Predictive maintenance and field service automation
Machine alerts that book the service visit before the line stops
An OEM with 2,000 machines in the field had telemetry it never acted on. Alerts went to a shared inbox, technicians were dispatched reactively, and customers called first. We built a workflow that reads machine data, predicts failures, and schedules the right technician with the right parts before the customer notices.
- Client
- Packaging machinery OEM
- Region
- Italy and United States
- Scale
- 500 to 800 employees
- Timeline
- 10 weeks, rolled out region by region
Representative engagement. Client details are anonymised and figures reflect typical outcomes from comparable deployments.
What the client got
0%
Fewer unplanned stoppages
Across machines under monitoring in the first two quarters.
0%
First-time fix rate improvement
Parts are allocated before the technician leaves the depot.
0%
Growth in preventive contract revenue
Account managers sell against real risk data instead of a price list.
The bottlenecks
Bottleneck 1
Telemetry nobody watched
Machines streamed vibration, temperature and cycle-count data to a cloud platform. Threshold alerts emailed a shared inbox with hundreds of messages a day, most of them noise.
Bottleneck 2
Dispatch was a phone call
When a customer reported a fault, a coordinator rang technicians to find who was nearest and free, then rang the warehouse to confirm parts. First visits often ended with a parts order and a second trip.
Bottleneck 3
Service contracts were undersold
With no data on which customers had machines at risk, the sales team could not make a case for preventive contracts. Renewals were argued on price alone.
What we built
Step 1
Signal scoring
The workflow ingests telemetry events, applies failure-pattern rules built with the OEM’s engineers, and uses an LLM to summarise each at-risk machine in plain language: likely component, estimated time to failure, and recommended action.
Step 2
Automatic scheduling
At-risk machines generate a service ticket. The workflow matches technician skills, location and calendar, checks parts stock at the nearest depot, and proposes a visit slot to the customer by email with a one-click confirm.
Step 3
Technician brief
Before each visit the technician receives a mobile brief: machine history, the predicted fault, parts already allocated, and the last three service notes summarised.
Step 4
Renewal intelligence
Account managers get a monthly list of customers with rising risk scores and out-of-contract machines, with a drafted note on what preventive cover would have avoided.
The automation workflow
Alert to scheduled visit flow
Telemetry to confirmed service appointment
Questions clients ask about this engagement
Does this replace the existing IoT platform?
No. It sits on top of it, consuming the events the platform already emits and turning them into decisions and bookings.
How accurate are the failure predictions?
Rules were calibrated against two years of service history. Predictions carry a confidence score, and low-confidence alerts are reviewed by a service engineer rather than auto-scheduled.
More in Manufacturing & Supply Chain
Industry overview
Custom AI operations platform
One AI dashboard for service, support and production
A mid-size manufacturer ran production, field service and customer support on three disconnected systems. We built a single AI dashboard that reads from all three, flags issues before customers call, and drafts the responses the team used to write by hand.
70%
Less time per support ticket
Procurement workflow automation
Supplier RFQs and purchase orders without the inbox
Buyers were sending requests for quotation one email at a time, chasing suppliers by phone, and retyping prices into the ERP. We automated the full loop from RFQ to approved purchase order, with AI reading supplier replies in any format.
80%
Less buyer admin per RFQ
Quote-to-order automation
Quotes for custom machined parts in 20 minutes instead of 3 days
A job shop quoting 60 to 80 custom parts a week was losing orders to faster competitors. Every RFQ needed an estimator to open the drawing, check material, count operations and price it. We built an AI quoting assistant that reads drawings and specs, drafts the quote, and lets the estimator approve in minutes.
20 min
Median time to quote
B2B outbound engine
Opening 14 new OEM accounts for a contract manufacturer in nine months
A contract manufacturer with two anchor customers wanted to diversify into European and North American OEMs. Trade shows produced business cards, not conversations. We built an account-based outbound engine that reached engineering and sourcing leaders with specific, relevant messages and handed sales a calendar of qualified meetings.
14
New OEM accounts opened
Have a similar bottleneck?
Tell us how the work happens today. We will show you the workflow we would build and prototype it on your own process.