> Markdown version of [/jobs/ext/2049523-ai-engineer](https://www.wearedevelopers.com/jobs/ext/2049523-ai-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # AI Engineer - **Company:** Pret A Manger - **Location:** London, UK - **Experience:** Experienced - **Salary:** £55,688.0 - **Contract:** Permanent contract - **Skills:** Artificial Intelligence, Microsoft Azure, Cloud Engineering, Continuous Integration, Software Debugging, Python (Programming Language), Machine Learning, Search Technologies, Software Engineering, Microsoft Power Automate, Large Language Models, Technical Debt, Google Cloud Functions, Low-code, Virtual Agents, Restful APIs, Software Version Control, Serverless Computing - **Published:** August 14, 2026 - **Apply:** https://www.adzuna.co.uk/jobs/details/5841922132 ## About the Role We're looking for passionate team players to bring joy to our customers every day. As a AI Engineer you'll play a central role in building how work gets done across Pret's global functions. This is not a traditional software engineering role. The focus is to design, build and operate practical AI agents, automations & orchestration across Copilot Studio, Azure AI Foundry & Claude that improve productivity, decision-making, customer experience and team effectiveness., * 2-6 years in applied AI, software or automation engineering, with a portfolio or real production examples. * A genuine, active interest in AI and automation: you're already building agents, workflows or infra-adjacent tools in your own time, not just at work. * Infrastructure, platform or cloud engineering background is a strong plus: comfort with production operations, monitoring and reliability, not just prototyping. * Demonstrable architecture-level judgement: able to make and defend platform and design decisions unsupervised, as this role has no engineering peer group to lean on day-to-day. * Hands-on Microsoft Copilot Studio, including topic authoring, generative orchestration, connectors, knowledge sources and debugging. * Power Automate and the wider Power Platform, including Dataverse data modelling and security roles. * Python development with sound engineering habits (testing, version control, CI/CD). * Hands-on Azure, ideally Function Apps, and cloud-native deployment. * Practical experience of LLMs in production, including prompt orchestration, RAG and output evaluation. * Ability to reason about Copilot-vs-Foundry and knowledge-design decisions and explain the trade-offs. * Excellent written communication and structured fault-finding. This role must document to a standard that lets someone else pick up the work. * Advantageous: Azure AI Foundry (Agent Service, Prompt Flow), Azure AI Search, Semantic Kernel or the Microsoft Agent Framework, machine learning/forecasting and certifications AB-620 and AI-102/103. ## Description This is a sole-builder role: you will own delivery end-to-end, from architecture and platform choice through to build, testing and handover, rather than leading a team from day one. You will work across Support Centre functions, shops, markets, technology teams and third-party suppliers to turn business opportunities into working solutions, moving them from proof of concept through to production using Copilot Studio, Azure AI Foundry, Power Platform and Azure. The role combines strong engineering fundamentals with AI fluency, automation and solution design. You will build agents across knowledge, actions and instructions, exercising clear judgement on which platform a use case belongs on, ensuring solutions are useful, safe, scalable and adopted by the teams they support. You will be mentored technically by the Automation Manager but are expected to work independently as the hands-on architect and builder for the AI agent estate, with real ownership over the calls that will shape it long after the team has grown. This is the start of a brand-new AI function at Pret, and you'd be building it. This is a genuine greenfield opportunity with no legacy estate to maintain or backlog of technical debt. The architecture decisions you make in the first few months are the ones the function is built on. If you want to work at the frontier of applied AI adoption inside a real, global, operationally complex business rather than a lab, this is that role. The model is simple and deliberately fast-moving: you'll work directly with the business and with AI Business Analysts to identify real opportunities and turn them into working proofs of concept. Where a POC proves out and drives measurable business value, it gets scaled into production, and the team scales with it. 2027 is already pencilled in for growth, tied to business demand and to the data becoming available to support more ambitious use cases. What you build in the first six months becomes the standard everything after it is built on: the patterns, the guardrails, the way agents get evaluated and shipped. This is a foundations role, not a maintenance one., Discovery and proof of concept: most of your time in the first phase * Work directly with the business and Business Analysts to identify high-value automation and AI opportunities that fundamentally re-design what we do today * Scope and build working proofs of concept fast, testing feasibility and likely value before committing to a production build. * Make and defend the platform call (Copilot Studio vs Foundry vs. Claude) for each opportunity, with the reasoning to back it up. * Support the appropriate documentation and knowledge repositories as we lay the foundations for required governance and controls Taking proven ideas to production * Take POCs that demonstrate real business value from prototype through to a production-grade agent or automation inc. all required testing & validation * Build Copilot Studio agents (conversation flows, fallback strategies, Adaptive Cards in Teams) where the use case calls for low-code. * Build pro-code Foundry agents using Python, Azure Functions, custom skills and workflows where the use case calls for it. * Develop Power Automate cloud flows with real orchestration: error handling, retries, long-running operations. * Model and integrate data, connectors, REST APIs and MCP servers, with security and auditability built in from the start. Platform, architecture and governance: ongoing, and what the 2027 team inherits * Own the end-to-end architecture of the AI agent estate: you make the platform and design calls, and escalate trade-offs to the Automation Manager. * Choose the right knowledge and retrieval strategy (managed RAG, owned RAG or live query) for each use case. * Build eval sets, retrieval-quality checks, guardrails and observability, and deploy via CI/CD. * Work within the environment strategy and DLP policies; ensure retrieval is permission-aware by design. * Document every agent and every promotion decision to a genuinely handover-ready standard. This is the foundation the 2027 team builds on, not paperwork for its own sake. * Ensure we build agents in a cost-effective way to maximise the ROI and ensure we deliver within budget ## Related Videos - [Agentic AI - From Theory to Practice: Developing Multi-Agent AI Systems on Azure](https://www.wearedevelopers.com/videos/1532-agentic-ai-from-theory-to-practice-developing-multi-agent-ai-systems-on-azure) - [Reimagining app development with Low-code and AI](https://www.wearedevelopers.com/videos/1651-reimagining-app-development-with-low-code-and-ai) - [Rest API Antipatterns](https://www.wearedevelopers.com/videos/100208-rest-api-antipatterns) - [Guiding Agentic AI with Vue](https://www.wearedevelopers.com/videos/2033-guiding-agentic-ai-with-vue) - [Coding an Immersive Copilot using Unity / .NET and Azure OpenAI!](https://www.wearedevelopers.com/videos/1204-coding-an-immersive-copilot-using-unity-net-and-azure-openai) - [Azure AI Foundry for Developers: Open Tools, Scalable Agents, Real Impact](https://www.wearedevelopers.com/videos/1541-azure-ai-foundry-for-developers-open-tools-scalable-agents-real-impact) ## Related Articles - [How to Become an AI Engineer](https://www.wearedevelopers.com/magazine/331-how-to-become-an-ai-engineer) - [Navigating the AI Shift](https://www.wearedevelopers.com/magazine/629-navigating-the-ai-shift) - [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development) - [Prompt Engineering is a Job of the Past](https://www.wearedevelopers.com/magazine/342-prompt-engineering-is-a-job-of-the-past) - [The Prompt Engineer ✍️](https://www.wearedevelopers.com/magazine/216-the-prompt-engineer) - [Dev Digest 121 - AI goes offline](https://www.wearedevelopers.com/magazine/456-dev-digest-121-ai-goes-offline)