Staff Conversation Designer (Support)

Our Global Structure Deliveroo is now part of DoorDash, bringing together teams with even greater reach, scale, and ambition. Depending on your role, you may collaborate with teammates, systems, and leaders across DoorDash and Wolt. Together, we’re unlocking new possibilities as one global team. Join us in our mission to transform the way people shop and eat, where impact, innovation and growth drive everything we do. Our Product team defines how millions of people experience Deliveroo, turning complex marketplace challenges into simple, high-impact solutions. Working at the intersection of data, design and engineering, we set direction, prioritise what matters most, and deliver outcomes that move the business forward. We're looking for a Staff Conversation Designer to join in London and set the design direction for AI-powered support across Deliveroo, DoorDash and Wolt. You'll bring a strong AI fluency, technical practice, and editorial point of view to build and shape conversations powered by our LLM models: deciding what a conversation needs to say, what to leave out and how to make each exchange clear, useful, and worthy of people's trust. You'll shape the personality, voice, tone and experience principles that express our brands, local language and support philosophy. You will craft conversation specs, prompts, and evals, annotate data, and co-build the tools and harness that drives our experience. Success means fewer avoidable problems, better resolutions when help is needed, and stronger design practices that help teams deliver those outcomes consistently. What You’ll Be Doing • Set support conversation strategy across support journeys, connecting what people need in the moment, what the business can promise, and what operations can actually do. Influence roadmaps with Product, Engineering and Operations, deciding where to invest, what to sequence first and how to balance quality, speed and impact. • Design and maintain conversation flows. Lead proactive guidance, chat and voice experiences from definition through launch, anticipating needs to prevent or resolve problems before people ask for help. Create and maintain multi-turn conversation specs covering context, interruptions, changing intent, clarification, error recovery, fallback paths and human handoffs without repetition. • Prototype new conversational experiences. Invent and test new conversational patterns, interaction paradigms, personas, and support experiences. Use LLMs and code to explore how people could ask for help, act on options and move between chat, voice and other interfaces. Work with Engineering to shape the information, actions, and human judgement these experiences need, using working prototypes and user feedback to decide what to pursue. • Shape voice, tone and behaviour. Set the editorial direction for support conversations across brands, markets and modalities. Write and iterate on system prompts and persona definitions that express this direction through model behaviour. Make deliberate choices about wording, pacing and tone, adapting to local voices, policies and access needs while keeping conversations useful and promises accurate. • Build evaluation datasets and testing plans. Establish how teams evaluate conversation quality before and after launch. Create evaluation datasets, testing plans and scoring rubrics to test conversation flows and model behaviour. Define annotation guidelines and annotate conversations for evaluation and model training, calibrating judgments so teams assess quality consistently. Combine these evaluations with research, performance data and reviews of hundreds of transcripts to improve accuracy, resolution quality and customer effort, and influence what the team prioritises. • Design conversational safeguards. Identify conversational risks and define, implement and test guardrails. Address hallucinations, harmful or biased responses, and requests outside the service’s scope. Define how conversations protect sensitive information, stay within the system’s authority to act, and acknowledge uncertainty or seek human judgement. • Raise conversation design capability. Help the wider Studio develop technical fluency and conversational judgement through mentorship, education and shared tools. Build reusable prompt guidance, evaluation tools and working practices with design and cross-functional peers, and lead their adoption across teams to improve conversation quality and inform product priorities. What You’ll Need to Thrive • Work samples. Work samples showing strong editorial and conversational judgement, whether through conversational systems, content programmes or shipped products. Show your personal contribution, the direction you set, alternatives you explored and evidence of improved outcomes. We welcome different formats that make your thinking, execution and influence clear. • Strategic leadership. Experience setting direction across teams and influencing product priorities, with examples of resolving ambiguity and making trade-offs between user needs, commercial goals and technical constraints. • Editorial judgement. A strong editorial point of view, backed by excellent writing and editing. You can explain why a conversation needs a particular message, sequence or tone, cut what gets in the way and give others specific feedback that improves the work. Your background may be in editorial, content design, linguistics, product design or another discipline; what matters is how you apply that judgement to interactive experiences. • Hands-on AI practice. Practical fluency with LLMs and code, demonstrated through working AI prototypes and direct investigation of model behaviour. You can distinguish conversational problems from limitations in data, tools or the wider system. • Service design judgement. The ability to connect individual conversations to the wider service, including user intent, context, available actions and agent workflows. Your decisions ac