Insights · Evidence
Evidence, tiered by rigour.
Every statistic used across TRIAI's materials is sourced and tiered by evidentiary strength. Headline claims draw only from primary research. Directional figures are always labelled as such. This is the same discipline we bring to a client's own adoption claims.
Tier 1 · HeadlineTier 2 · Directional
Tier 1 · Headline
Headline-grade evidence
Primary research — safe to state as fact anywhere on the site when attributed.
| Claim | Source |
|---|---|
| 88% of organisations use AI in at least one function; only 39% report enterprise-level EBIT impact | McKinsey, The State of AI |
| 95% of GenAI pilots show no measurable P&L impact; ~5% reach production and create value | MIT NANDA, The GenAI Divide (2025) |
| AI governance/data incidents: financial loss in 77% of cases, reputational loss in 55%; 5.4% average annual EBITDA improvement where agentic AI scales properly | KPMG / Neurons Lab analysis |
Tier 2 · Directional
Directional evidence
| Claim | Source |
|---|---|
| 88% of AI agent projects fail to reach production | Bantech Solutions |
| ~171% average ROI for production-stage agentic deployments | Landbase |
| 36% higher AI ROI under centralised/hub-and-spoke operating models | IBM via AI Assembly Lines |
| 21% of organisations have mature governance for autonomous agents | Paul Okhrem |
| >40% of agentic AI projects to be cancelled by end-2027 (cost/unclear value) | Gartner (via compilation) |
| 72% report increased employee satisfaction post-agentic-AI | SoundHound AI |
| Time-to-ROI ranges from ~2 weeks (customer service) to 12+ months (supply chain) | Joget |
| Finance close redesigns report 30–50% faster cycle times | Joget / Master of Code |
| Documented supply-chain exception redesign savings of over $20M | AI Monk |
| Retail customer-operations redesign lifted NPS to 65 alongside a $77M gross profit gain | OneReach.ai |
Discipline
How TRIAI uses evidence
- The rule
- External headline claims use Tier 1 only. Tier 2 figures appear only with attribution and as directional context. Maximum three to four statistics per external asset.
- The honesty clause
- All figures are third-party benchmarks and documented case outcomes — illustrative ranges, never TRIAI performance guarantees. Actual outcomes depend on the specific process, data readiness, and how faithfully a build follows the agreed design — which is precisely why Implementation Assurance and Outcome Optimisation exist as ongoing disciplines within Scale, not one-off checks.
- The forward commitment
- As TRIAI completes engagements, our own anonymised evidence — baseline-to-outcome results, with clearly defined terms for production, adoption, and scale — will progressively replace Tier 2 references across our materials.
Sources are reviewed at least annually. External citations stay attributed to the original publisher.
See how this evidence shapes our approach.