decision date.
- 04Decision gateScale, adjust or close.
Where should you start?
AI Use Case Portfolio Workshop (1 day) or Diagnostic (2–3 weeks).