Industrials/Multi-Industry: Industrial AI: practical lessons from real-world implementations

Research evidence excerpt

Industrials/Multi-Industry: Industrial AI: practical lessons from real-world implementations

Accessible version

Industrials/Multi-Industry

Industrial AI: practical lessons from real-

world implementations

Industry Overview

Industrial AI implementations – not for the faint of heart 11 May 2026

We attended the ARC European Industry Leadership Forum in Barcelona on May 4-5, a Equity

leading conference for automation professionals. Artificial intelligence (AI) is at the top Americas

of the agenda for industrial leaders, but adoption remains challenging. Topics discussed Industrials/Multi-Industry

included 1) challenges with data quality, availability, and fragmentation within Andrew Obin

enterprises, 2) the role of human oversight within AI decision making, 3) domain Research Analyst

expertise as both a key requirement and a constraining factor, and 4) the advantages of BofAS+1 646 855 1817

incumbency for industrial software vendors. andrew.obin@bofa.com

David Ridley-Lane, CFA

Data quality/availability: an underappreciated bottleneck ResearchBofAS Analyst

According to participants, ~95% of AI proof-of-concept projects never scale. Several +1 646 855 2907

david.ridleylane@bofa.com

speakers cited data quality as the single biggest reason. While data can be “cleaned up”

Devin Leonard

for a small project, this labor-intensive process means the AI project becomes difficult Research Analyst

to scale. Companies store engineering data in unstructured formats (e.g., PDF, BofAS

+1 646 855 3698

spreadsheets) and across disconnected systems. We provide an example of process devin.leonard@bofa.com

industries on inside.

Humans need to be in the loop – for now

"Would you sign off on an AI-assisted safety-relevant decision?" The answer for most