Thoughtworks

thoughtworks.com

Thoughtworks offers separately purchasable advisory, implementation and ongoing operations rather than only research or software licenses. Buyers can commission technology roadmaps, application and platform delivery, legacy migration, data modernization or production AI systems. Its Enterprise AI offering explicitly covers AI strategy and agent development; data engineering and AI-assisted modernization should be evaluated separately from those AI services. Procurement scope should distinguish recommendations, prototypes, production deployment and operational responsibility. ^[1] ^[2] ^[3] ^[4] ^[5]

Services and delivery

Technology strategy and modernization roadmaps

Advisory. Advises on technology priorities, investment alignment and modernization sequencing. The deliverable is a roadmap; migration implementation requires a separate delivery scope. ^[1]

AI strategy advisory

Advisory. Assesses AI maturity, model costs and vendor dependency, and develops an implementation roadmap. This is advisory work rather than a commitment to deploy the recommended systems. ^[2]

Production AI agent development

Implementation. Builds production-grade agents using AI/works and Agent/works, combining software engineering, customer domain expertise and governance controls. ^[2]

AI Factory strategy and architecture

Advisory. Advises on an enterprise-owned AI operating model, model selection, token economics and hybrid architecture. Options described range from external model APIs to dedicated GPU infrastructure and on-premises models. ^[6]

AI Factory implementation

Implementation. Provides hands-on capability to build an AI Factory. Buyers should specify the architecture and deployment deliverables separately from strategy work; the evidence does not establish a standard managed-service package. ^[6]

Data platform modernization

Implementation. Migrates legacy data platforms to cloud infrastructure and develops data products. The offering includes data-mesh implementation approaches with domain ownership and governance; these services do not by themselves establish AI-model delivery. ^[4]

Legacy and mainframe modernization

Implementation. Delivers incremental modernization of enterprise systems, mainframes, data platforms and products. Mainframe work combines AI/works with Mechanical Orchard capabilities; AI-assisted engineering is not equivalent to delivering a customer ML application. ^[3]

Developer portals and self-service engineering platforms

Implementation. Implements developer-experience portals and self-service platform tools to bring fragmented engineering tools and documentation together. ^[7]

Digital product design and delivery

Implementation. Provides interaction and interface prototyping, testing, refinement and engineering-led product delivery. General product strategy and organizational coaching are distinct from this implementation scope. ^[8]

Application maintenance and enhancement

Managed services. Provides ongoing application maintenance and enhancement. Buyers should establish the supported application estate, release responsibilities and incident obligations in the service agreement. ^[5]

Infrastructure operations

Managed services. Manages cloud and on-premises infrastructure using site reliability engineering practices. Service hours and contractual availability commitments require confirmation. ^[5]

Managed data operations

Managed services. Operates data pipelines and addresses data quality and cloud-spend optimization. The service is classified as data operations, not automatically as AI or machine-learning operations. ^[5]

Technology Radar

Research. Publishes a twice-yearly technology snapshot. This knowledge-sharing publication is not evidence that Thoughtworks implements every technology it discusses. ^[9]

Company details

Technology and platforms

AI/works and Agent/works are identified as development and agent-governance platforms used in AI delivery. The supplied evidence does not establish standalone licensing terms, hosting arrangements or post-engagement usage rights. ^[2]

Track record

Thoughtworks reports productionizing a chatbot for Bayer’s PRINCE preclinical information platform after user feedback. The implementation used LLMs hosted by Bayer’s myGenAssist platform, retrieval-augmented generation over preclinical reports, source references and Langfuse observability. This supports production application delivery, not a claim that Thoughtworks trained the foundation models. ^[10]

BMW assigned Thoughtworks and other providers to build its Connected AI Platform using AWS, Kubernetes and Kubeflow. Thoughtworks reports a live pilot for identifying upcoming vehicle-maintenance needs, with infrastructure provisioning, access controls, ML workflow orchestration and version tracking. Delivery attribution is collaborative rather than exclusive. ^[11]

For an unnamed manufacturer, Thoughtworks reports migrating four batch jobs from mainframe JCL to Python on AWS Batch, moving Db2 schemas to PostgreSQL, and implementing automated validation and feature-toggled cutovers. Mechanical Orchard’s Imogen platform was involved. ^[12]

For Powerpay, Thoughtworks reports building an AWS landing zone, supporting PCI-related requirements, deploying privileged-access management and monitoring, and providing ongoing maintenance and incident management. This demonstrates specific infrastructure security controls, not an unrestricted cybersecurity practice. ^[13]

Engagement and delivery

A US bank engagement illustrates advisory-only scope: Thoughtworks delivered a market scan, software-development lifecycle review, GenAI opportunity roadmap, operating-model recommendations, workshops and policy guidance. The case states that the client subsequently executed recommendations; it does not establish that Thoughtworks implemented the proposed integrations or production tooling. ^[14]

Scope and limitations

The supplied material establishes service scope and delivery examples, but not pricing, standard project sizes, support hours, service levels or data-residency commitments.
Case-study outcomes are Thoughtworks’ own reports and have not been independently verified here.
AI platform licensing, hosting dependencies, handover rights and exit arrangements remain unverified.
The evidence supports specific security controls within delivery work, not a comprehensive cybersecurity-service classification.

Questions to ask

Does the statement of work end at assessment, prototype, production deployment or ongoing operations?
Who owns delivery and acceptance across Thoughtworks, customer teams and other providers?
What AI evaluation thresholds, human-review controls and incident responsibilities will apply?
What AI/works and Agent/works licenses, hosting dependencies and exit rights remain after handover?
Which references match our workload, regulatory setting and platform stack?
What support hours, service levels, data-residency commitments and pricing are contractual?

Overview