For chief executives
Where AI belongs in a manufacturer’s P&L
Service is where industrial manufacturers make their best margins. A map of the lines in the profit and loss that technical support touches, and where AI can move them.
Lingo
5 min read
Most of what is written about AI in manufacturing is about the factory floor: vision systems, predictive maintenance, scheduling. The less glamorous opportunity is after the product has shipped, in the hours a manufacturer spends helping other people install and fix it. That is where the margin is.
In short
- Service is where industrial manufacturers make their best margins: McKinsey and Bain both put aftermarket margins well above new equipment.
- Technical support touches seven lines of the profit and loss account. A guide acts on two of them itself; the rest depend on what installers do next.
- The dependable use of AI here is a specific one: getting the documented answer to the installer and passing other questions to a person.
Service is the more profitable part of the business
Two of the large consultancies have published the gap between what manufacturers earn on new equipment and what they earn on service after the sale. They measure different margins on different samples, so the two panels below cannot be compared with each other. Each compares service with new equipment inside one study.
Margin on service against margin on new equipment
Two studies, two different margin measures. Compare within each panel, not across them
Bain, gross margin
McKinsey, EBIT margin
McKinsey’s 2023 view is blunter: margins on aftermarket services are almost always at least double the margins on new units. Bain found service was already 22% of revenue in its sample, and had grown at 9% a year from 2010 to 2013 against 5% for equipment.
For a manufacturer that sells through distributors and installers, a large part of “service” is not a contract at all. It is the technical helpline, the field engineer who sorts out what the installer could not, and the warranty claim when a product was set up wrongly. Those costs come straight off that margin.
Where technical support sits in the P&L
Seven lines, in three parts of the account. On two of them a guide does something itself. On the other five the effect depends on what installers do next.
Lines in the P&L that technical support touches
What a guide does on each line. A map, not a measurement
Revenue
Service contracts and partsDepends on what installers do next
An installer who gets the answer on site can finish the job. Whether that shows up in parts and contract revenue is yours to measure.
New product sales through installersDepends on what installers do next
Installers recommend the brand that answers. A published guide answers documented questions at any hour.
Cost of sales
Warranty claims caused by installationDepends on what installers do next
A guide puts your documented order in front of the installer. Whether claims fall is something to measure.
Return visits and no-fault-found partsDepends on what installers do next
A guide can take the installer through your documented checks before a part is condemned.
Operating costs
Technical support salariesWhat the product does
Documented questions can be answered from a published guide at any hour.
Engineering time spent on supportWhat the product does
Questions passed to a person arrive in the same conversation, with the earlier messages visible.
TrainingDepends on what installers do next
The procedure taught on the course can be sent to the installer on site afterwards.
What AI is actually good for here
The best evidence on AI in support work comes from a study of 5,179 customer support agents at one company, published by the US National Bureau of Economic Research. In the working paper, a generative AI assistant raised the number of issues resolved per hour by 14% on average. The effect was concentrated in novice and low-skilled staff, who improved by 34%, with minimal impact on the most experienced. That was a generative assistant helping agents at a software company, not a guide answering installers.
The tool did not make experts faster. It gave everyone else access to what the experts knew.
That is the same job a guide does for an installer network: the installer who has never fitted this model gets the answer your best engineer would give.
McKinsey’s 2023 estimate that generative AI could raise productivity in customer care by an amount worth 30 to 45% of the function’s costs is a modelled estimate, not a measurement. It is a ceiling to reason against, not a number to put in a budget.
What still needs people
AI does not fix a product that fails in the field, write the documentation you do not have, or replace the engineer who diagnoses the fault nobody has seen before. For safety-critical and regulated products, an AI that improvises an answer is a liability, not a saving; the companion article on deterministic and generative AI covers why.
The useful question for a chief executive is narrower than “what is our AI strategy”. It is: how many hours a year do our people spend repeating what our manuals already say, and what would they do with them instead? The cost build-up shows how to put a figure on those hours.
Sources
- Winning in industrial serviceBain and Company, August 2015. Survey of 45 European industrial manufacturers, 2014. New equipment margin shown in the report’s figure 1.
- Industrial aftermarket services: growing the coreMcKinsey and Company, July 2017.
- Aftermarket sales and service are vital to manufacturers’ strategiesMcKinsey and Company, March 2023.
- Generative AI at workBrynjolfsson, Li and Raymond, National Bureau of Economic Research working paper 31161, April 2023, revised November 2023. The version later published in the Quarterly Journal of Economics (2025) reports 5,172 agents and a 15% average gain.
- The economic potential of generative AI: the next productivity frontierMcKinsey Global Institute, June 2023.
Questions people ask
How can a manufacturer use AI to improve profit?
The most dependable place is service, where industrial manufacturers earn their best margins and where much of the work is answering documented questions. AI that takes installers through steps a manufacturer has approved can answer documented questions without taking an engineer off other work. Effects on return visits and warranty claims are plausible but need measuring in your own business.
Are aftermarket services more profitable than new equipment?
Consultancy research says so consistently. McKinsey reported an average EBIT margin of 25% for aftermarket services against 10% for new equipment across 30 industries (2017). Bain found service earned a 39% gross margin against 27% for new equipment in a survey of 45 European industrial manufacturers (2015). The two use different margin measures.
What does research say about AI in customer support productivity?
A study of 5,179 support agents published by the US National Bureau of Economic Research found a generative AI assistant raised issues resolved per hour by 14% on average, and by 34% for the least experienced agents, with little effect for the most experienced.
