Building an Intelligent Chatbot Framework for Cloud Operations

25%

Reduction in Development Cycle Times

8

AWS Enterprise-grade Deployed Accounts

60%

Reduction in Average Query Time

Background

Our client is a global management consulting firm known for helping leading organizations solve complex business challenges through strategic insight, innovation, and digital transformation. With a presence in over 50 countries, they serve Fortune 500 companies, governments, and nonprofits across industries. As part of its commitment to driving next-generation solutions, they invest heavily in emerging technologies, including AI and cloud. Their innovation unit focuses on building advanced digital products and AI-powered platforms to help clients achieve scalable, data-driven outcomes. This combination of deep industry expertise and cutting-edge technical capability makes them a key player in shaping the future of business and technology.

Challenge

The key challenge our client faced in building their Slackbot IT Agent was how to integrate hundreds of tools in a way that was both scalable and user-friendly. The original prototype allowed users to perform common AWS and Azure infrastructure tasks via Slack, but required them to manually switch between multiple agents during a conversation. This led to a frustrating user experience and made the system difficult to maintain, as each agent required extensive, custom-coded logic to interpret and process requests. To address this, the team standardized how tools were described, defining consistent functions, inputs, and outputs, and introduced dynamic agent generation based on user queries. This reduced code duplication, simplified tool integration using JSON stubs, and enabled the chatbot to grow without increasing complexity. The result was a more flexible, intuitive Slackbot capable of delivering a seamless experience across a wide range of IT tasks.

Approach

Keyrus partnered with the client to design a scalable, production-grade Slackbot that streamlines internal IT operations using AWS-native services. To address usability and maintainability challenges, we implemented a dynamic agent-routing system powered by Amazon Bedrock, enabling the chatbot to understand user intent and automatically route requests to the appropriate AI agent. This eliminated the need for manual switching and reduced backend code complexity. Supporting services like OpenSearch (for RAG workflows), Redis (for session continuity), and DynamoDB (for lineage tracking) were deployed in a multi-AZ setup to ensure performance and fault tolerance. The solution followed AWS best practices around CI/CD, security, and operations. Infrastructure was managed via Terraform and GitHub Actions, while security was enforced through Secrets Manager, Parameter Store, and IAM policies. The system integrated with tools like Jira, GitHub, and ServiceNow, enabling seamless automation across the enterprise without disrupting existing workflows. With improved modularity, observability, and AI-driven interactions, the chatbot now runs across eight AWS accounts—supporting their cloud-native vision while driving real-world efficiencies in IT support.

Key results

Eliminated backend code sprawl through modular Bedrock agent architecture, reducing maintenance overhead and accelerating development cycles.

Enabled fault-tolerant, scalable IT automation aligned with cloud-native best practices.

Reduced the average code volume per agent significantly by being able to re-use code and utilizing lambda layers for shared code.

Benefits

This engagement demonstrates how Keyrus leverages Generative AI and AWS-native architectures to solve real enterprise challenges, transforming fragmented, manual IT processes into a seamless, intelligent support experience. By combining deep cloud expertise with a user-first design, we delivered a scalable solution that not only met technical goals but also improved employee efficiency, reduced operational burden, and aligned with the client’s broader digital transformation strategy.