Skip to main content
ThinkscoopEngineering
EngagementsCapabilitiesWorkApproachNotesAboutStart a build
EngagementsCapabilitiesWorkApproachNotesAboutStart a build

Thinkscoop Engineering

Senior engineers, AI augmented. Not AI washed.

Engagements

  • AI MVP Sprint
  • AI Integration Pod
  • Agentic Workflow Build
  • Embedded AI Pod
  • All four, with prices

Capabilities

  • Retrieval and context
  • Evaluation and quality gates
  • Agent orchestration
  • Guardrails and escalation
  • Observability, cost and drift
  • Product and platform engineering

The practice

  • Engineering home
  • Delivered work
  • How we work
  • Engineering notes
  • About the practice
  • Questions we get asked
  • Start a build

Reach us

contact@thinkscoopinc.com

Other practices

  • Business Applications
  • Growth
  • Thinkscoop, the parent company

Legal

  • Privacy policy
  • Terms
  • Cookies
Thinkscoop Technologies LLPTeam based in India. Clients across the US, Australia and the UAE.
Engineering/Work/Booking.com

Intelligent Support Agent Resolving 68% of Queries Autonomously for Booking.com

Travel & eCommerce10 weeks
68%Autonomous resolution rate
<2minFirst response time
+22ptsCustomer satisfaction
10k+Daily queries handled

What this engagement is evidence of

Retrieval over a very large policy corpus where the correct answer depends on the version of a policy in force on the booking date, not the current one. The escalation surface was designed before the resolution path, which is the design decision the whole engagement turns on.

The situation

At Booking.com's scale, a 4-hour average support response time isn't just frustrating - it's a retention problem. Customers who can't resolve accommodation disputes, cancellation questions, or payment issues quickly don't rebook. The team had tried off-the-shelf chatbots and rule-based automation. Both failed for the same reason: the policy surface was too large and too dynamic for static decision trees to cover. Policies vary by property, country, booking type, and promotional tier. Any bot that couldn't reason across that complexity was destined to escalate.

What was breaking

  • 10,000+ daily queries across accommodation disputes, cancellations, payment failures, and policy questions - with no triage
  • Existing chatbots resolved less than 20% of queries without escalation, flooding human agent queues with fully resolvable issues
  • 4-hour average first-response time during peak periods, directly tied to negative reviews and reduced rebooking rates
  • 500,000+ policy documents across 50+ jurisdictions - impossible to maintain as static rules in any existing system
  • No context handoff between bot and human: agents received escalations with no background, forcing customers to repeat themselves

Escalation design is as important as resolution design.

Most support AI projects are measured on resolution rate. We argued from day one that resolution rate is the wrong metric - a bot that resolves 80% of queries incorrectly is worse than one that resolves 68% correctly and escalates the rest gracefully. We built the escalation experience first. Every escalated query arrives at a human agent with the booking data, the policy evidence, the conversation history, and the agent's uncertainty reasoning already assembled - so the agent can resolve in 90 seconds rather than 8 minutes. This changed how the team thought about the project: instead of trying to maximise automation, we focused on making the human-AI handoff seamless.

  • Confidence-threshold escalation: agent escalates with full assembled context rather than a cold handoff
  • Policy RAG with versioning: the knowledge base tracks when policies change and which version applies to which booking date
  • Reasoning traces exposed to human reviewers: agents can see exactly which policy clauses the AI cited for any decision
  • Separate evaluation dataset for each query category (refunds, disputes, payment failures) to prevent cross-category accuracy masking

A reasoning agent with live system access

The system is a multi-step LangGraph agent with direct tool access to Booking.com's booking retrieval API, payment and refund processing API, and a Pinecone vector store containing 500,000+ policy documents chunked and tagged by jurisdiction, property type, and booking tier. The agent retrieves the relevant booking, identifies the policy applicable to that specific booking context, checks refund eligibility, and - within defined thresholds - can initiate a refund or send a policy-based resolution without any human involvement. For edge cases, it routes to a human queue with a pre-populated resolution context card.

