Industrial customers often arrive at a part-replacement with a single piece of information: a photo. A faded label on a failed contactor inside a switchboard, a phone snap of an old motor-starter nameplate, a scanned drawing from a maintenance dossier. Until now, identifying that part still required typing out what you could read into a search box. From today, EPFINDER accepts the image directly.

How it works

Upload a photo through the EP-AI chat widget — or via the new “search by image” button on the catalogue page — and EPFINDER does three things:

1. Reads the visible text on the part (OCR). Manufacturer name, part number, electrical rating, certification marks, and any other label content are extracted.
2. Identifies the product family from the visual silhouette. Contactor vs MCCB vs MCB vs surge-protection vs terminal block — each has a distinctive form factor.
3. Cross-references against the EPFINDER catalogue for an in-stock match, ranked by closeness of the OCR-extracted attributes against the catalogue specification.

What this is good for

• Field replacements — a maintenance engineer needs to source a same-or-equivalent replacement for a part that’s already in service. Photo the nameplate; get the SKU.
• Legacy panels — older switchboards with hand-written tags or worn labels. The OCR extracts what’s legible; EP-AI’s equal-or-higher rule handles the rest.
• Drawing pages with spec callouts — paste a screenshot from a CAD drawing or a single-line diagram. EPFINDER picks the parts out of the annotation text.
• Procurement from suppliers’ technical PDFs — saves typing out tabular part lists from supplier datasheets.

Accuracy & limits

OCR works well on clean nameplates and recently-printed labels — typically 90%+ first-shot accuracy. For severely weathered or partially-obscured labels, EP-AI flags what it could and could not read, and asks one or two clarifying questions (“Is this a 25 A or 32 A version? I cannot read the current rating cleanly”). When the visual evidence is genuinely ambiguous, the assistant hands off to a sales engineer with the photo attached.

Privacy

Photos uploaded for search are processed for the matching task and then discarded. They are not added to the public catalogue, not shared with manufacturers, and not used to train external AI models.

Try it: take a photo of any part you need, drop it into EP-AI on the catalogue, and see what comes back.