Sources & context
Context for the September 2026 Imperium presentation. Source review updated 25 September 2026.
Imperium’s published indicators
The five figures on slide 5 reproduce indicators published on the Imperium website. They are company calculations from selected interviews, including targets, midpoint estimates and extrapolations. They are not measured portfolio averages or audited savings, and do not predict an individual installation’s results.
These indicators combine different interviews. They are not the results of the person pictured beside them. The 52% calculation combines Sean, Isaac and Alberto’s accounts; it is not Shanny’s reported metric.
| Published figure | Company calculation | Interpretation |
|---|
52%
Estimated time returned | Mean of 60% conditional time-return,45% midpoint automation estimate and 50% inferred from a doubling-workload target. | Combines estimates, a conditional outcome and a target; not a measured time-return average. |
3.1×
Estimated output uplift | Mean of 6× reported task-output midpoint,2× target,2.5× inferred capacity and 1.8× inferred capacity. | Different tasks and inferred capacity measures; not a measured average productivity uplift. |
7.5
Estimated hiring avoided | Mean of five planned roles avoided and a ten-hire account published by Imperium. | Counterfactual hiring requirements, not actual headcount reductions or a representative portfolio average. |
$175,000
Modelled annual payroll avoided | Mean of $150,000 annualized from a $12,500 monthly midpoint and $200,000 assigned to another account of avoided specialist costs. | The second account does not establish an annual period. This annualized company model is not a directly reported or audited annual saving. |
5.6×
Estimated task-speed uplift | Mean of 6× task-output,8.7× inferred elapsed-time compression and 2× workload target. | Combines output, elapsed time and a target; not a measured like-for-like speed benchmark. |
Machine-readable metric record · Client accounts
Individual client accounts
Elissa Doyle describes 25 creative concepts in a reported 30-minute task. Ben Barnes describes a four-week plan prepared in about four days. Shanny Poijes, Core Concept Living describes a revised office staffing plan and estimates $10,000–$15,000 in monthly avoided costs. These are client accounts with their own circumstances, not standardized measurements.
Further examples: GeoNet, Trident Treehouse, BDL and Imprint Genius. Continuum and property-company examples describe engagement scopes where stated, not independently audited financial results.
Product and availability
ARX interface imagery uses source-backed product-kit illustrations with fictional demonstration data. The mobile companion view follows the actual mobile layout; availability depends on deployment. The Chrome pairing view is a private preview. Connections, permissions, dashboards and actions vary by installation. Claude is the default model engine; native Codex is a private preview.
The 1 September 2026 catalog contains 1,467 entries, supporting “1,400+ integration options”. Native and MCP variants may be separate entries. This is not a count of unique vendors or connections active in each installation. Integration catalog · Operating system · Security and deployment.
Development and experience
The timeline separates company-account origins and architecture, dated engagement and catalog records, and planned October 2026 direction. Its rail summarizes development themes, not exact release dates. The selected logos include company and team experience, including earlier professional relationships. SAP, Aldi, Stefanini and Vysiion appear as earlier professional experience, not current ARX customer claims or endorsements.
The founder profiles follow the company leadership page. Adoption themes draw on the audit information and Archie Nash’s published company films.
Research reference
Slide 24 refers to Project NANDA’s preliminary July 2025 report, The GenAI Divide: State of AI in Business 2025, hosted by a third party. Its findings have sampling, measurement and observation-period limits. It does not establish a universal AI failure rate, endorse Imperium or measure Imperium’s results.
Photography and illustrations
The presentation uses published film frames, client interview frames and company photographs. Architectural B-roll is editorial material, not a claim about Imperium facilities. Slide 4 retains one generated architecture concept. View the current photography and illustration ledger.
September 25 refinements
The desktop and mobile interface views are source-backed ARX product-kit configurations with fictional demonstration data. The phone view is deployment-specific, and the Chrome extension remains a private preview. OpenAI-related native Codex support remains a private preview. DeepSeek is shown as a self-hosted model option to assess for an installation, not as evidence of generally available native ARX integration. Official local deployment reference.
The founder background is a founder-reported account of earlier software, data procurement, migration, enterprise resource planning, and commercial operations work. SAP and ALDI marks refer to founder-reported prior professional experience. No project detail, current Imperium customer relationship or endorsement is asserted. Educational affiliations are omitted.
The ontology plate illustrates sources, connected business objects and decisions. It is a conceptual operating model, not a screenshot or certification of a particular backend architecture. The progress rail groups development themes rather than exact feature release dates; October remains planned.
Only Earned photographs come from the published Correlating AI to Revenue presentation. No unverified event year is stated. Other real frames come from the client interviews and the published company film.