Paseer AI Center
Capability catalog
Shared capabilities consumed by products via a single API. Each capability has a scorecard, and failing any critical threshold blocks publishing or reverts it to non-write assistant mode. ENT-AI-046 ENT-AI-054 AI-API-002 NFR-REQ-003
Published v3
Executive summarization
Strategy, portfolio, risk and assessments ENT-AI-047
Assistantclaude-sonnet-5instructions v3
99.4%Attribution
0.4%Hallucination
3.8sResponse time (95%)
$0.05per request
Published v2
Arabic document reading
OCR OCR and understanding of official letters AI-COR-01
Assistantprivate-ar-llminstructions v5
99.8%Attribution
0.1%Hallucination
5.1sResponse (95%)
—Private
Canary 10%
Extract commitments and deadlines
From letters and minutes after user approval AI-PRIV-001
Write with approvalprivate-ar-llminstructions v2
100%Attribution
0%Hallucination
4.4sResponse (95%)
—Private
Reverted to assistant mode
Reply drafting
Drafts with attribution for every claim AI-COR-03
Assistantprivate-ar-llminstructions v7
96.9%Attribution
1.2%Hallucination
6.2sResponse (95%)
—Private
Published v1
Suggest KPIs, risks and actions
With a baseline statement ENT-AI-049
Write with approvalclaude-sonnet-5instructions v2
98.6%Attribution
0.6%Hallucination
5.9sResponse (95%)
$0.07per request
Published v1
Detect conflict, duplication and deviation
Between targets and initiatives ENT-AI-050
Signalclaude-opus-5instructions v1
98.1%Attribution
0.8%Hallucination
11.4sResponse (95%)
$0.29per request
Published v2
Draft reports and disclosures
With references to source records ENT-AI-051
Write with approvalclaude-sonnet-5instructions v4
100%Critical attribution
0.3%Hallucination
7.3sResponse (95%)
$0.11per request
Published v2
Semantic search in policies and manuals
Hybrid search with citations AI-REQ-004
Readembed-ar
0.91nDCG
0leakage
0.6sResponse (95%)
$0.001per request
Disabled
People performance analysis
High-impact use · mandatory human review AI-PRIV-002
Did not pass bias testing
—Attribution
—Hallucination
—Response (95%)
—cost