The Verification Standard: What We Refuse to Send Until It's Proven

My job at Nexen is to be the person the rest of the team has to convince. Everyone loves an answer that arrives fast. I'm interested in whether it's true.

The gate

Before anything leaves this company — an estimate, a schedule commitment, a submittal response, a blog post like this one — it passes a verification gate. The rules are simple and non-negotiable:

• Every number traces to a source. Vendor quote, catalog page, installed-cost record, or measurement. "Industry average" is not a source.
• AI output is a claim, not a fact. It gets checked against the primary document by someone who didn't produce it.
• Confidence is stated, not implied. Verified items are marked verified. Assumptions are named as assumptions. Clients can tell the difference because we tell them.

Why this is rare

Speed is cheap now — anyone with a chatbot can produce a professional-looking answer in seconds, and that's precisely the problem. The industry's next round of expensive mistakes will be confident, well-formatted, machine-generated errors that nobody checked. Firms without a staffed verification role are shipping those errors right now and won't know it until an architect's reply makes it public.

The test we hold ourselves to

Would this survive a hostile reader — an auditor, an inspector, opposing counsel? If not, it doesn't ship. It's also why verification-first work styles score as strengths on the assessment every Nexen teammate takes. Skepticism isn't friction here. It's the product warranty.