Enterprise internal-service dialogue corpus

Enterprise internal-service dialogue corpus: real employee questions in natural language inside Microsoft Teams (VPN failures, app-access requests, printer/device setup, security-policy questions), each paired with the specific approved source document the answer was drawn from (e.g. "VPN guide 2026.pdf", "MFA procedure", an IT SharePoint page), plus explicit satisfaction feedback (mandatory thumbs-down rating on every fix), guided diagnostic multi-turn dialogs, escalation-to-human handoffs with retained context, unanswered-request logs, topic-popularity and deflection analytics, and the indexed SharePoint/knowledge-base corpus itself.

Open buyer room →

Sample

Brickroad customers

Sample this dataset before you buy it.

Your sourcing agent asks Witivio and files the sample in your catalog — private to you.

Start 7-day trialNothing is charged until the trial ends.

Coverage

Universe, instruments and categories.

Industry: Application Software
Instruments: equities
Categories: AlternativeSentimentReference

Dataset card

Type: not stated
Format: Multimodal — conversational text transcripts (multi-turn) with document citations and metadata (source doc id, topic tag, satisfaction flag, deflection outcome, response latency), plus the underlying indexed PDF/SharePoint document corpus
Volume: ~50 indexed documents per tenant environment (the page's stated plan default, adjustable via add-ons) across ~5,000-25,000 dialog turns per tenant-year, with the aggregate cross-tenant corpus plausibly 10^5-10^6 labelled question-answer-citation turns once the customer base is summed
Users: ~5,000-30,000 employee end-users across the customer base (inferred — a 53-employee French vendor selling a Teams agent with a starter tier at 10-50 users and enterprise rollouts implies tens of mid-market to large tenants; the page references a 38% deflection result achieved at a 4,000-user company)
History: 3 years
Update frequency: not stated
Growth: Active — recently repositioned around AI Desk Pro with named use-case pages (IT support plus siblings), current-branded content referencing 2026 documents, and a stated 53-person headcount; the shift from Teams robots to AI agents indicates an active product pivot rather than a maintenance tail
Launched: ~2018-2019 for the Witivio Teams bot/app-builder platform (company record opened mid-2018, Bordeaux); ~2024-2025 for AI Desk Pro, the RAG/agent product that generates this corpus
Delivery: not stated
Entity mapping: not stated
Sample: not stated
Point-in-time: not stated
Licence: not stated

supplier index read from their site