The newest title in architecture, and already one of the most confusing. Here is what each AI architecture role does, what to look for in a CV, and where organisations get them wrong when they hire.
DuI Architect is the newest title in architecture, and it is already one of the most confusing. Almost every architect has now added AI to their CV. For some, that means years of machine learning in production. For others, it means a Copilot rollout or a proof of concept that never went live.
Many organisations asking for an AI Architect actually need one of three quite different people. Before you can successfully recruit an AI Architect, you must first decide whether you need someone to set the direction, build the solutions, or run the models in production.
Also seen as: Head of AI Architecture, AI Strategy Architect, Responsible AI Lead
The Enterprise AI Architect sets the direction for how the organisation adopts AI. That covers which use cases to invest in, which platforms and models to standardise on, and how AI is governed, secured and kept within the law. They make sure AI is adopted as part of the enterprise architecture, rather than as a collection of disconnected experiments.
Look for AI strategies and roadmaps, use case prioritisation linked to business value, and AI platform decisions across Azure AI Foundry, Amazon Bedrock or Google Vertex AI. You should also see AI governance and responsible AI frameworks, and awareness of regulation such as the EU AI Act or standards such as ISO/IEC 42001. They will usually have a strong Enterprise or Data Architecture background behind the AI work.
There are very few people who have genuinely set AI strategy for a large organisation, because the discipline is so new. Be wary of CVs where the AI strategy is a slide deck with no investment decisions or governance behind it. Often the best candidate is an experienced Enterprise or Data Architect who has led AI adoption, rather than someone with AI in every job title since 2023.
Also seen as: GenAI Architect, LLM Architect, Agentic AI Architect
The AI Solution Architect designs solutions built on large language models and other AI services: assistants, retrieval augmented generation (RAG), document processing and, increasingly, agents that take actions on their own. They decide how the model is grounded in the organisation’s data, how it connects to existing systems, and how its answers are tested, monitored and kept safe.
Look for Azure OpenAI, Amazon Bedrock, Google Gemini or Anthropic’s Claude, and RAG designs using vector databases such as Azure AI Search, Pinecone or pgvector. You should also see orchestration frameworks such as LangChain or Semantic Kernel, agent frameworks and the Model Context Protocol (MCP), and evaluation, guardrails and cost control. The best ones will talk about what happened after the proof of concept.
The gap between a demo and a production system is enormous in AI. Plenty of people have built an impressive proof of concept in a fortnight. Far fewer have taken one into production, with the security, testing and running costs that come with it. Ask them what is running today, who uses it, and what it costs each month.
Also seen as: Machine Learning Architect, MLOps Architect, AI Platform Architect
The ML and MLOps Architect designs the platform and processes that take machine learning models from experiment to production, and keep them working once they are there. That covers training pipelines, feature stores, model registries, deployment, monitoring and retraining.
Look for Azure Machine Learning, Amazon SageMaker, Vertex AI or Databricks, and MLflow for experiment tracking and model management. You should also see feature stores, CI/CD for models, model monitoring and drift detection, and Python throughout. Most come from data engineering or data science, and have worked on predictive models in areas such as pricing, fraud, demand forecasting or risk.
ML Architects are easily confused with Data Scientists and ML Engineers. A Data Scientist builds the model. An ML Engineer productionises it. The ML Architect designs the platform that lets the organisation do both repeatedly and safely. If every example in the CV is a model they built themselves, you are looking at a very good Data Scientist.
If you are hiring for an AI role and are not sure which of these you actually need, please get in touch. It is surprisingly common for the job description to describe one of these roles while the job title asks for another.
Ben Clark
bclark@konvergent.co.uk
