Structural Shift Reshaping Industrials: AI Integration
-
noviembre 18, 2026
-
Industrials* have adopted artificial intelligence (“AI”) broadly but not deeply. The value sits in operations (asset uptime, inventory, spend and contracts), and imperfect data is no longer a reason to wait, because AI can now help fix the data while delivering the value.
As we noted in the first article in this series, AI is now the top value creation lever for industrials CFOs (51%), ahead of operational and cost optimization (45%), and the shift underway is about application.1
As our article on skilled labor scarcity showed, contractor AI use is concentrated in office functions, with craft execution barely represented.2 Many leaders fear AI will simply replace work. It does replace certain tasks, but adopted deliberately, its larger effect is to multiply what scarce, skilled people can do, a thread that runs through this article.
What the Market Is Telling Us
Four patterns explain why AI value is not yet showing up in industrials’ results, and where it could.
Adoption is broad but shallow. Roughly seven in ten manufacturers use generative AI tools, and nearly 90% plan to spend more on AI over the next two years, according to a Manufacturing Leadership Council survey.3 But use clusters in knowledge management (57%), while operational applications trail: 20% in predictive maintenance, 12% in supply chain and procurement and 7% in inventory management.4 Half of manufacturers surveyed have no metrics to measure what AI is doing for operations.5
Individual gains are not reaching the profit and loss statement. FTI Consulting’s 2026 Private Equity AI Radar shows the pattern in private equity (“PE”): 95% of funds report AI initiatives meeting or exceeding their original business cases, but only 17% say they are significantly exceeding them, and just 7% of portfolio companies have reached enterprise-scale deployment.6 In a May 2026 survey of more than 2,200 maintenance and operations leaders, 58% of teams were using AI, yet 79% saw unplanned downtime stay flat or rise over the past year.7
The same gap shows up inside individual companies.
Industrials are data-rich but fragmented. Contract data is scattered across an average of 24 systems.8 In our experience, sensor data is similarly locked in vendor-specific systems. Unsurprisingly, data issues are the top barrier to AI in manufacturing, cited by more than 60% of manufacturers surveyed and more than 20 percentage points ahead of any other challenge.9
Labor is no longer the main cost lever; process and execution inefficiency is. U.S. manufacturing employs about 12.7 million people today, versus 17.3 million in 2000.10, 11 After two decades of shrinking workforces, the bigger opportunity now sits in unplanned downtime, maintenance, consumables, excess inventory and contract leakage. Unplanned downtime still costs the Fortune Global 500 roughly 11% of revenue, or $1.4 trillion. Plants see fewer downtime incidents than in 2019 (25 a month versus 42), but in most sectors each one now costs more and takes longer to fix. An hour of downtime in automotive now costs twice what it did in 2019.12 Skilled capacity is the binding constraint: Manufacturers reported roughly 522,000 job openings in August 2026, up from 416,000 a year earlier.13
Because most unplanned downtime is ultimately a maintenance problem, this is where the labor constraint meets AI most directly. Hiring alone cannot close the skills gap, so leaders are funding tools that amplify the workforce they already have.14 Manufacturers surveyed agree: “more efficient use of workforce” was the operational benefit most often ranked as having the highest AI potential, ahead of predictive maintenance and asset uptime.15 The demand is for AI that multiplies the workforce rather than replaces it.
What This Means for Cost, Cash and Capacity
Value leaks from industrial AI programs in three places:
Spend that does not deliver measurable business impact. AI tools deployed at individual levels improve working experiences but leave the underlying cost base unchanged, or even higher once license and token costs are added. That spend can still build AI capability, but only inside a structured program: Time saved matters only when it is redeployed to constrained work or the process itself changes.
Pilots that do not scale. With no baseline, no owner and no financial metric, success cannot be proven and funding stalls. For the half of manufacturers with no metrics to measure what AI is doing for operations, this is the norm.16
Use that varies across sites and teams. Within one organization, some sites apply AI with rigor and shared standards while others use it for ad hoc requests. Results cannot be compared, repeated or scaled, so best practice never becomes standard practice.
