Contents · 7 min read
01 / 05 · The choice
Should AI change what your customers buy?
In short
AI may let you sell uptime or output alongside the machine. That move makes sense only when you can stand behind the result with proof and action.
In your next offer review, sales wants to put uptime beside the machine price. Service can show which response worked, but your leadership team still needs machine data that proves the result and contracts that allow its use. If either is missing, keep the outcome promise off the invoice.
That may add service value, but revenue per customer can stay flat. Putting AI features inside the hardware price does not test whether customers will pay for a result.
You do not need to stop selling machines. One outcome-priced service can sit beside the hardware sale while your team tests the new offer.
The wider case for AI in industrial companies starts with what you know about each customer site. AI can help your teams use that knowledge in the offer as well as in the product.
Before your sales team can price a result, your installed-base data must show whether it occurred.
02 / 05 · Measurement
Can your installed-base data prove the result?
In short
You cannot sell a result with confidence unless both sides can see whether it happened. Your installed-base data must back that answer before the promise reaches a contract.
Start with the data that shows how your machines perform at customer sites. Then check the installation records and service cases that explain why the result changed.
Your right to use the data matters as much as the data itself. Existing service contracts may not give you access to the machine data an outcome offer needs. The B2B industrial product briefing shows why.
Audit your leading product families first. The manufacturing sales product briefing sets out that starting point. It then links customer calls to the rules used to fit products to each site.
The audit should leave your team with clear answers to three questions:
What result can you see?
Name the machine data that proves whether it occurred.
What explains the change?
Link the result to service work and site records.
What may you use?
Confirm that your company has the right to use the needed data.
Clear proof tells you what happened, but the next test is whether your company can change it.
03 / 05 · Control
Can your company influence the promised result?
In short
Measurement shows whether the result occurred, but your company must also be able to affect it. Otherwise the guarantee moves risk onto your P&L while the main causes stay beyond your control.
| Cannot measure | Can measure | |
|---|---|---|
| Cannot influence | Keep the machine model. | Use the data to learn, but do not guarantee. |
| Can influence | Build proof before pricing the result. | Run one protected contract. |
A live signal is useful because it tells your team when performance has changed. But the signal alone does not show that your company can fix the cause.
Your product may shape part of the result while the customer site shapes the rest. If your teams cannot act on the main causes, the promise gives the customer cover without giving you enough control.
Past work can show where your control begins and ends. Installation records reveal site issues, while service cases show which response worked. Customer calls add what the buyer was trying to achieve.
That knowledge is more useful when it stays with your company. It helps service staff act sooner and helps product teams remove causes that keep returning.
Once you can measure and influence the result, those teams must shape the promise sales makes.
04 / 05 · Team work
How should product and service shape the promise?
In short
Product and service must turn site evidence into a promise sales can defend. The offer should cover only what your company has seen and can change.
Sales knows what the customer wants, while service sees what fails after the sale. Product sees patterns across the installed base, but those facts often sit in separate systems or in someone's memory.
That split is common because customer work spans sales and delivery as well as support. The record may show what happened without showing why one response worked.
The AI-augmented customer service and success paper sets out a weekly flow for joining the work. Product receives themes tied to source calls. Sales gets proof it can use, while operations gets failures tied to installations.
Use that flow for your offer:
Service brings the cases
Show where the result held and where it failed. Add what the team did next.
Product finds the pattern
Check which causes return and which ones your company can change.
Sales tests the promise
Ask whether the customer will pay for the measured result.
Finance checks the risk
Put the expected income and cost of failure into one test.
Leadership narrows the offer
Promise only what your company can measure and influence.
This gives sales a sound basis for the offer, but only a paid contract can test the model.
05 / 05 · First contract
What is the smallest commercial test worth running?
In short
Run one service contract priced by outcome on its own protected P&L. Expand only when the customer result and the contract economics support the next test.
A broad launch may feel faster, but it would mix many sites and promises before your team knows what works. One contract keeps the result clear and the risk visible.
The industrial automation product briefing supports this starting point: one outcome-priced service contract shaped with the CFO and commercial leader.
Run it in five steps:
The commercial leader picks one result
Choose something the customer values and your company can measure.
Product and service show your control
Use site records and past service work to show what your teams can change.
The CFO protects the test
Keep its income and costs visible on their own P&L.
The team reviews the evidence
Check the customer result and the contract economics before you widen the offer.
Management decides what follows
Add another contract only when the evidence supports more risk.
One contract is enough to test the offer, but it does not show that the guarantee itself makes more customers stay.
The contract must cover measurement disputes and liability. Both need separate legal and finance work. Your team should set them for the result you promise and the control you hold.
That first review tells your leadership team whether to stop or offer the model more widely.
Test the first promise
Your leadership team now has to decide whether one customer result is ready for a protected contract test, based on whether your company can measure it and influence its main causes. The manufacturing sales CEO briefing shows how customer knowledge can stay with the company and support a stronger commercial model. Use it to examine the installed-base evidence behind one result before sales puts an outcome promise into a contract.
What AI changes in manufacturing, written for the CEO. Free to read, no form.
Questions CEOs ask about this
Should a manufacturing company use AI to move from selling machines to selling uptime, output or other customer outcomes?
Keep selling machines, and add an outcome-priced service only when you can measure the customer result and change its main causes. If you cannot do both, a guarantee shifts risk onto your P&L without giving your team enough control. Your decision is whether one result is ready for a protected contract test.
Which installed-base data is required?
Start with the data that shows whether the promised result occurred. Add installation records and service cases that explain changes in performance, then confirm your right to use the data.
When do outcome guarantees create too much risk?
The risk is too high when your company cannot measure the result or change the main causes. Detailed terms for disputes and liability depend on the contract, so the test cannot replace legal and finance review.
How should service and product teams work together?
Service should bring source-linked cases from customer sites, while product looks for causes that return. Sales can then base its promise on what those teams have seen and can change.
Can the model improve recurring revenue and customer retention?
It gives you a way to test recurring service income, and service work can help protect repeat sales. But the evidence does not prove that the guarantee itself raises retention, so compare results before expanding.
What is the smallest commercial test worth running?
Run one service contract priced by outcome on its own protected P&L. Have the CFO and commercial leader shape it, then expand only when its results support another contract.
Founder, GRAIL
Johan has spent fifteen years in management consulting and executive leadership, including as Chief Product Officer of a SaaS company. He founded GRAIL to help European leadership teams work out what AI changes at the core of their business. He is the author of The Execution Revolution and the host of the ThinkRoom podcast.
Related reading
Sources
- AI for Industrial Companies: why knowledge from customer sites can shape the commercial use of AI. (accessed October 29, 2026)
- B2B industrial CTO and CPO briefing: why machine-data rights and installed-base depth matter. (accessed October 29, 2026)
- Manufacturing sales CTO and CPO briefing: how to audit installed-base data and connect customer calls to product rules. (accessed October 29, 2026)
- What does an AI-augmented customer service and success team look like?: how evidence can move each week between service and product teams. (accessed October 29, 2026)
- Industrial automation CTO and CPO briefing: why the first test should be one protected outcome-priced service contract. (accessed October 29, 2026)
- Selling Smarter, Not Bigger: why customer knowledge should stay with the company rather than one seller. (accessed October 29, 2026)
Next scheduled review: February 16, 2027