Turning Data and Claude AI Into Commercial Intelligence on AWS

Project snapshot

For data-rich businesses, the challenge is no longer simply collecting information. The real opportunity lies in turning proprietary data into intelligence that improves products, helps customers make better decisions, and creates new sources of value.

GBG, a global expert in identity and location technology, set out to do exactly that. Following years of product development and acquisitions, the company held enormous volumes of identity-verification and product data across multiple platforms, regions, and teams. GBG saw an opportunity to turn this fragmented data estate into a strategic intelligence capability.

Working together, GBG, Intellias, and AWS created a unified cloud data platform capable of supporting analytics, machine learning, and generative AI at enterprise scale. The platform became the foundation for automated product recommendations, benchmarking and customer intelligence — and is now being extended with Claude-powered conversational analytics and agentic workflows.

Rather than approaching generative AI as an isolated experiment, GBG built it on top of governed, production-ready data infrastructure. This allows Claude to work with trusted business context while preserving the privacy, security, and tenant isolation required in a highly regulated identity environment.

Key initiatives

Starting with customer value

GBG treated the programme as a customer-value initiative rather than simply a technology modernization project.

The ambition was to move beyond reporting what had already happened and help customers understand what they could do next: identify performance opportunities, benchmark results, optimize identity-verification journeys, and ultimately interact with complex analytics more naturally.

This led GBG and Intellias to define automated product insights as the programme’s initial Backbone Use Case — a high-value use case around which the wider data and AI roadmap could be organized.

Creating one intelligence layer from fragmented data

GBG’s product portfolio had grown both organically and through acquisitions. Individual platforms had their own data architectures, schemas, and operating models, making cross-product analysis difficult.

Intellias and GBG consolidated these data sources into a unified AWS data layer capable of supporting analytics across regions and product lines.

The resulting platform provides the governed context required not only for dashboards and machine learning but also for generative AI systems that need reliable access to enterprise data.

Designing security into the AI architecture

GBG operates in identity verification and fraud prevention, where privacy and data isolation are fundamental requirements.

Security, governance, and responsible AI therefore had to be designed into the platform from the beginning. The architecture incorporates controlled access, auditability, role-based permissions, and governance mechanisms to support sensitive data workloads.

These principles became even more important as the platform expanded into generative AI. Claude-powered applications are designed to operate within GBG’s own AWS environment, with tenant-level controls, row-level security, guardrails, evaluation gates, and observability built into the solution.

Moving from analytics to AI-powered interaction

The original platform made it possible to generate automated recommendations, benchmarking, alerts, and customer intelligence.

The next step was to change how people interact with that intelligence.

GBG and Intellias began developing a conversational analytics experience using Claude Sonnet through Amazon Bedrock. Instead of requiring users to navigate predefined dashboards or ask a data team to answer every non-standard question, the solution is designed to allow users to query governed product and identity-verification data using natural language.

This represents an important evolution of the platform: from delivering predefined insights to enabling users to explore trusted data interactively.

Business challenge

GBG operates across highly regulated markets and processes large volumes of identity-verification data.

Over time, global expansion and acquisitions created a complex technical estate in which product and customer data was distributed across independent systems. This created several barriers to extracting commercial value from that information.

Different schemas and architectures made cross-platform analysis difficult. Data often needed to be extracted and transformed manually before teams could analyze it. Potential churn signals, anomalies, and commercial opportunities could therefore take time to surface.

The company also lacked a unified intelligence layer capable of providing a consistent view across accounts, products, geographies, and customer segments.

Perhaps most importantly, the fragmented architecture limited what GBG could do with AI.

Machine learning and generative AI are only as useful as the information they can reliably access. Without governed, compatible, high-quality data, AI applications risk becoming isolated proofs of concept rather than production capabilities.

GBG therefore needed more than a new data platform. It needed a secure intelligence foundation that could support analytics, machine learning, generative AI, and eventually agentic workflows across the organization.

Solution

AWS provided the scalable cloud foundation while Intellias worked with GBG to translate the company’s product and data strategy into a delivery roadmap.

A combined team of data product managers, data platform engineers, data scientists, MLOps specialists, AI engineers, and GBG product and data teams worked through an incremental release model.

Intellias began with an assessment of GBG’s data strategy, governance, architecture, AI engineering, and MLOps capabilities. These findings were translated into a prioritized roadmap designed around measurable customer value.

The resulting architecture included several interconnected capabilities.

Unified AWS data platform

Separate identity platforms were integrated into a common data architecture, making information from different products and global regions compatible for analysis.

AWS services provide the processing, storage, analytics, machine-learning, and generative-AI capabilities required to operate the platform at scale.

Production MLOps

Model training, deployment, versioning, monitoring, and retraining were formalized into a controlled lifecycle.

This allowed GBG to move beyond experimental models toward repeatable production AI capabilities.

Automated customer intelligence

The platform supports automated product recommendations based on real customer behavior and performance.

GBG can identify potential configuration improvements, compare customers with relevant peer groups, detect unusual usage patterns, and identify opportunities to improve outcomes or introduce additional products.

Claude-powered conversational analytics

With the underlying data foundation established, GBG and Intellias began developing a conversational interface over GBG’s Insights data.

