A single opening draws 3,000+ resumes, and manual first-pass screening takes days. MOI structures batches of PDF resumes, scores them against job rules with evidence, and completes a full pass in under an hour.
Aramco Gulf Operations Company is a wholly owned subsidiary of Saudi Aramco, founded in 2000 to manage the Kingdom of Saudi Arabia's petroleum interest in the offshore Partitioned Zone shared with Kuwait. Its operations cover four offshore fields-Khafji, Hout, Lulu and Dorra-alongside the Kuwait Gulf Oil Company in the jointly operated zone.
AGOC hires at volume. When a role opens, recruiters face résumés arriving from different channels in different formats and different languages — PDF, Word, scans and images; some cleanly laid out, many not. A popular role routinely draws several hundred applications.
The problem was never too few résumés. It was that they are too varied, too scattered, and too hard to compare. What consumed the team was not the final hire-or-not judgement but the long, repetitive, error-prone stretch of reading and organizing that comes before it.
Four things went wrong. First, résumés resist structuring: education, years of experience, skills, projects and past employers sit in different sections and layouts, and AGOC's pool spans both Arabic and English, so extracting them by hand is slow and lossy.
Second, the information cannot be compared. Every résumé has its own structure, so they never land in a single view. What a recruiter lacks when ranking a batch is not résumés but a comparable structured result.
Third, screening is inconsistent. Two recruiters filtering the same role weigh things differently; the same recruiter drifts between a quiet day and a busy one. Fourth, the output is slow and does not carry forward — the criteria, scoring conventions and field mappings used in one round are hard to retain, so the next hire starts over. At AGOC's volume, screening capacity rather than candidate supply set the pace of recruitment.
MOI's starting point was not a chatty recruiting assistant but a decomposition of screening into concrete business steps, reassembled into one continuous chain.
First, unstructured résumés become structured input. Résumés are pulled automatically from their source channels and parsed with multi-modal document understanding into education, tenure, skills, certifications and project history along a common set of fields — whatever the file format, and in Arabic or English alike. Extracted fields can be corrected by a person, so scoring rests on facts that are actually right.
Second, the role and the résumés are brought onto the same analytical track. The job description is parsed into an explicit set of screening criteria — required skills, years of experience, education threshold, preferred extras — and each résumé's structured fields are aligned against it, producing a view built for this role and directly comparable across candidates.
There is a deliberate trade-off here: this was not built as free-form Q&A. Open natural-language questioning over résumés looks more impressive, but in high-volume screening that someone has to stand behind, it brings inconsistency, unstable accuracy and slow response. Turning the role into explicit scoring criteria and scoring every résumé the same way is more controllable, more explainable and more reusable.
The agent retrieves the candidate pool by hybrid search and scores each candidate across the role's criteria — structured, repeatable scoring rather than an answer that varies each time — returning a ranking with per-criterion scores and the résumé evidence behind them.
The agent processes around 2,000 résumés per hour — roughly 100× manual screening throughput. Retrieval and scoring run end to end in one agent, so recruitment moves at the speed of interviewing rather than of reading.
Recruiters see a ranked shortlist rather than a pile: candidates ordered by score under each role, with the scoring basis, key strengths and risk flags attached, ready either for human review or for export into the downstream recruiting system.
With field extraction, scoring conventions and ranking rules standardized, outcomes no longer depend on an individual recruiter's experience or state of mind — the same criteria apply to every résumé.
The reasoning is also visible: why this candidate ranks where they do, which requirements they meet, where they fall short, and which part of the résumé each score came from.
The lasting value is not only the thousands of résumés nobody had to read, but that a process previously carried by individual experience is turning into platform, rule and data capability that the next hire inherits.