Skip to content
SME AI Impact

Research gaps

What this record does not know

A case-study library that hides its gaps is a marketing document. These are the specific things this research could not establish, listed by category rather than smoothed over.

Practitioner BenchmarkClosing the Transparency Gap

Demonstrating what public award citations leave out

The 30 winners cannot exceed what IMDA and SBF published in their official citations — leaving all 30 cases without paired before-and-after baselines and 60% without named foundation models or vendors. In contrast, our Practitioner Cases (Maple Commerce implementations) set the documentation standard: complete model selections, verified architecture diagrams, paired operational baselines, and production telemetry are published openly with full conflict disclosures.

Winners without independent corroboration13

These cases rest on the official award citation and, in most instances, the company's own website. No third-party source addressing the implementation was located.

Undisclosed technology vendors19

The case study describes what the system does, but no source names a model provider, platform or cloud host. No vendor has been guessed.

Missing baseline metrics29

The citations publish relative changes, such as a percentage reduction or a multiple, without the underlying before-and-after values. No improvement figure has been calculated from an assumed baseline.

All 29 case studies fall into this category. Where a citation reports an absolute volume, such as manhours redirected or sites processed, that figure is shown as published. Where it reports only a percentage, only the percentage is shown.

Unverified or thin implementation timelines0

Only the award announcement date is documented. The citation does not state when the implementation began, how long it took, or when the reported results were measured.

    Insufficient detail for technical inference29

    For these cases the public record does not describe the workflow precisely enough to justify even a labelled inference. No inference section appears.

    Method limitations4

    These limits apply to the whole project.

    • The awards were announced on 13 October 2026. Research was completed five days later, so independent coverage of individual winners is sparse.
    • General web search returned a high proportion of irrelevant results for small-company queries. Company sites were fetched directly where a domain could be identified, which biases coverage toward companies with findable websites.
    • No winner was contacted. No company was invited to confirm, correct or expand its case study.
    • Reported outcomes are company-reported and were verified by SBF site visits during the award process. This project did not independently audit any figure.