Electronic data deliverables, drafted from your signed reports in each client's format.

Today a data specialist exports results from the LIMS, maps them onto the client's template, fills in the codes the format demands, fixes whatever the checker rejects and sends the file with the report. Each format has its own columns, codes and rules, and the work starts over when a project moves to a new version.

In the proof of concept, AI reads copies of the signed report, your LIMS export and the client's format files, and drafts the EDD row by row: sample codes, methods, analytes with CAS numbers, results, units, limits and lab qualifiers, each carried from your report. Any row it can't place goes on a short list for your data group, with the reason. Nothing is loaded or sent.

The format belongs to the agency; the file usually belongs to the consultant. New York DEC takes remediation and materials management data in a standard EDD format built on EarthSoft's EQuIS, and says labs and subcontractors hand their data to the primary consultant, who prepares the EDD. DEC's checking tool is the EQuIS Data Processor, version 7.24.2 for the current format, and a common error DEC names is a value missing from the valid values list. A merge EDD goes through DEC's automated data loading; an update or replace EDD goes to DEC's EQuIS team to be reviewed and loaded by hand, and a file over 25 MB needs the team's help to split. So the draft uses the consultant's codes, not yours, and lists every value it couldn't find on the list.

Sample identity has rules of its own. DEC's EDD manual keeps the sample code the consultant or DEC assigns apart from your lab sample ID. A field duplicate or matrix spike must name the sample it came from as its parent, trip blanks and field blanks carry no location, waste characterization samples are typed WC and clean fill CF, and characters such as # or @ stay out of sample names. The draft ties each row to the client sample ID on your report and flags any sample whose parent or type it can't pin down.

One reportable result per chemical, per sample. The same manual wants every initial run and every retest in the file, with exactly one result for each chemical in each sample marked reportable. Its example is a benzene result of 1,000 flagged E, over the calibration range and not reportable, followed by a dilution at 650 that is. Selected ion monitoring runs and second-column results raise the same question. The draft never makes that choice: it marks reportable the result your signed report issued, and lists any sample where the report leaves two candidates.

Non-detects and qualifiers have exact places. For a non-detect the result value stays empty, the detect flag reads N, and the sample-specific reporting limit carries the number, alongside the method detection limit and the quantitation limit. Labs fill in their own lab qualifiers; the interpreted qualifier is filled only when the lab itself submits to DEC, and the validation level stays blank in raw lab data. Fractions are coded total, dissolved, TCLP, SPLP or SEM, solids are reported wet or dry, and the lab's sample delivery group ties each row to the hard-copy report. The draft copies each qualifier exactly as your report prints it and never adds, drops or converts one.

Other receivers, other rules. EPA's Staged Electronic Data Deliverable is a hierarchical file created by a lab's LIMS, with stages that add sample-related QC results, then instrument QC, then instrument output detailed enough to recalculate results; its specification 5.2 dates from March 2019, and the Contract Laboratory Program's statement of work sets the stage. Virginia's Office of Drinking Water takes every lab result through EPA's Compliance Monitoring Data Portal, by Excel, XML or web entry, after at least three successful test files, while positive coliform results and MCL exceedances also go to a field office within 24 hours. That alert stays with your staff. Your reviewer will check codes against the valid values, parent links, reportable flags, limits and units, and that every row matches the signed report.

With a yes against the success criteria, new jobs get draft EDDs the same way for about 30 days.

Each file is checked and released by your data group before it goes to a consultant or an agency, and every load or submission stays with your team or the client.

The side-by-side scoring continues through the month. Adding another job waits until this one is judged, and it gets its own price.