The business trigger
Manual triage repeated the same questions and lost important context between chat, web forms and specialist follow-up.
Engagement: AI-assisted enquiry operations← All confidential case files
Representative confidential case file · 06
Reducing enquiry friction with human-governed AI triage
A business receiving service enquiries through web forms, chat and messaging, then manually forwarding them across specialist teams.
Anonymous client profileInternational multi-service business
Client baseUnited Kingdom-registered international group
Delivery marketsUnited Kingdom, UAE and Türkiye service enquiries
Primary usersProspective clients, concierge staff and specialist service owners
◇
Client identity protected
Client identity, prompts, knowledge content, system connections, volumes, performance data and internal controls remain withheld. This dossier describes the engagement structure without publishing private evidence, numerical performance or information that could expose the client.
The actual build
An approved knowledge base, AI classification and summarisation flow, guided questions, CRM routing rules, human review gates and audit-dashboard specification.
Designed for: Prospective clients, concierge staff and specialist service ownersWhere the engagement began
Staff repeatedly read the same introductory questions, requested missing details and decided who should respond. Enquiries crossing property, formation, technology or private-client topics could be routed inconsistently. Valuable context disappeared between the first message and specialist follow-up.
Why a standard service package would fail
Automation could improve speed but could not be allowed to provide uncontrolled advice, make eligibility decisions or commit the business. The system needed approved knowledge, privacy rules, confidence thresholds and visible human ownership.
How Tengri framed the mandate
The workflow automated preparation, not accountability. AI could classify, summarise, retrieve approved information and propose routing; a responsible person remained in control of advice, exceptions and commitments.
From diagnosis to operating handover
Workflow diagnosis
Mapped channels, enquiry types, repeated tasks, risk points, missing data and existing ownership.
Knowledge control
Separated approved public answers from restricted, professional or case-specific information.
Assisted workflow
Designed classification, structured summaries, follow-up questions, CRM fields and routing logic.
Governance & review
Added confidence thresholds, human checkpoints, audit fields, exception handling and improvement reviews.
What the mandate produced
- Enquiry and risk map
- Approved knowledge architecture
- AI triage logic
- Structured summary format
- CRM routing rules
- Human-approval checkpoints
- Exception and escalation model
- Review dashboard specification
06 · Business value
What changed operationally
- Specialists received more structured context before first follow-up.
- Repetitive sorting and information requests could be reduced.
- Sensitive questions remained subject to defined human review.
- The business gained a controlled foundation for additional automation rather than isolated AI tools.
CONFIDENTIAL
What is intentionally not published
Client identity, prompts, knowledge content, system connections, volumes, performance data and internal controls remain withheld.