The Fight Radiology Already Had
If you work in weld inspection and you're skeptical of AI reading radiographs, you're not wrong to be. You're just late to a fight that a nearly identical profession already had and already settled.
Diagnostic radiology is the same imaging discipline pointed at people instead of pipe: a trained, certified specialist reads an image and makes a call that carries real consequences if it's wrong. When AI-assisted image review started showing up in radiology, the reaction was the same one weld RT is having now — not curiosity, alarm. Radiologists had spent years training for exactly the skill a piece of software was now claiming to do. The obvious fear was that the job would shrink to "supervise the AI" and eventually disappear.
That's not what happened. What happened is AI became a second reader: it flags likely abnormalities, catches the thing a fatigued eye misses on the fortieth study of a shift, and the radiologist still reads the image and signs the report. The certification requirement didn't move. The liability didn't move. What moved was where a radiologist's time went — less of it on studies that were never going to be the hard call, more of it on the ones that actually needed a specialist's judgment.
Weld RT in Canada has the same structure, almost exactly. A CGSB Level II or III still has to make the call and sign it — CSA Z662 doesn't allow that to move to software, any more than a hospital would let an AI model sign a radiology report. What's up for grabs is only the part before the signature: who or what looks at the image first, and how much of a certified reader's day goes to films that were always going to come back clean.
The industries that fought this fight before weld inspection didn't end up with fewer specialists, smaller signatures, or lighter accountability. They ended up with specialists spending their certified judgment on the cases that actually needed it. That's the version of this HotPass AI is built toward — not instead of your Level III, in front of them.
Where this goes next: if this is the argument that actually lands with your team, the natural follow-up isn't "buy the AI reader" — it's "what does our shop's workflow look like once the queue behind the Level III gets shorter." That's a build conversation, and it's the one Enrisk is set up to have alongside the read itself.