The bottom line. With capital more expensive and more closely scrutinized, as described earlier in this series, AI spend must clear the same bar as any other investment. Operational use cases that move cash and cost can clear it. Across the AI use cases PE leaders were asked about in a recent FTI Consulting survey, 54% to 67% of responses reported return on investment above 10%, a rough proxy for the cost of capital.17 Broad productivity tools often cannot show that they clear the bar at all.
What Leaders Should Consider
Ask where the AI budget actually lands. What share of AI spend touches operating cost or cash, versus individual productivity? Who genuinely needs a general-purpose seat, and where would the same money do more embedded in an operational process?
Start before the data is clean. AI can now reconcile, classify and extract meaning from messy, siloed legacy data, making the cleanup part of the use case rather than a multi-year precondition. Top-performing PE funds stand apart for where and how they deploy AI, not for fund size or AI spend, and even Gartner, which expects organizations to abandon 60% of AI projects unsupported by AI-ready data, recommends aligning data to specific use cases rather than fixing everything first.18, 19 Fix the upstream processes that produce bad data alongside each use case, or the cleanup will have to be repeated.
Prioritize four use cases that pay back and tolerate imperfect data:
- Asset uptime and predictive maintenance. AI combines sensor, maintenance and log data across vendor silos to prioritize interventions and schedule scarce technicians, while giving them live troubleshooting support and AI-guided augmented and virtual reality for complex repairs. The value is in downtime, defects and maintenance spend, and it leaves behind a usable asset register and failure history. NIST research found that plants relying more on predictive than preventive maintenance had 18.5% less unplanned downtime and 87.3% fewer defects, within a U.S. maintenance cost-and-loss pool the same study puts at about $222 billion a year.20
- Inventory segmentation and working capital. AI segments SKUs and helps improve demand forecasts on imperfect enterprise resource planning data, releasing cash and lifting service levels while leaving cleaner material master and lead-time data. The prize is large: J.P. Morgan estimates about $707 billion of liquidity trapped in working capital across S&P 1500 companies.21 For an educational publisher that had to commit to inventory months before customers ordered, an FTI Consulting machine learning demand forecast came within 0.3% accuracy, helped cut obsolescence by more than half and freed about $1.5 million of working capital in the first year.
- Procurement and spend analytics. AI classifies spend, de-duplicates vendors and normalizes material data to surface consolidation and price-variance opportunities, leaving behind a clean spend cube and vendor master. In a recent industrial merger, the kind of portfolio action we examined in our article on portfolio realignment, two companies recorded comparable purchases under different supplier names, legal entities and category structures. A machine learning model matched differently written supplier names, with a generative AI review step resolving borderline cases instead of manual review. Across roughly $4 billion of combined spend, the approach reached about 90% accuracy and cut vendor classification for 9,000 vendors from 30 to 50-plus hours to minutes. It produced a single rationalized supplier spend profile without first harmonizing the two companies’ systems and helped surface approximately $70 million to $90 million in procurement synergies.
- Commercial contracts and revenue leakage. AI can extract pricing, escalators, rebates and obligations from contracts scattered across systems and email inboxes, then reconcile them against billing to recover leakage, leaving behind structured contract data. World Commerce & Contracting research finds that poor contracting erodes almost 9% of annual revenue on average, and 15% or more in complex industries.22 For an industrial client, a large language model-based tool extracted critical commercial terms from contracts in different formats and compared actual performance with entitlements, identifying at least 150 basis points of margin leakage on received revenue. What began as a pilot can run as an ongoing agent that protects margin at minimal infrastructure cost.
