Calibration certificate reviews, read against the PO and your template before release.

Release review starts once the technician finishes. The signatory reads the certificate beside the work order, the customer's purchase order and any cover letter attached to it, and your template for that instrument type. They confirm the instrument is the one received, the data is complete, the standards were in date, the conditions are stated, and the uncertainty and scope lines say what they may. Anything off goes back before the certificate is signed.

In the proof of concept, AI does that first read on copies and writes a difference list for each certificate: the line, what the PO or template asks for, what the certificate shows, and where it found each. It signs nothing and changes nothing. Your reviewer sets each list beside the corrections your own release check made on the same certificates.

The uncertainty line has its own grammar. ILAC's policy on uncertainty in calibration wants the result as the value plus or minus its expanded uncertainty, both with units, and the coverage factor and coverage probability stated with a short explanatory note. The expanded uncertainty carries at most two significant digits, rounded only after every calculation is done. A2LA's December 2024 article on common certificate findings puts significant figures first, usually from a template that prints extra digits or keeps trailing zeros, which count. The reported value may never be smaller than the CMC on your scope, and A2LA also bars copying the CMC across as if it were the uncertainty of this calibration. ILAC's policy explains why: the reported figure has to include what the customer's own device contributes. Where coverage factors differ from item to item, NIST's SOP 1 wants them in their own column. The list marks each of these by line. It never recomputes a value.

A decision rule has to say how uncertainty counts. When a certificate states in or out of tolerance, the wording of the rule matters. A2LA's G136 guidance calls a rule described as simple acceptance with uncertainty not taken into account technically incorrect, because a decision rule by definition accounts for it; the statement has to say how, for example through a test uncertainty ratio constraint. Some customers add their own line: the cover letter the Frederick National Laboratory for Cancer Research attaches to its vendor calibration requests wants the test accuracy ratio printed on the certificate whenever the standard used falls short of the ratio it asks for. The list quotes the rule as printed and flags a missing ratio.

The scope line is narrower than the letterhead. Under A2LA's rules on referring to its accreditation, a claim is made only under the legal name shown on the scope, for activities the scope covers, at the locations it lists, and with the lab's A2LA certificate number. Work outside the scope may share a certificate only if it is clearly set apart. Results from a subcontracted lab can appear on an endorsed certificate only if that work is on your own scope, you take responsibility for it, the subcontractor has approved the excerpt and is itself accredited for it. Your reviewer will check each measured quantity and range against the scope you send.

The PO and its cover letter are a checklist of their own. That laboratory's equipment calibration SOP, revision 11, asks every vendor for at least three data points spread across a set share of the instrument's range, at least three measurements on each channel of a multi-channel pipette, both the as-received or as-found readings and the as-returned or as-left readings on the certificate, repairs approved before they are made and described on it, and a signature from a responsible person. A2LA's article adds two common misses: a certificate that shows the calibration date but not the date of issue, and test points that don't match the range the customer actually uses, such as one point where low, mid and high were needed. The list sets each certificate against the terms of the account it went to.

Standards, methods and conditions vary by discipline. Among the routes ILAC's traceability policy accepts for a lab's own standards is calibration by a national metrology institute whose service the CIPM arrangement covers, and it names the BIPM's key comparison database, not a logo on the certificate, as the place to confirm that. SOP 1 has the procedure cited by its full title and version date, with units and symbols reviewed against NIST's SP 811. NVLAP's handbook wants a force report to say whether it followed ASTM E74, ISO 376 or both, and a mass report to give the ranges of environmental readings taken. When conformity of upper-class weights is assessed, it requires a statement whenever magnetism, surface finish or density went unchecked, and it leaves the choice of class to the customer. Your reviewer will look for each standard's due date against the calibration date, too. These summaries are context for your reviewer, not legal or technical advice.

If your reviewer finds the lists met the success criteria, new certificates get the same first read for roughly 30 days before they reach a signatory.

Every certificate is still released by your authorized signatory, and corrections are made by your staff in your own system.

The side-by-side continues on live certificates. A second job waits for its own decision and its own price.