Policy RAG engine

500,000+ policy documents indexed in Pinecone with jurisdiction metadata, property category, and effective date ranges. Every policy retrieval includes the document version applicable at the time of the booking, not the current version - preventing incorrect application of retrospective policy changes.

Booking context retrieval

Live integration with Booking.com's booking API gives the agent real-time access to booking status, payment history, cancellation window, and property-level terms - so responses are always grounded in the actual booking, not a generic policy summary.

Refund authority layer

The agent can initiate refunds up to a defined threshold autonomously. Above that threshold, it prepares a refund recommendation with supporting evidence for a human agent to approve in a single click.

Context handoff card

Every escalation generates a structured handoff card: booking summary, conversation history, policies retrieved, agent's confidence score, and the specific reason for escalation. Human agents resolve escalated queries 6× faster than before.

Multi-jurisdiction compliance logging

Every customer interaction is logged with the policy version cited, the jurisdiction applied, and the resolution taken - satisfying data retention and consumer protection requirements across 50+ countries.

What it produced

The resolution rate improvement was significant. The first-response time improvement changed the customer experience entirely.

Autonomous resolution rate rose from 20% to 68%. First response time dropped from 4 hours to under 2 minutes. Customer satisfaction up 22 points.

What we would tell the next team

L1

Policy versioning is a first-class engineering problem

We underestimated how complex Booking.com's policy versioning was. Policies change at the property level, country level, and platform level - and the correct policy for a dispute is the one that applied at booking time, not today. Building a time-aware policy retrieval system added two weeks to the project but was the difference between a legally defensible system and a liability.

L2

Confidence calibration requires category-specific evaluation

A single confidence threshold across all query types led to overconfident responses on refund decisions and overconservative escalation on simple FAQ queries. We moved to per-category confidence thresholds with separate evaluation datasets for each - immediately improving both resolution accuracy and escalation precision.

L3

The handoff experience determines agent adoption

Human agents initially viewed the AI system with scepticism. What changed their minds wasn't the resolution rate - it was the quality of the escalation handoff card. When they realised the AI was assembling all the context they'd previously had to gather themselves, and that escalated queries now took 90 seconds instead of 8 minutes, they became active advocates for the system.

“We went from a chatbot that frustrated customers to an AI that actually solves their problems. The escalation design is what makes it trustworthy at scale - our human agents love it as much as our customers do.”

Head of Customer Experience, Booking.com

Vitals

  • Client: Booking.com
  • Sector: Travel & eCommerce
  • Duration: 10 weeks

Disciplines exercised

  • Retrieval and context
  • Guardrails and escalation
  • Agent orchestration

Platform and tooling

  • LangGraph
  • Pinecone
  • FastAPI
  • React
  • PostgreSQL
  • Datadog
Start a build

Model versions are deliberately not listed. The architecture treats the provider as a configuration decision, and a version printed on a page is out of date within the quarter.

$engineering / other engagements

Five more, and what each one cost to learn.

72%EY (Ernst & Young)A multi-agent AI system operating entirely within EY's Azure environment that ingests audit data, cross-references regulatory frameworks, and generates source-cited draft reports for human partner review.12 weeks · Professional Services78%Global FinTech (Confidential)A multi-step LangGraph reconciliation agent that reasons over discrepancy reports, queries source systems via API, and drafts resolution actions for human review - with a full evaluation harness measuring accuracy and escalation rate weekly.10 weeks · Financial Services4.2minHindustan Unilever LimitedA unified AI knowledge assistant with RAG over 200,000+ internal documents, role-based access control tied to SSO, full audit logging, and citation-grounded answers - operating entirely within HUL's Azure environment.14 weeks · FMCG / Enterprise3×SAMCoA real-time portfolio intelligence platform with live market data feeds, AI risk modelling, and a natural language query interface - so managers get cited answers to risk exposure questions in seconds, not a day later.8 weeks · Financial Services35%PixisA multi-touch attribution engine with AI-assisted causal inference, processing 5M+ events daily across Google, Meta, and TikTok - giving marketing teams a defensible, explainable model for budget decisions.6 weeks · MarTech / AI Marketing

Is your constraint the same as this one?

Send the task, the data it runs on and the systems it has to touch. You get a written read from an engineer within a working day.

Start a build