Sopra Steria

Technology and platforms

The AI knowledge-access offering gives examples of regulatory search, report generation and compliance chatbots. Named example stacks include Google Cloud with Vertex AI Search and Conversation, AWS with LangChain, and SmartSearch with InnerData. These are examples rather than a guaranteed compatibility catalogue; the page describes its illustrated solutions as proven in-house. ^[5]

The application-development service describes Agile, Product and DevOps methods, alongside cloud-native engineering and low-code/no-code integration. Managed platform services separately cover AWS, Red Hat OpenShift, Kubernetes-based platforms, Microsoft hybrid cloud and OpenStack. Platform hosting for AI applications should not be confused with a commitment to develop or maintain their models. ^{[2] [7]}

Engagement and delivery

The generative AI service describes a progression from opportunity workshops and readiness assessment to an implementation roadmap, tailored development, deployment and continued refinement. Cloud migration similarly includes execution, testing, acceptance, observation and handover. These descriptions support buying delivery beyond an initial assessment, although contractual responsibilities and acceptance criteria still need agreement. ^{[4] [3]}

Ongoing services include application support and continuous improvement, managed cloud platforms, and security monitoring and response. The Google Cloud security offering describes automated or orchestrated incident response as well as threat hunting. E-STAP 360 is a product-enabled managed service with analyst review, not simply a software licence. ^{[8] [7] [9] [10]}

Track record

For EDF, Sopra Steria describes jointly designing and deploying AMY, a natural-language voicebot for password resets, within the existing telephone support architecture. The case identifies Natural Language Understanding models. Its March 2023 account says rollout followed a five-month pilot and took place in June 2022, reporting a 23% reduction in support-team workload. This is a historical, firm-reported outcome, not a general performance commitment. ^[11]

An unnamed Norwegian municipality case documents Microsoft Azure migration from analysis through execution. Technologies and activities include Azure Migrate, Windows and Linux rehosting, SQL rehosting or refactoring, Terraform and CI/CD pipelines. It provides concrete cloud-delivery evidence, but customer anonymity limits independent reference checking. ^[12]

Scope and limitations

• The supplied notes do not establish pricing, minimum engagement size, staffing commitments, geographic service availability or contractual service levels.

• Bespoke AI model development is explicitly offered, but the notes do not provide a matching customer reference or model-training benchmark.

• Knowledge-access solution illustrations are described as proven in-house; they should not be presented as external customer deployments.

• Platform versions, deployment restrictions and current availability require confirmation. Responsible AI regulatory commentary is not a substitute for legal advice or contractual compliance criteria.

• Detailed claims from the Iberpay, ADEO and POST Luxembourg cases were excluded because the notes record unsuccessful repeat access.

Questions to ask

• Which activities will Sopra Steria own directly, and which will platform vendors or specialist partners deliver?

• Can you provide production references matching our proposed AI task, data sensitivity and operating environment?

• Which models and platform versions are supported, and what restrictions apply to deployment, data residency and data handling?

• What acceptance tests will cover AI quality, security, human oversight and operational readiness?

• Who owns application support, model updates, monitoring and incident response after handover, and under what service levels?

• What migration testing, rollback arrangements and documentation are included in the proposed scope?