How should the evidence in these AI case studies be read?

Metric value

The value shown on the corresponding public case page.

Measurement definition

What was counted, compared, or timed in the project validation snapshot.

Source and limitation

A canonical case URL and an explicit evidence boundary.

Not an independent audit

The values have not been certified by an external assurance provider.

Not a universal benchmark

Project-specific samples do not predict results for another organisation.

Not client identification

Commercially sensitive client identities, raw records, and confidential baselines remain unpublished.

Turning retail transaction data into operating decisions

• 51 · stores covered — Cross-store sales, inventory, and operating analysis
• 289 · legacy tables mapped — Fields, sources, and metric ownership documented
• 99.6% · reconciliation consistency — Daily sales and inventory metrics checked against source systems
• 77.2% · actionable inventory alerts — Alerts followed by replenishment, transfer, or stock-reduction action
• 21 min · median decision-data time — From a cross-store question to a reviewable answer

Evidence boundary: The published metrics come from project records and describe the delivered scope and pilot validation. They should not be extrapolated as expected results for another organisation.

An AI-native production line for accounting operations

• 12 · business domains modelled — Customer, invoice, bookkeeping, filing, and compliance operations
• 52 · screens and flows mapped — Interface operations translated into orchestrated production steps
• 9 · production workspaces — Batch work, exception takeover, and permission governance
• 72.8% · low-risk task automation — Tasks completed without rewriting and passed deterministic checks
• 99.2% · evidence-chain completeness — Critical judgements retain source, rule, version, and action record
• 4.7 h · average month-end cycle — From complete documents to a result ready for human review

Evidence boundary: The published figures separate mapped product scope from pilot workflow validation. They do not claim financial outcomes, and regulated decisions and official submissions remain subject to qualified human review.

Legal AI that returns every conclusion to its evidence

• 2 · operating chains unified — Law-firm operations and large-dossier analysis
• 4 · evidence coordinates — File, page, source text, and confidence
• 98.6% · citation-location accuracy — Lawyer sample matched citations to the correct material and page
• 92.4% · priority-material recall — Lawyer-labelled key facts entered the candidate set
• 11.8 h · case initial-review cycle — Effective work time to the first issue list; 18.5 h before

Evidence boundary: The validation figures come from a controlled project sample and lawyer-labelled reference set. They demonstrate the tested workflow, not legal accuracy for every jurisdiction or matter.

Recruiting automation with context, evidence, and send control

• 3 · automation modes — Manual, Copilot, and controlled Autopilot
• 4 · operating workspaces — Candidate queue, conversation, profile, and operations
• 98.9% · conversation-sync coverage — Visible historical messages successfully backfilled
• 99.3% · send reread success — Sent text reread from the recruiter side after action
• 15.3 min · candidate handling time — First review to confirmed next action; 22.4 min before

Evidence boundary: The published validation covers synchronisation, send verification, and workflow time in a controlled pilot. It does not claim improved hiring outcomes without a longer recruiting cohort.

A trusted data path from industrial operations to management decisions

• 4 · field processes connected — Weighbridge, laboratory, vehicle, and energy
• 1 · unified operating entry — ERP, mobile access, and industrial data
• 96.1% · field-data coverage — Critical shifts, equipment, and business events recorded
• 99.2% · on-time collection — Data arrived within the defined dashboard window
• 27 min · exception acknowledgement — Alert to accountable operator confirmation; 41 min before

Evidence boundary: Validation covers data-record completeness, collection timeliness, and operator acknowledgement in the delivered project scope. Safety-related actions remain under authorised site personnel.

Credit research automation without hiding the source

• 2 · research source classes — Internal material and current public financial information
• 3 · orchestrated stages — Retrieval, cross-checking, and report drafting
• 91.6% · priority-fact recall — Reference facts entered the evidence set
• 98.3% · conclusion citation coverage — Material report judgements retained a reviewable source
• 4.4 h · research first-draft cycle — Task creation to analyst-ready draft; 7.6 h before

Evidence boundary: The validation sample measures fact recall, citation coverage, and first-draft time. The system does not replace credit, investment, or trading decisions.