AWS Partner Network (APN) Blog
Category: Amazon Machine Learning
Taking control of agents on Amazon Bedrock AgentCore with Dataiku
As agent estates grow from a handful of pilots to hundreds of agents in production, teams need to decide which agents to stop and which to promote. This post shows how Dataiku Agent Management works with Amazon Bedrock AgentCore and Amazon CloudWatch to discover, monitor, and control agents, using the signals AgentCore already emits.
Automating compliance testing with Fisent, Launchpad.io, and Amazon Bedrock
Learn how Kaufman Rossin partnered with Fisent Technologies, Launchpad.io by Pegasystems, and AWS to transform manual Regulation E compliance testing into a scalable Outcomes-as-a-Service solution. The service combines deterministic workflow automation, applied generative AI, and cloud-native AWS infrastructure. It has processed more than 100,000 dispute cases across 30+ fintechs. Along the way, Kaufman Rossin achieved 70% faster testing and 50% fewer manual hours.
A Data-First Approach to Mainframe Modernization with AWS Transform and Qlik
AI agents are making mainframe modernization programmable. This post walks through a data-first migration pattern that pairs Qlik Replicate’s log-based Change Data Capture (CDC) with the agentic AI of AWS Transform and Kiro. Using the strangler fig pattern and blue/green deployment, it migrates wave by wave, keeping the mainframe as system of record until each module is validated and the mainframe is finally decommissioned. Rising hardware costs, a shrinking COBOL and PL/I talent pool, batch-processing limits, and constrained innovation are driving organizations to make the move.
Optimize AI Costs with New Relic Cloud Cost Intelligence on AWS
Managing AI costs is a top challenge for organizations scaling generative AI on AWS. This post shows how AWS CUR 2.0, Amazon S3, and AWS Trusted Advisor form the data foundation that New Relic Cloud Cost Intelligence builds on for Amazon Bedrock workloads. The result is token-level cost visibility inside the observability workflows engineering teams already use, shifting AI from opaque experimentation to disciplined, ROI-driven operations.
Scaling agentic AI for governed banking operations with Apexon and AWS
Banking, financial services, and insurance (BFSI) institutions need AI systems that can act autonomously at market speed, but autonomy without governance is a nonstarter in an industry defined by regulatory intensity and operational risk. In this post, Apexon and AWS present an architecture-first approach to operationalizing agentic AI on AWS, with governance built in from the start across three layers: intelligence, orchestration and execution, and governance, security, and audit. Learn how governed agentic patterns apply to core BFSI workflows like customer onboarding, credit processing, and payments exception handling, enabling autonomy at scale while keeping every action traceable and auditable.
Governing AI agent identity with Amazon Bedrock AgentCore and Saviynt Zuma
AI agents authenticate, hold permissions, and act on data, yet most organizations manage them as application components rather than governed identities. Learn how Saviynt Zuma works with Amazon Bedrock AgentCore to govern agents as first-class identities, from registration and least-privilege provisioning through runtime authorization, certification, and deprovisioning. Security teams can then answer which agents exist, who owns them, and what they can access.
Building trusted AI with data and AI governance partners on AWS
The race to deploy agentic AI is outpacing the governance needed to put it into production – 74% of companies plan to deploy by 2027, yet only 21% have mature governance models. AWS launched GEAR (Governance Enabling Agentic Readiness) to close that gap, identifying partners whose governance accelerators compress months of setup into weeks and turn AI pilots into production-ready systems with measurable results.
Mastering automated agent performance evaluations in Amazon Connect Customer
TEKsystems Global Services (TGS) helped a large media and entertainment enterprise transform their Amazon Connect Customer automated evaluation system from 66.67 percent accuracy to 100 percent. They developed a strategic classification framework that matches each evaluation question to the optimal automation method, either generative AI, conversational analytics categories, or contact metrics. This layered approach eliminated AI hallucinations, enabled full interaction coverage, and turned quality assurance from an unreliable manual process into a trusted, production-ready system that delivers actionable coaching insights at scale.
As generative AI adoption accelerates, the gap between deploying automation and making it trustworthy is becoming the critical differentiator. The next frontier of multi-modal insights and real-time coaching starts with the right foundation.
Designing governance-first AI agents with Fenergo using Amazon Bedrock
Autonomous AI agents are transforming financial services — but in regulated industries, capability alone doesn’t drive adoption. Trust does. In this post, Fenergo and AWS engineers walk through the architecture behind the Fenergo Agent Operating System (Agent OS): a governance-first layer built on Amazon Bedrock, Amazon DynamoDB, Amazon SNS, Amazon SQS, and Amazon Timestream for InfluxDB that makes every agent decision traceable, explainable, and audit-ready. Learn how to design agentic AI systems that answer the question regulators always ask: why?
Mission Cloud accelerates data validation with Amazon Bedrock AgentCore
A data migration consulting firm’s growth was limited by dependence on a small group of expert practitioners to execute complex validation workflows, creating bottlenecks that delayed client deliverables and constrained profitability. Mission Cloud and AWS built a production AI agent on Amazon Bedrock AgentCore that guides consultants through structured validation workflows and supports natural language SQL analysis, democratizing expertise across the entire team. New employees are onboarded significantly faster, and validation cycles that previously took weeks now complete in days.









