Director of Support and Services

Academic Key; Greater London; Hybrid; GBP 150,000 - 190,000; Full time

Job summary

Sona London is seeking a Director of Support and Services to lead an AI-first support function. This hands-on leadership role focuses on architecting agent behaviour, configuring systems, and scaling operations without proportionally increasing headcount.

You will partner with Product, ML, and Engineering to raise agent capability and improve customer outcomes. The ideal candidate brings 7+ years in customer-centric roles, deep experience with AI agents and LLMs, and a track record of

Qualifications

• 7+ years in customer support, customer success, or operations, with at least 2 years in a leadership role.
• Experience at scale with support economics and AI-first workflows.
• Defined AI system tasks, failure modes, and iterative improvements.
• Deep familiarity with agentic systems and LLM behaviour.
• Ability to read traces and communicate fixes to engineers.
• Proven track record of multiplying impact without headcount growth.

Responsibilities

• Design and own the agentic support operating model end-to-end.
• Architect multi-tier routing and escalation to optimise resolution.
• Define tiered support models and premium services.
• Lead a lean orchestration team focused on configuration and quality monitoring.
• Collaborate with Product and ML to improve agent behaviour.

Skills

Leadership

AI-first workflows

Agentic systems

LLM behaviour

Trace analysis

Problem solving

Job description

# Director of Support and Services**Sona** · LondonAbroad London◷ Full Time✓ Link validatedAbout Sona Frontline work keeps the world running. Restaurants, hospitals, schools, nurseries, airlines, stores, emergency services. Together they employ 80% of the global workforce, over 3 billion people, and they're among the hardest businesses anywhere to run. By-the-minute scheduling, HR and payroll make them far more operationally complex than the average office, yet just 1% of venture funding has ever gone to software built for them. Rising costs, wage increases, labour shortages and tighter regulation have made fixing that urgent. Sona is the AI-native workforce management platform for frontline organisations, with agents that match staffing to real demand and take the admin load off managers. That's a $500bn+ market the software industry has mostly walked past. We've raised over $100M from N47, Felicis, Gradient and Northzone, signed more than 100 enterprise customers across the UK and US, and opened offices in London, New York, Austin and Lisbon. More on working at Sona here ( https://www.sona.ai/careers ). About the Role: The Director of Support and Services is responsible for building and scaling an AI-first support function. This is a hands-on leadership role combining agent architecture, operational design, product collaboration, and team leadership to solve support at scale without scaling headcount at the same rate. The role owns the support operating model end to end, defining what agents solve, when they escalate, how they improve, and how humans supervise them. You'll work closely with Product, ML, and Engineering to architect agent behaviour, identify the systemic friction agents miss, and continuously raise agent capability. You'll also lead a lean team focused on configuration, quality monitoring, and exception handling, rather than on taking tickets directly. Success looks agentic: resolution rates, accuracy, cost per resolution, and how much human time the agents save. If you built a traditional, high-performing support team here, we'd consider that a miss. The aim is support infrastructure where agents are the production layer and humans orchestrate them. Responsibilities: Design and own the agentic support operating model, defining what problems agents solve end to end, which require escalation, at what thresholds, and what triggers human review. Own the configuration layer that lets you evolve agent behaviour without shipping code. Architect multi-tier agent routing and escalation logic, matching agent assignments to customer segment (self-service, assisted, premium, white-glove) to maximise agent resolution while maintaining quality thresholds. Own the cost model around agent operation, human supervision overhead, and escalation. Use agent data to decide when to invest further in agent capability and when to escalate to humans instead. Own agent quality and capability metrics, establishing KPIs that track resolution rates, accuracy, customer satisfaction, and human exception workload. Run regular retrospectives on agent failures to identify systematic fixes. Work closely with Product and ML to improve agent behaviour, identify friction points agents miss or mishandle, and make the commercial case for whether to train the agent, escalate, or fix the product. Prototype new agent capabilities and guide their build. Direct a dedicated engineering resource to build agent infrastructure, including orchestration logic, intelligent routing, multi-turn resolution agents, predictive issue detection, and self-service expansions. You define what agent behaviours to build. They build the plumbing. Build and lead a lean, high-performing orchestration team, small by traditional support standards, that owns agent configuration, quality monitoring, exception handling, and continuous improvement. The focus is systemic improvement, not heroic individual performance. Define tiered support and services models based on customer segment and agent capability, and define white-glove or premium services where human expertise is the differentiator. Requirements: 7+ years in customer support, customer success, or operations roles, including at least 2 years in a leadership capacity. You've operated at scale and understand support economics. Hands-on experience building or scaling AI-first workflows where AI is the production layer. You've defined what an AI system should do, found its failure modes, and iterated on its behaviour yourself, rather than just evaluating vendor solutions. Deep familiarity with agentic systems and LLM behaviour. You understand agent limitations, failure modes, and hallucination risks, and how to structure problems so agents succeed. You can read traces, identify where agents are going wrong, and communicate fixes to engineers. A track record of solving problems systematically through configuration and process design, not by scaling headcount. You've built tiered models, designed escalation logic, and found ways to multiply your team's impact without adding headcount. Able to reaso