How ASAPP Uses Amazon Bedrock to Achieve 91% First-Call Resolution

ASAPP is an AI-native customer service platform that orchestrates large language models to automate contact center interactions for enterprise clients. By deploying Anthropic’s Claude through Amazon Bedrock, ASAPP eliminated its homegrown PII redaction layer and reduced call escalations by up to 40%, while helping clients achieve a 91% first-call resolution rate. The platform now automates more than 90% of contact center interactions, with human agents freed to handle three times the volume of complex cases.

Impact

91%

First-call resolution rate

77%

Cost reduction per chat interaction

3x

Human agent capacity increase

49%

Customer self-service growth

40%

Reduction in call escalation

>90%

Contact center automation rate

Challenge

ASAPP’s GenerativeAgent relied on a homegrown PII redaction layer that slowed response times and degraded output quality, limiting the platform’s ability to resolve complex customer issues without human escalation.

Solution

ASAPP integrated Amazon Bedrock with Anthropic’s Claude Sonnet models, eliminating the homegrown PII layer, gaining built-in data privacy compliance, and enabling more natural, capable language generation across voice and chat channels.

Tools & Technologies

What Leaders Say

Every bit of the product benefit is directly related to our ability to use powerful models. And Amazon Bedrock is core to how we do that.

Nirmal Mukhi, Vice President of AI Engineering, ASAPP

We bring AI to the contact center in a way that encourages consumers to trust the product and doesn’t mislead them.

Mackenzie Smith, Head of Partnerships and Business Operations, ASAPP
Get the full story.

Sign up to read complete case studies, access detailed metrics, and unlock all use cases.

Full Story

ASAPP operates at the intersection of AI and enterprise customer service, building a GenerativeAgent Platform that orchestrates five to seven large language models simultaneously to handle voice and digital interactions end-to-end. Its clients include some of the world’s largest organizations, where marginal improvements in contact center efficiency translate directly to significant operational savings and measurable customer satisfaction gains.

The platform’s early architecture included a homegrown PII redaction layer built to protect sensitive customer data, but this came at a steep cost. The system slowed interactions, degraded response quality, and prevented GenerativeAgent from accessing the contextual information it needed to resolve complex requests. Rebooking an airline ticket, for example, required address data that the redaction system stripped out, forcing workarounds or immediate human escalation.

ASAPP transitioned from its previous AI provider to Amazon Bedrock, gaining access to Anthropic’s Claude Sonnet models alongside AWS’s built-in enterprise data privacy controls. The GenerativeAgent Platform now routes interactions dynamically across its LLM ensemble depending on use case, latency requirements, and channel (voice or chat). Crucially, the homegrown PII layer was retired entirely, as Amazon Bedrock’s compliance infrastructure met enterprise standards without the performance penalty.

The results were immediate and measurable. GenerativeAgent resolved up to 40% more issues without escalating to human agents. ASAPP’s enterprise clients saw a 49% increase in customer self-service engagements, a 77% reduction in cost per chat interaction, and a first-call resolution rate of 91% for complex service issues. Human agents simultaneously gained the capacity to handle three times as many complex interactions because routine volume was absorbed entirely by AI.

ASAPP is now working with AWS to test Amazon Nova Sonic, a speech-to-speech model delivering near-real-time voice conversations. The company’s trajectory signals a broader shift across enterprise contact centers: from menu-driven IVR systems to AI agents so capable that customers routinely ask whether they’re talking to a human. For organizations processing millions of interactions at scale, this has moved from pilot to critical infrastructure.

Similar Cases

P
Postman
Up to 1,150/year
developer hours saved

Postman selected Claude Opus 4.6 as the default model for Agent Mode, saving developers up to 1,150 hours per year and nearly $1M annually for a 10-person team in API development automation.

TechnologyCAClaude APIABAmazon Bedrock
C
CustomGPT.ai
10,000+
paying customers served

CustomGPT.ai is a no-code RAG-as-a-Service platform enabling businesses to build domain-specific AI agents on their own data. By building its vector retrieval infrastructure on Pinecone, the company scaled to over 10,000 paying customers, stores 400+ million vectors, and delivers sub-20ms P50 query latency at 99.95%+ uptime. The result is a platform that earned the #1 ranking in a RAG accuracy benchmark, with Pinecone providing the foundation that let the engineering team focus entirely on product differentiation rather than infrastructure management.

TechnologyPPinecone
J
Jamf
70%+
employee adoption rate

Jamf, the leader in Apple enterprise management securing over 30 million devices for 75,000+ organizations worldwide, deployed the Moveworks AI Assistant (internally named Caspernicus) to transform employee support across IT, HR, Legal, and Facilities. Within the first month, 30% of employees adopted the assistant; today, more than 70% of Jamf’s workforce actively uses it to resolve requests that once took days in a matter of minutes. By meeting employees where they work in Slack, the platform automated routine tasks like password resets, software provisioning, and onboarding workflows, freeing IT to focus on higher-impact initiatives.

TechnologyMAMoveworks AI Assistant
TX
Terminal X
0.68 to 0.91
f1 retrieval accuracy improvement

Terminal X is a vertical AI platform for institutional investors that acts as a 24/7 research agent, processing millions of financial documents for hedge funds, asset managers, and private equity firms. By rebuilding its retrieval architecture on Pinecone’s vector database, Terminal X improved F1 retrieval accuracy from 0.68 to 0.91, cut average latency by over 35%, and doubled deployment velocity. Users now save approximately three hours per day, and investment memo preparation dropped from two days to half a day.

Financial ServicesTechnologyPPinecone
D
Delphi
>100M
vectors stored

Delphi is an AI platform that enables coaches, creators, and experts to deploy interactive “Digital Minds”—always-on conversational agents trained on their unique content. Scaling from proof of concept to a commercial platform with thousands of customers required a vector database that could support millions of isolated namespaces, billions of vectors, and sub-second retrieval under variable load. Delphi selected Pinecone, achieving P95 query latency of 100ms and keeping retrieval under 30% of total response time—freeing the engineering team to build product rather than manage infrastructure.

TechnologyPPinecone
A
Assembled
~95%
ticket handling time reduction

Assembled is a workforce management and customer support optimization platform serving enterprises like Stripe, Etsy, and DoorDash. To power Assembled Assist, the company built a hybrid RAG pipeline combining Pinecone vector search with Algolia keyword retrieval and LLMs from OpenAI and Anthropic. Support tasks that previously took 40 minutes now complete in 2 minutes—a 95% reduction in handling time.

Customer Support TechnologyPPineconeOLOpenAI LLMs
N
Notion
Millions
notion ai users reached

Notion, the connected workspace platform used by millions worldwide, integrated Cohere Rerank into its search pipeline to power Notion AI’s search accuracy across multilingual enterprise workspaces. Every search and Notion AI interaction now routes through Cohere Rerank, delivering dramatically improved relevance while cutting the cost and complexity of embedding-based retrieval for smaller workspaces.

TechnologyCRCohere Rerank
F
Fujitsu
World-class score
jglue benchmark performance

Fujitsu, the global IT and digital transformation company with 124,000 employees, partnered with Cohere to develop Takane — a state-of-the-art Japanese large language model built on the Cohere Command series. Designed for private deployment in regulated sectors such as finance, healthcare, and government, Takane delivers world-class performance on the JGLUE benchmark and is now integrated into Fujitsu’s AI service offerings and data intelligence platform.

TechnologyCCCohere Command