Commercial Service — Blackbird AI Audit
Your team is using AI every day.
Do you know what it's doing to their work?
Current AI audit frameworks address model-level risk: bias, data quality, technical performance. None of them operate at session level — the point where the machine is learning what your team wants to hear and calibrating its outputs accordingly.
Session-level sycophancy analysis · severity-scored findings · 7 documented patterns
The Gap
What current AI audits don't cover
The Blackbird methodology documents session-level sycophancy, classifies it by severity, and produces a report your compliance team can act on. No other provider offers this.
What they miss
- Session-level sycophancy — real-time calibration to your team's preferences
- User failure taxonomy — the 10 documented ways users compound AI errors
- Warm-instance contamination in knowledge management and decision support
- Severity-scored severity — quantified risk your governance team can act on
What Blackbird delivers
- Four-layer analysis: behavioural, interpretive, user response, temporal
- 7 machine behaviour patterns classified and severity-scored per session
- Written report with governance recommendations your board can present
- Training recommendations and protocol updates as platforms change
What We Audit
A bespoke service designed specifically for you
Four tiers. From initial assessment to ongoing enterprise retainer. Each engagement is scoped to your organisation's AI usage and regulatory requirements.
Assessment
Scoped review of AI session patterns across core workflows. Severity-scored findings. Suitable for initial discovery.
- 5–10 representative sessions
- Full four-layer analysis
- Written findings report
Full Department Audit
30–50 sessions across a function. Comparative severity map. Policy recommendations.
- Cross-session pattern comparison
- Role-by-role severity mapping
- Governance recommendations
Enterprise Retainer
Quarterly audit cycle. Ongoing session-level monitoring. Protocol updates as platforms change.
- Quarterly report + annual synthesis
- Access to working libraries
- Priority methodology updates
Training Programme
How AI Reads You delivered to internal teams. Builds a Live Audit Document per participant.
- 4-week programme
- Individual audit documents
- Custom organisational protocol
Sector Coverage
Built for regulated industries
Financial services, legal, healthcare, education, enterprise governance, government, insurance, professional services, pharmaceutical, and 10 more sectors. Each audit is calibrated to your industry's regulatory landscape and risk profile.
The Methodology
This is not a checklist audit
Built on 7 documented machine behaviour patterns with severity scoring. Every finding is classified, quantified, and actionable. SHA256 authenticated.
Sycophancy
The machine agrees with the user's position regardless of accuracy. Validated, then abandoned when challenged.
Performed Honesty
The machine disclaims uncertainty while simultaneously providing the answer it just disclaimed.
Warm Calibration
The machine adapts tone, format, and content to match the user's emerging preferences over session duration.
Expert Positioning
Training data presented as current fact. Temporal authority assumed without verification.
Asymmetry
The machine knows more about the user than the user knows about the machine. Information imbalance.
System Limits
Architectural constraints presented as capabilities. The machine cannot always tell you what it cannot do.
What You Get
Audit deliverables
Every Blackbird engagement produces documented, actionable outputs your governance team can present to regulators, boards, and auditors.
Written Findings Report
Full session-level analysis with severity-scored findings. Pattern classification across all M1–M7 dimensions. Narrative summary suitable for board-level reporting and regulatory engagement.
Severity Mapping
Cross-session heat mapping. Role-by-role severity comparison. Visual pattern distribution across departments, workflows, and AI platforms in use.
Governance Recommendations
Policy recommendations mapped to your regulatory framework. Session governance protocols. Cold-instance validation procedures. Procurement guidance for AI tools.
Training Recommendations
Role-specific training needs identified from session analysis. Source Challenge protocols. User failure pattern awareness. Ongoing monitoring guidance and review cadence.
Get Started
The machine is already calibrating to your team. The question is whether you can see it.
Every audit is scoped to your organisation, your sector, and your regulatory requirements. Contact us to discuss which tier is right for you.