Embedding AI Into Legacy Hardware
Facilities and manufacturing · Embedded AI for devices
Thread AI Editorial
September 28, 2026
A Worker that a device maker ships with its fax and multifunction devices, classifying each page that lands and filing it in the customer's system of record.
Targeted Impact
Time to launch
Weeks from spec to a document feature running on devices already in the field
Turnaround time
From a fax landing to its fields in the customer's system of record
Manual keying
Retyping left for staff at the customer's site once pages are classified and filed
The challenge
Referrals, orders and records still arrive by fax in sectors such as manufacturing. The device captures the page, then someone at the customer's site reads it, decides where it goes and types the details into another system. It’s an inefficient process.
The solution
Fax machines already in the field become a front end for the latest frontier AI models. A faxed order that lands can open a purchase order in the ERP and message the account owner that it's there.
The device maker's team builds those Workers on Lemma and sells each one as a cutting edge feature of hardware its customers already own. Any action that writes to a customer's system waits for a person to approve it.
Capabilities
1
Classification on arrival.
Referrals, orders and everything else are sorted the moment the device captures
2
Reads and writes back to each customer's systems.
The customer's system of record or ERP is registered once as a Function, so each new integration is configuration work for the device maker.
3
Incorporate latest AI and agentic features into legacy systems.
Lemma allows the customer to push cutting edge features into existing products by deploying infrastructure that can power new agentic patterns.
4
A person for every uncertain page.
Anything below the customer's confidence threshold goes to a review form for human approval.