Add Microsoft IQ account brief recipe with the Copilot harness

notebooks/data/copilot-harness-foundry-fabric-iq/customers.csv

@@ -0,0 +1,2 @@
+customer_id,customer_name,customer_domain,customer_tier,knowledge_base_count,index_count
+CUST-001,Adatum Corporation,adatum.com,Strategic,2,2

notebooks/data/copilot-harness-foundry-fabric-iq/knowledge-bases.csv

@@ -0,0 +1,3 @@
+knowledge_base_id,customer_id,knowledge_base_name
+KB-001-01,CUST-001,Product Support KB
+KB-001-02,CUST-001,Employee Knowledge Hub

notebooks/data/copilot-harness-foundry-fabric-iq/offline-fixture.json

@@ -0,0 +1,24 @@
+{
+  "provenance": "Fictional Customer 360 evaluation data. This answer is hand-authored for offline wiring tests, not a live response or model-quality evidence. Bracketed labels identify fixture records, not service-generated citations. Wrong-ID test inputs are mocks, not additional imported customers.",
+  "known_id": "CUST-001",
+  "known": {
+    "answer": "Customer: found. Adatum Corporation (CUST-001) has two recorded knowledge bases: Product Support KB (KB-001-01, idx-001-01) and Employee Knowledge Hub (KB-001-02, idx-001-02). Latest samples TS-009967 and TS-009924 on 2026-11-05 are labeled Healthy; those historical labels do not establish present health. Discuss TS-004243: sampled p95 reached 4,609 ms on 2026-10-16 19:15 UTC for KB-001-02 / idx-001-02 [TS-004243]. This sparse sample does not establish an outage or its cause. Report: found. FSR-CUST-001-20261105, dated 2026-11-05, recommends: \"The proposed next step is a small set of approved support scenarios with a reviewer-scored answer-quality rubric.\" [FSR-CUST-001-20261105] That is a proposal, not completed work. Confirm the business question, content owner, current operating state, and evaluation results with the account team before discussing expansion.",
+    "ontology": {
+      "customer_id": "CUST-001",
+      "customer_name": "Adatum Corporation",
+      "knowledge_bases": [
+        {"id": "KB-001-01", "name": "Product Support KB"},
+        {"id": "KB-001-02", "name": "Employee Knowledge Hub"}
+      ],
+      "sample_count": 283,
+      "latest_sample_ids": ["TS-009967", "TS-009924"],
+      "highest_p95_sample_id": "TS-004243"
+    },
+    "retrieval": {
+      "customer_id": "CUST-001",
+      "report_reference": "FSR-CUST-001-20261105",
+      "report_date": "2026-11-05",
+      "documented_action": "The proposed next step is a small set of approved support scenarios with a reviewer-scored answer-quality rubric."
+    }
+  }
+}

notebooks/data/copilot-harness-foundry-fabric-iq/reports.jsonl

@@ -0,0 +1 @@
+{"id": "FSR-CUST-001-20261105", "customer_id": "CUST-001", "title": "Field sales report: Adatum Corporation", "report_reference": "FSR-CUST-001-20261105", "report_date": "2026-11-05", "documented_action": "The proposed next step is a small set of approved support scenarios with a reviewer-scored answer-quality rubric.", "content": "SYNTHETIC DEMO | CUST-001 | Fictional field-sales context Page 1 SYNTHETIC DEMO - NOT A REAL CUSTOMER SALES REPORT Field sales report: Adatum Corporation Support readiness before wider employee rollout Account key: CUST-001 | Tier: Strategic | Report date: 2026-11-05 | Reference: FSR-CUST-001-20261105 This document is fictional demo content. Customer identity and operational evidence are copied or calculated from the supplied synthetic telemetry. Conversations, opportunity hypotheses, stakeholders, and next actions below are invented; they are not actual field engagements, commitments, customer feedback, or sales forecasts. Account situation - fictional field narrative The fictional account scenario centers on a support-team knowledge assistant for product and troubleshooting content. The seller's working objective is to reduce the effort needed to locate an approved answer while keeping source context visible. This is a proposed evaluation story, not a claim that the organization has purchased a capability or that the named workflow is already deployed. The imagined stakeholder group includes support operations and an application owner. The discussion favors a small set of approved support scenarios with a reviewer-scored answer-quality rubric. The field team would keep the first evaluation narrow enough to distinguish content-quality issues from application or service behavior. No budget, contract value, procurement date, or named individual is asserted. Illustrative conversation note - invented \"The account team wants a concise answer with a useful source, not a longer list of documents.\" This quotation was authored for the demonstration. It is not attributed to an actual customer representative. The next discussion would test the underlying requirement with examples rather than treat this narrative as confirmed discovery. Operational evidence - synthetic dataset facts Evidence Observed value Telemetry coverage 2026-10-02 01:45 to 2026-11-05 16:15 UTC Samples / sampled KBs / sampled indexes 283 / 2 / 2 Customer inventory totals on latest sample 2 KBs; 2 indexes Regions / SKUs represented in samples East US / Basic Requests / errors / throttles in observed samples 4,825 / 111 / 30 Statuses recorded on samples Degraded: 3; Healthy: 269; Throttled: 11 Highest sampled p95 latency 4,609 ms (not a period-wide p95) SYNTHETIC DEMO | CUST-001 | Fictional field-sales context Page 2 Adatum Corporation | account follow-through Customer key: CUST-001 - SYNTHETIC DEMO Technical review - evidence-linked interpretation The latest recorded sample for this account is TS-009967 at 2026-11-05 16:15 UTC, labeled Healthy. It references KB-001-01 (Product Support KB) and idx-001-01, with 1 partitions and 1 replicas in that sample. These are sample-level configuration facts, not a verified description of every deployment owned by the account. The highest sampled p95 is attached to TS-004243 at 2026-10-16 19:15 UTC: 4,609 ms, status Degraded, KB KB-001-02, index idx-001-02. This identifies a concrete observation to investigate; it does not establish a root cause, an outage duration, an SLA breach, or a link to a particular sales conversation. Across only the observed samples, errors divided by requests equal 2.30%. Samples are sparse rather than a continuous account-level series. Do not extrapolate their traffic counts to a complete day or month, or average their p95 values into a global percentile. The source's customer inventory totals are repeated snapshot values, not measures to sum across samples. Sampled resources available for a grounded demonstration KB-001-01 - Product Support KB; index idx-001-01. Latest observed document count: 5,400; storage: 0.24 GB. Observation: 2026-11-05 16:15 UTC. KB-001-02 - Employee Knowledge Hub; index idx-001-02. Latest observed document count: 6,840; storage: 0.23 GB. Observation: 2026-11-05 12:40 UTC. Opportunity hypothesis and next field actions - fictional The proposed next step is a small set of approved support scenarios with a reviewer-scored answer-quality rubric. The seller would first confirm the business question and the content owner, then agree on a reviewer-scored answer-quality check. A technical review would separately examine the cited samples, the intended SKU and region, and a realistic workload before recommending any configuration change. Expansion remains conditional on evidence from that evaluation. The report does not recommend a particular upgrade, claim compatibility from a synthetic metric name, or promise performance. It suggests revisiting the customer-wide inventory totals with the account team because the sampled resources need not cover that entire inventory. Provenance and retrieval guidance Source: foundry_iq_realtime_telemetry_10000.csv; managed table dbo.telemetrysample. Join this report to structured data using customer_id = CUST-001, not a fuzzy company-name match. KB and index references above retain source values. Source coverage ends 2026-11-05 19:00 UTC. All named metrics are demo fields, not an asserted Foundry IQ product telemetry contract."}