The solution uses Claude Sonnet through Amazon Bedrock, with an agent architecture orchestrated through Amazon Bedrock AgentCore and a LangGraph runtime.

Users can ask questions in natural language while the system works against a governed semantic layer rather than exposing the model directly to unrestricted enterprise data.

For a regulated identity business, this required more than connecting a large language model to a database.

Intellias designed tenant isolation and row-level security controls so that each organization can only access information it is authorized to see. Bedrock Guardrails, evaluation-gated promotion, separate production and non-production environments, observability, agent-memory controls, and per-user cost attribution provide additional operational safeguards.

The initial release is being introduced to GBG internal users before a wider customer-facing rollout.

During an early measured period in September 2026, the environment recorded 367 platform invocations and 239 chat completions with a 98.74% success rate, providing the team with operational data to continue evaluating and tuning the experience ahead of broader adoption.

Claude-powered engineering and delivery

Claude is also changing how the platform itself is developed and operated.

The combined GBG and Intellias team introduced Claude Code and Claude-based agentic workflows across engineering and delivery management.

Secure integrations with Jira, Confluence, GitHub, and AWS allow Claude to assist with tasks that previously required significant manual effort but were too context-dependent for traditional automation.

One example is programme reporting.

GBG’s Data Innovation organization manages nineteen initiatives across three squads. Previously, reconciling the leadership roadmap with Jira, reviewing individual initiatives, cleaning up board inconsistencies, and preparing an accurate status update could represent roughly two days of work.

The team developed a Claude-powered workflow that reads the board, proposes progress figures with supporting evidence, prepares an update for human review, publishes the approved information, and then updates the corresponding Jira records.

Human approval remains part of the workflow at critical stages.

The overall process can now be completed in approximately two hours.

Additional Claude-assisted workflows support pull-request review, engineering triage, access-request processing, SQL view generation, infrastructure management, and recurring data reconciliation.

Rather than replacing engineering judgement, the approach gives engineers and delivery teams an agentic layer that can perform repetitive investigative and coordination work while keeping people responsible for approval and higher-value decisions.

High-value use cases

The combination of unified data, machine learning, and Claude is creating a growing portfolio of intelligence capabilities.

Recommendation-led product adoption

GBG can use real product and transaction data to identify opportunities to improve customer outcomes.

Instead of relying only on generic product messaging, teams can demonstrate expected improvements associated with particular products or configuration changes.

Advanced peer benchmarking

Customers can be benchmarked against relevant organizations across sectors, operating contexts, and geographies.

This gives customers a clearer view of their own performance and helps GBG provide more contextual guidance.

Algorithmic prospecting and growth targeting

Sales and account teams can use recommendations, clustering, and behavioral signals to identify potential growth opportunities and prioritize customer engagement.

Proactive account intelligence

Automated monitoring can identify unusual behavior that may indicate a customer issue, operational anomaly, or commercial opportunity.

In one instance, unusual transaction activity identified by the platform was traced to a penetration test being accidentally run against a live environment. Proactive intervention allowed GBG to address the problem before it developed into a larger customer issue.

Conversational exploration of enterprise data

Claude adds another interaction model on top of these capabilities.

Instead of limiting users to predefined dashboards and reports, conversational analytics can allow authorized users to ask follow-up questions, investigate trends, and explore governed enterprise data using natural language.

The goal is to make sophisticated analytics accessible without compromising the data controls required by GBG’s regulated environment.

Agentic engineering operations

Claude-powered agents are also being applied internally to engineering and delivery workflows where traditional automation is difficult because decisions depend on context spread across multiple systems.

This includes delivery reporting, engineering triage, code review, infrastructure workflows, and operational data tasks.

Business outcomes

The programme has evolved from a data-platform modernization initiative into a foundation for scalable AI.

A unified intelligence platform

Data that previously lived across separate products and architectures can now be analyzed through a shared platform, giving GBG a common foundation for customer analytics, machine learning, and generative AI.

New commercial capabilities

GBG has transformed product and transaction data from an operational byproduct into a source of customer intelligence.

Automated recommendations and benchmarking help teams provide more proactive guidance and support new commercial propositions.

A scalable foundation for Claude

Because Claude is being introduced on top of a governed enterprise data platform rather than as a standalone AI experiment, GBG can combine generative AI with trusted proprietary context.

The architecture provides the security, tenant isolation, evaluation, observability, and operational controls needed to move conversational AI toward production use in a regulated environment.

Faster engineering and delivery operations

Claude-powered workflows are reducing manual work across the joint engineering organization.

Processes that previously required extensive reconciliation across tools can increasingly be delegated to agents while retaining human review and approval.

Incremental AI adoption

GBG and Intellias have deliberately expanded AI in stages: first creating the data foundation, then delivering analytics and machine-learning use cases, followed by conversational and agentic AI.

This approach allows each new capability to build on infrastructure, governance, and operational practices that have already been proven.

Key metrics

Technology

Amazon Web Services, Amazon Bedrock, Claude Sonnet, Amazon Bedrock AgentCore, LangGraph, Bedrock Guardrails, Amazon SageMaker Unified Studio, AWS Glue, Amazon EMR, Amazon S3, AWS Lambda, PostgreSQL, Airflow, Python, custom role-based access control, audit logging, Claude Code, GitHub, Jira, and Confluence.