🤖 Chat with Souvik AI
← Back to Portfolio Homepage
MakeMyTrip Connect Proposed Product Concept

HostAssist AI

AI-assisted partner onboarding for independent hotels

"Turn existing partner documents into structured onboarding data, with AI assistance and human control."

Guest Response Time 3.2 Min (was 4.2h)
Host RevPAR Revenue Lift +21.4% Revenue
Guest AI CSAT Rating 94.8% Rating
PROBLEM

Manual partner onboarding creates unnecessary data-entry effort and drop-off risk.

INSIGHT

Much of the required information already exists in documents partners already possess.

SOLUTION

HostAssist AI extracts and structures data, then lets partners review before risk-based verification.

SUCCESS

Improve onboarding completion and reduce effort without compromising listing quality or trust.

Strategic Framework Product Logic
USER Independent Hotels

Small and mid-sized hotel & homestay owners onboarding properties.

FRICTION Manual Form Entry

Repetitive typing of rates, GST codes, and room amenities.

INSIGHT Existing Documents

Details already exist in rate cards, tax certificates, and photos.

PRODUCT BET AI + Human Control

AI extracts structured data; partner reviews and confirms.

OUTCOME Higher Completion

Faster onboarding → More active room inventory on MakeMyTrip.

Interactive Product Prototype UX Demonstration

Compare the legacy onboarding experience with the proposed HostAssist AI flow.

Test Step 3 Risk Model:
9:41 MMT Partner 100% 🔋
MMT Connect | Host Onboarding V1 Concept
PM Rationale (Step 1 of 3)
V1 Improved Flow
PRODUCT DECISION Loading decision rationale...
AI ROLE Loading AI role...
QUALITY GUARDRAIL Loading guardrail logic...
KEY METRIC Loading metric...
ROLLOUT STRATEGY Loading rollout plan...
LEGACY FLOW
Manual multi-page forms with repeated typing.
HOSTASSIST AI
Document extraction with partner review.
Human-in-the-Loop Architecture Responsible AI
Layer Responsible Actor Core Task & Boundary
1. Extraction AI Multimodal Vision Extracts text & structures room data from document images
2. Communication AI System Surfaces field-level uncertainty tags (High, Needs Review, Low)
3. Review & Edit Hotel Partner Reviews pre-filled data, edits low-confidence fields, confirms output
4. Escalation Human Desk Team Reviews high-risk submissions or critical document mismatches

* Note: Confidence states (High, Needs Review, Low) are illustrative UX concepts, not production model accuracy benchmarks.

Proposed Verification Logic Risk Controls
Risk Tier Trigger Conditions System Verification Action
LOW RISK High AI confidence, verified GST format, complete documents Automated verification and fast listing activation
MEDIUM RISK 1+ low-confidence field, manual correction made by partner Partner re-confirmation required before activation
HIGH RISK Document mismatch, unreadable tax ID, potential duplicate Verification paused; escalated to Human Verification Desk
Core Product Principle: "AI should reduce partner effort, not remove partner control."
Measurement & Experimentation A/B Testing Framework
Primary Outcome Metric

7-Day Listing Completion Rate
Formula: (Partners completing onboarding within 7 days) / (Eligible partners starting onboarding).

Supporting Metrics

• Median Onboarding Completion Time
• AI Extraction Acceptance Rate
• Manual Correction Rate
• Time to Activation

Guardrail Metrics

• Incorrect Listing Information Rate
• Verification Failure Rate
• Duplicate / Fraud Detection Rate
• Support Ticket Spike Rate

CONTROL VARIANT

Partners use existing manual onboarding flow (multi-page text forms and manual rate grids).

TREATMENT VARIANT (HostAssist AI)

Partners use HostAssist AI document-assisted flow (document upload, extraction preview, partner review, risk-based verification).

SUCCESS Completion ↑ + Guardrails stable → Roll out V1
PARTIAL SUCCESS Completion ↑ + Guardrail worsens → Investigate & iterate
NO IMPACT Completion unchanged → Revisit UX & AI extraction
NEGATIVE Completion ↓ or Quality worsens → Pause / Roll back
Pilot to Scale Strategy Controlled Deployment
STAGE 1 PILOT

Small representative cohort of independent hotel partners.

STAGE 2 LEARN & ITERATE

Analyze field corrections, diagnose root causes, refine UI guidance.

STAGE 3 CONTROLLED ROLLOUT

Expand segment by segment after metrics & guardrails pass check.

STAGE 4 SCALE

Full deployment across all eligible independent hotel partner channels.

Rollback & Pause Conditions

Rollout pauses immediately if any trigger condition occurs:

  • ! Listing information quality or rate accuracy materially deteriorates
  • ! Verification failure rate increases materially
  • ! Fraud or duplicate property detection rates worsen
  • ! High-risk submissions are incorrectly auto-approved
Rule: Pause rollout → Investigate → Remediate → Re-test
If I joined the team tomorrow

1. Validate onboarding baseline funnel metrics
2. Interview a small cohort of hotel partners
3. Instrument HostAssist analytics events
4. Launch a controlled pilot with clear readiness checks
5. Review primary + guardrail metrics weekly

Expected Business Pathway: Reduced Effort → Higher Listing Completion → More Active Partners → More Room Inventory → Greater Booking Potential

Final Product Takeaway

"HostAssist AI is not about replacing partner judgment with AI. It is about removing repetitive work, making uncertainty visible, and applying automation in proportion to risk."

Next validation: Run a controlled experiment against the existing onboarding flow.