A vision system catches defects on one line. The demo goes well, and everyone agrees it should go to every line. A year later it is still running where it started.
The World Economic Forum calls this pilot purgatory: more than 70% of companies investing in advanced analytics, AI or digital solutions never get past the pilot phase.
A successful pilot proves the idea. It does not prove the plant is ready to scale it.
Many barriers sit beyond the model
Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept, and gave four reasons:
- Poor data quality — uncleaned sensor data, and tags that differ between lines.
- Inadequate risk controls — nobody owns model risk or checks for drift.
- Escalating costs — every new line needs its own pipeline.
- Unclear business value — no baseline KPI to prove the gain.
None of the four is about the algorithm. BCG puts 70% of AI implementation challenges in people and processes, and only 10% in algorithms.
Plenty of pilots, very little at scale
95% of manufacturers have invested in, or plan to invest in, AI. Yet only 2% of manufacturing COOs told McKinsey that AI is fully embedded across their operations. The gaps are in the foundations: 56% report data issues, and 44% have unclear or no accountability for AI failures.
The money is coming. The question is whether it lands on a foundation or on another round of pilots.
The price of standing still
Unplanned downtime costs the world’s 500 largest companies about $1.4 trillion a year (Siemens). Plants that scale see the difference: the 2025 Global Lighthouse cohort reported 40% higher labour productivity. Gartner found that high-maturity organisations keep AI in production for three years or more at more than twice the rate of low-maturity ones.
Higher maturity does not guarantee returns. It is where sustained returns tend to start.
The next pilot is not the answer. The next foundation is.
Before the next budget round, count your pilots honestly, baseline the KPIs you would defend to the CFO, and test your best pilot against a rollout on paper. Then name an owner for model risk, and get an independent readiness assessment before you commit the money.
Read the full guide
Stuck on Line One is a 20-page guide drawing on 29 sources, and the first in INCIT’s Industrial AI Insight Series.
A pilot tests an idea. Readiness assessment helps identify the capabilities needed to scale it.
Official AIMRI Assessments are conducted by Certified AIMRI Assessors. To scope one for your plant, contact INCIT.