These four use cases are not a complete list. For software development, financial planning and analysis, and customer service, mature vendor AI tools already exist, so buy them. Industrials’ investment belongs in operations, where their proprietary data can unlock the most value. Toward this end, companies should:
- Build reusable data assets from each use case. Each use case should leave the data better than it found it. That is how infrastructure improves along the way without a multi-year program first. Avoid throwaway, one-off pipelines.
- Define the measure before building. Agree on the baseline, owner and financial metric upfront. The answer cannot be “we will see what it does.”
- Match ambition to data trust. Start where a person reviews the output and move toward automation only as the data earns trust.
Closing Thoughts
AI’s value in industrials will be measured by how deeply it is embedded where cost and cash actually sit. Done well, AI multiplies the workforce. It lets the same skilled people keep more assets running, more inventory moving and more contracts honored. That outcome is not automatic. Nearly half of manufacturers surveyed (47.4%) expect AI to reduce plant headcount by 2030—yet 51.2% have no dedicated budget for AI training and education.23 Leaders who treat AI as a force multiplier and fund the skills to use it will capture the larger effect.
* In this series, “industrials” refers to asset-intensive businesses that make, build, move or maintain physical goods, from manufacturers and distributors to contractors and equipment service operators.
Footnotes:
1: FTI Consulting, “2026 Global CFO Report,” Industrials Extract [Unpublished Raw Data] (2026).
2: FTI Consulting, “Structural Shift Reshaping Industrials: Skilled Labor Scarcity” (Sept. 16, 2026).
3: Brousell, David R., “Survey: GenAI Adoption Surges In Manufacturing,” Manufacturing Leadership Council, Questions 1 and 6 (Apr. 1, 2026).
4: Id. (Link) Question 3.
5: Id. (Link) Questions 10 and 21.
6: FTI Consulting, “2026 Private Equity AI Radar: Continued AI Acceleration and Impact,” Pages 3 and 4 (May 19, 2026).
7: “Factories Have Gone AI, Yet See Zero Downtime Relief,” Manufacturing Business Technology (May 5, 2026) (reporting MaintainX, State of Industrial Maintenance 2026).
8: World Commerce & Contracting, “Contract Management: An Overlooked Driver of Business Agility and Financial Performance,” (Aug. 2025).
9: Brousell, supra note 3, Questions 25 and 27.
10: U.S. Bureau of Labor Statistics, News Release, “The Employment Situation: September 2026” (Oct. 2, 2026).
11: U.S. Bureau of Labor Statistics, Monthly Labor Review, “Manufacturing wage and salary employment, 2000–20,” chart data (2021).
12: Siemens, “The True Cost of Downtime 2024,” Automation.com, Pages 3, 4, 8, 10 and 11 (June 28, 2024).
13: “US manufacturing jobs grow as broader hiring slows,” Manufacturing Today (Oct. 2026) (reporting U.S. Bureau of Labor Statistics Job Openings and Labor Turnover Survey data).
14: MaintainX, “The State of Industrial Maintenance 2026” (2026).
15: Brousell, supra note 3, Question 18.
16: Brousell, supra note 3, Questions 10 and 21.
17: FTI Consulting, supra note 6, Page 7.
18: FTI Consulting, “AI Is Creating a New Performance Tier in Private Equity” (June 4, 2026).
19: Gartner, Press Release, “Lack of AI-Ready Data Puts AI Projects at Risk” (Feb. 26, 2025).
20: Thomas, Douglas S. and Weiss, Brian A., “Maintenance Costs and Advanced Maintenance Techniques in Manufacturing Machinery: Survey and Analysis,” National Institute of Standards and Technology (Apr. 25, 2021).
21: J.P. Morgan, “Increasing efficiency: Working Capital Index 2024,” (August 2024).
22: World Commerce & Contracting, supra note 8.
23: Brousell, supra note 3, Questions 22 and 24.
Artículos relacionados
Publicado
noviembre 18, 2026
Contactos clave
Senior Managing Director
Senior Managing Director
Managing Director
Director