// THE PRODUCTION RECORD
Our own record, published in full.
The industry number is that 95% of enterprise AI pilots return nothing. Rather than quote it and move on, here is our own record: every AI and data system we have built, and what actually happened to it. It is a small dataset by design, 10 systems, and it is our own, not an industry survey. We publish the ones we refused and corrected alongside the ones that shipped, because that is the whole point.
Record current as of July 2026
8
In production
1
Refused
1
Corrected
| System | Sector | Outcome | What happened |
|---|---|---|---|
| Water-quality consumer product | Energy | In production | Shipped on the fifth iteration, after four approaches failed. Now processes 50,000+ calls a month. |
| Operations unification across five platforms | Insurance | In production | One operational picture in 12 weeks; the engagement is ongoing. |
| Property enrichment pipeline | Distribution | In production | 100,000+ properties enriched. We caught and reported 19% of our own spend as preventable waste. |
| Court intelligence pipeline | Legal | In production | 75% cost reduction; first data in 8 weeks. |
| Country-risk intelligence platform | Enterprise security | In production | Country risk tied to enterprise exposure, built on official government advisory sources; ongoing. |
| Daily customer-intelligence briefing suite | Operations | In production | Automated daily briefings generated from unified operational data. |
| Voice agent for live customer calls | Operations | In production | Handles live inbound customer calls. |
| Three predictive ML models | Multiple | Refused | Built and tested. Accuracy collapsed once we removed leaked labels and circular targets, so we kept them shadow-only and shipped none of them. |
| Edwin (AI memory system, our R&D) | Own product | In production | Provisional patent filed. Two of 35 experiments disproved our own hypotheses, and the filing discloses them. |
| Consumer marketplace directory | Own product | Corrected | Found much of our own inventory had been inferred by a language model rather than observed; corrected the public count from 1,290 to 906 genuinely bookable. |
Methodology
- In production means a system is processing real data in production use.
- Refused means we built and tested it, then chose not to ship it.
- Corrected means we found and fixed a material error in our own output.
- This is our own engagement record, not an industry survey, and the sample is small on purpose. Two security-assessment engagements are excluded pending review, so they do not appear here.