Start HerePUBLIC EVIDENCE REGISTER / REVIEWED 2026-07-20
Know what is proven.
Know what is not.
Nexius Labs separates measured outcomes, attributed client testimony, engagement context, and illustrative examples so every claim can be interpreted at the right level.
Review the evidence01 CLASSIFY / MEASURE / DISCLOSE
THE PUBLIC REGISTER
Every public claim,
with its limits visible.
These records state the evidence basis and the boundary of what a reader may reasonably conclude.
View machine-readable registerAI handles repetitive bookkeeping work end to end
“We went from being constrained by manual bookkeeping throughput to a workflow where AI handles the repetitive work end-to-end.”
- Attribution
- Kate Yap, Co-founder, Daisy Accounting
- Evidence basis
- Named client testimony published with attribution.
- Limitation
- No quantitative baseline, sample size, observation period, or delivery timeframe is publicly claimed.
300+ employees / Multi-country operations
Presence in 8 countries. Engagement focus included ai literacy, role-based fluency, shared governance.
- Attribution
- Nexius Labs engagement record
- Evidence basis
- Organisation size, footprint, and sector supplied from Nexius Labs engagement records.
- Limitation
- The client identity is confidential. Focus areas describe engagement context and are not outcome claims.
120 employees / Construction materials
Import, export, and distribution. Engagement focus included ai capability, workflow mapping, trade & operations.
- Attribution
- Nexius Labs engagement record
- Evidence basis
- Organisation size, footprint, and sector supplied from Nexius Labs engagement records.
- Limitation
- The client identity is confidential. Focus areas describe engagement context and are not outcome claims.
20 employees / Automotive parts
Import, export, and distribution. Engagement focus included ai fluency, commercial workflows, logistics.
- Attribution
- Nexius Labs engagement record
- Evidence basis
- Organisation size, footprint, and sector supplied from Nexius Labs engagement records.
- Limitation
- The client identity is confidential. Focus areas describe engagement context and are not outcome claims.
10+ employees / Education and training
Singapore. Engagement focus included ai fluency, customer workflows, admin productivity.
- Attribution
- Nexius Labs engagement record
- Evidence basis
- Organisation size, footprint, and sector supplied from Nexius Labs engagement records.
- Limitation
- The client identity is confidential. Focus areas describe engagement context and are not outcome claims.
80 employees / Engineering
Singapore. Engagement focus included role-based fluency, workflow redesign, project support.
- Attribution
- Nexius Labs engagement record
- Evidence basis
- Organisation size, footprint, and sector supplied from Nexius Labs engagement records.
- Limitation
- The client identity is confidential. Focus areas describe engagement context and are not outcome claims.
EVIDENCE CLASSIFICATION
Four labels.
No blurred claims.
An illustration can explain an idea, but it cannot prove a result. A client statement can describe change, but it should not be converted into an invented percentage.
Measured outcome
A quantified result with a stated baseline, method, sample or volume, observation period, and limitations.
No public records yetAttributed testimony
A named client statement describing an observed change without presenting unpublished numbers as measured proof.
1 public recordEngagement context
Factual organisation and engagement information that establishes relevance but does not claim an outcome.
5 public recordsIllustrative example
A clearly labelled scenario used to explain a design or measurement concept. It is never presented as a client result.
Excluded from proof countsMEASUREMENT METHOD
Measure the whole
operating outcome.
Speed alone is not enough. A credible agentic AI result includes quality, exceptions, human effort, operating cost, and the conditions under which the result was observed.
- 01
Define the unit
Name the task, accepted output, accountable owner, and the boundary between one completed case and the next.
- 02
Record the baseline
Measure representative manual cases before intervention and state the sample size, period, and operating conditions.
- 03
Separate time measures
Report active human time, agent runtime, elapsed turnaround, and review time separately rather than combining them.
- 04
Test quality and exceptions
Track acceptance, rework, error, exception, escalation, and human-override rates alongside speed.
- 05
Measure operating cost
Include model, token, tool, infrastructure, retry, review, and correction costs per accepted outcome.
- 06
Publish limitations
State what was observed, what was not measured, where the result may not transfer, and who supplied or checked the data.
DISCLOSURE RULES
What Nexius will
and will not publish.
- Evidence before estimatesMeasured figures need a baseline, observation period, method, and limitations.
- Confidentiality remains explicitAnonymised contexts are labelled and never presented as named endorsements.
- Illustrations stay illustrativeDesign examples such as “four hours to seven minutes” remain excluded from proof counts until replaced by measured data.
- Corrections remain possibleEvidence records carry identifiers and review dates so material changes can be disclosed.
FROM CLAIM TO MEASUREMENT