Skip to main content

Centers of Competence

Purpose​

Centers of Competence (CoC) are organizational units that concentrate expertise in specific technical areas of AI. They ensure delivery quality, knowledge accumulation, and standardization of approaches.

Types of centers of competence​

Data Science CoC​

  • Expertise in machine learning, statistical modeling, and data analysis
  • Development and validation of ML models
  • Methodologies for running experiments and assessing model quality

MLOps / Platform CoC​

  • Tools and infrastructure for developing, testing, and deploying models
  • CI/CD for ML pipelines, monitoring models in production
  • Managing the ML Platform as an AI product

NLP / LLM CoC​

  • Expertise in natural language processing and large language models
  • Developing the Knowledge Assistant (RAG), AI Copilot, and other LLM products
  • Evaluating and testing LLM solutions, prompt engineering, fine-tuning

Functions of centers of competence​

Regardless of specialization, each CoC performs the following functions:

  • Developing standards and best practices — unified approaches to experiments, code review, testing, and documentation
  • Expert support for working groups — assigning specialists to initiative working groups, consultations
  • Developing AI products and platforms — technical ownership of products, their development and support
  • Architecture reviews — taking part in reviewing technical solutions before they go to production
  • Accumulating and transferring knowledge — maintaining a knowledge base, internal training, retrospectives, a catalog of delivered solutions
  • Research and innovation — tracking new technologies, running PoCs to assess applicability

CoC versus the AI Office​

Centers of competence and the AI Office perform different functions:

AI OfficeCenters of Competence
FocusPortfolio, methodology, governanceTechnical expertise, delivery
Responsible forWhat to do and whenHow to do it
DecisionsPrioritization, stage gates, resourcesTechnical approaches, architecture
ArtifactsPortfolio reporting, process standardsCode, models, platforms, technical standards

The AI Office coordinates the work of CoCs but does not manage their technical decisions. CoCs are autonomous in choosing tools and approaches within corporate standards.

Scaling​

  • In small organizations, the functions of several CoCs may be combined into a single team
  • As the practice grows, CoCs are spun off as expertise and workload accumulate
  • Each CoC has a leader who works with the AI Office Lead