Case Study – Retail

Executive Summary

DoorDash’s multi-sided marketplace needed to modernize how it served its growing base of U.S. merchants. The legacy touch-tone IVR lacked flexibility, led to caller frustration, and created bottlenecks for merchants seeking quick resolution to time-sensitive issues. Many callers bypassed the system entirely, pressing “0” to reach an agent—resulting in misrouted calls, long hold times, and inconsistent support experiences. DoorDash set out to improve call routing accuracy, increase self-service adoption, and reduce operational costs through AI-powered automation.

To achieve this, DoorDash partnered with CloudCookies —an AWS Conversational AI partner with deep expertise in Amazon Connect and Amazon Lex. CloudCookies deployed a phased strategy that included intent discovery, conversational design tailored to merchants, and targeted self-service automation for 23 high-impact call reasons. The result was a dramatic improvement in routing accuracy, caller adoption, and containment, along with a 218% increase in self-service compared to the legacy experience, delivering an additional annual savings exceeding $800,000.

By transforming the merchant IVR into a modern, intelligent voice experience, DoorDash not only streamlined agent queues and reduced costs, but also laid the foundation for scalable AI-driven support across its entire platform.

:)

“In less than 4 months, we achieved measurable success with conversational AI and self-service while building repeatable processes for future engagements. We couldn’t have done it without CloudCookies!”

-Chaitanya Hari, DoorDash Product Manager

:)

:)

Customer Challenge

DoorDash, a multi-sided marketplace, faced a critical challenge in delivering fast, effective phone support for one of its most important segments—Merchants. The existing IVR relied heavily on outdated touch-tone menus, making it difficult for merchants to quickly resolve urgent issues like delivery problems, payment inquiries, and order cancellations. This led to widespread caller frustration, with many merchants bypassing the menu entirely by pressing “0” to reach a live agent—causing frequent misroutes, longer handle times, and increased operational costs.

With thousands of merchants relying on voice support daily, DoorDash set an ambitious goal: to implement a modern conversational AI solution that would achieve over 80% routing accuracy across 23 merchant call reasons, increase adoption of self-service workflows, and improve containment compared to the legacy experience. This transformation needed to drive efficiency without compromising the responsiveness merchants depend on.

To succeed, DoorDash required a partner with deep expertise in conversational design, Amazon Connect, and AI-powered automation. The ideal partner also needed to bring analytical rigor and a scalable framework for routing logic, self-service development, and machine learning model tuning. The vision was clear: build a smarter, more intuitive support experience that would reduce costs, empower merchants, and set a new standard for service excellence.

:)

:)

:)

Partner Solution

CloudCookies partnered with DoorDash to completely transform the merchant support experience, migrating from a legacy touch-tone IVR to a modern, conversational AI system built on Amazon Connect and Amazon Lex. This shift required not only a technical deployment but a ground-up design of merchant-facing interactions that could match the fast-paced, high-pressure environments in which these users operate.

The first phase of the engagement centered on conversational design and system architecture, with CloudCookies identifying 23 high-impact intents—ranging from background check inquiries and order cancellations to account setup and pay issues. Using DoorDash’s call recordings and transcription data, CloudCookies performed theme and pattern detection, created word clouds, and mapped user utterances to expected intents. A self-service benefit heatmap was built to prioritize where automation could provide the most value.

The design also included identifying real-time API integration points that would enable merchants to perform tasks such as checking an order status or reporting issues without speaking to an agent. Conversational storyboards were developed for each use case, incorporating best practices in UX, error handling, and fallback logic.

Recognizing the challenges merchants face—background noise, frequent interruptions, and multi-step inquiries—CloudCookies designed the system with features such as:

• Barge-in functionality, allowing callers to interrupt prompts and provide their input immediately
• Touch-tone (DTMF) fallback, dynamically offered when background noise is detected or no response is given
• Context-aware prompts, which adapt to user input and skip unnecessary steps to reduce friction

In Phase 2, CloudCookies launched and refined seven self-service workflows, including canceling orders, confirming pickup status, and handling red card issues. These workflows were tightly integrated with DoorDash’s APIs via AWS Lambda, enabling real-time data exchange and action completion directly from the IVR.

To ensure performance and reliability, CloudCookies monitored KPIs such as:

• Intent accuracy (average: 89.87%)
• Adoption (average: 98.35%)
• Self-service containment (↑ 81% over legacy)
• Error rates, with 93.2% of callers experiencing fewer than four errors

By continuously tuning prompts, improving confidence thresholds, and monitoring real-world usage data, CloudCookies delivered a merchant experience that felt responsive, efficient, and intuitive—replacing frustration with functionality and paving the way for ongoing AI innovation.

:)

:)

Additional features included fuzzy logic for order IDs, free-form input collection, custom slot types, voice collection for complex alphanumeric data, and CloudCookies self-service performance benchmark for regression testing. To measure the impact, containment, deflection, and engagement return on investment (ROI) were assessed to develop a business plan for future engagements. The implementation process aimed for operational excellence, efficient tuning, and the promotion of self-sufficiency, with CloudCookies providing break/fix support and staff augmentation.

:)

:)

Why CloudCookies

Results

CloudCookies’ solution for DoorDash Merchants delivered powerful results, modernizing voice support and significantly improving the efficiency of merchant operations:

• Achieved $800,000 in annual operational cost savings (OPEX) through improved routing, reduced handle times, and increased self-service usage.
• Drove a 227% increase in self-service adoption compared to the legacy touch-tone IVR system.
• Reached an average of 82% across 27 key merchant intents.
• Reached 90% adoption, ensuring merchants engage meaningfully with the system.
• Reduced error rates, with 85% of callers experiencing fewer than 4 errors, even in noisy or high-disruption environments.
• Enabled real-time resolution of high-volume merchant issues such as order cancellations, order status, and point of sale issues through API-integrated self-service flows.
• Laid the foundation for scalable, AI-powered voice experiences across DoorDash’s global merchant network.

Building on this success, DoorDash plans to expand self-service capabilities to additional languages, introduce new merchant workflows, and enhance predictive routing using machine learning—continuing to set the standard for voice-first merchant support.

:)

About CloudCookies

CloudCookies is a distinguished Advanced Tier AWS Technology and Consulting Partner, specializing in business process automation, conversational design, and self-service solutions. We exclusively harness the power of AWS services to provide innovative solutions for our clients. Our services cover consulting, migration, integration, custom development, training, and support, ensuring our clients receive top-tier expertise. What sets CloudCookies apart is our extensive experience and deep knowledge, making us the preferred choice for those seeking transformative AWS solutions that break the mold.

:)