What Weld Inspection Can Steal From Radiology
Radiographic testing is, literally, an X-ray. So it's worth asking why weld inspection and medical radiology — same imaging physics, same "a trained human reads the image and makes a call" workflow — ended up in completely different places on AI adoption.
Medicine got there first, and not gently. Radiologists spent years hearing that AI was coming for their jobs. What actually happened: AI became a second reader. It flags the frames worth a closer look, catches things a tired eye missed on the fortieth study of a night shift, and the radiologist still makes the call and signs the report. Diagnostic radiology didn't lose its human sign-off requirement. It got a faster first pass.
Weld RT is set up almost identically. In Canada, CSA Z662 requires 100% RT or UT coverage on the highest-risk welds — road and rail crossings, water crossings, anything that won't be pressure-tested — plus a minimum 15% of every other production weld poured each day. Every one of those images still needs a CGSB-certified reader, and a CGSB Level III has to review and approve the procedure behind it. That sign-off requirement isn't going anywhere, and it shouldn't. It's the same shape as the radiologist's signature.
What's different is everything upstream of that signature. A pipeline contractor pouring welds across a multi-kilometre right-of-way generates a lot of film and digital radiographs, fast, often from remote sites, often on tight schedules. Someone certified still has to look at every one of them before the weld is accepted. That's the queue radiology solved with a second reader instead of asking every radiologist to read faster.
HotPass AI is built to be that second reader for weld radiographs — not a replacement for the CGSB Level II or III who signs off, but the pass that happens before it reaches them. It flags likely indications, surfaces the frames that need a closer look first, and leaves the call exactly where CSA Z662 already puts it.
The industries that fought this fight first didn't end up with fewer radiologists. They ended up with radiologists spending their time on the images that actually needed a human judgment call, instead of the ones that didn't.
Where this goes next: HotPass AI isn't only the read layer. Because it comes out of Enrisk — a firm that builds AI workflow automation, not just AI point-tools — the same engagement can cover the reporting, chain-of-custody, and intake automation around the read itself. If your bottleneck is the paperwork after the image lands as much as the read itself, that's a conversation worth having in the same call.