AI & Agentic Technical Delivery Lead
Accenture
- Level
- Senior
- H-1B history
- 998 approvals (FY2023)
- Posted
- 16h ago
Skills
About this role
AI & Agentic Technical Delivery Lead Senior Manager (CL6) London, Manchester Accenture Song - Data & AI We Are Accenture Song accelerates growth and value for our clients through sustained customer relevance. Our capabilities span ideation to execution: growth, product and experience design; technology and experience platforms; creative, media and marketing strategy; and campaign, content and channel orchestration. With strong client relationships and deep industry expertise, we help our clients operate at the speed of life through the unlimited potential of imagination, technology and intelligence. Visit us at: https://www.accenture.com/gb-en/about/accenture-song-index The Role Generative AI is moving from experimentation to enterprise delivery — and the gap between the two is where most organisations struggle. The difference between a compelling demo and a production system that works at scale, handles edge cases, stays within cost, and actually changes business outcomes is enormous. We are looking for an AI & Agentic Technical Delivery Lead who has been there before — multiple times. Someone who has personally designed and shipped GenAI and agentic solutions, led the engineering teams that built them, and carried the accountability when things did not go to plan. Someone who has made the hard calls: when to stop a POC that is not ready for production, when a proposed architecture will not hold, when a team is underestimating complexity, and when an estimate is wishful thinking. This is a senior technical leadership role, operating at the intersection of client advisory, solution quality, and delivery execution What Sets You Apart You have shipped GenAI products. Not just prototypes that lived in a demo environment. Not just POCs that got handed to another team. Actual production systems, with real users, real data, real edge cases, and real consequences when something went wrong. You have led engineering teams — hands-on enough to review a pull request and challenge a design decision, senior enough to set the technical direction and make it stick. You know how development teams work from the inside, which means you know where estimates go wrong, where technical debt accumulates silently, and where a team is about to hit a wall they have not seen yet. You are the person in the room who can tell the difference between an AI solution that will work and one that sounds good in a slide. Not because of intuition — because you have built both, and you know the signals.
What You Will Do
Technical Leadership & Quality Serve as the senior technical authority on GenAI and agentic solution delivery across client engagements — setting the bar for what production-ready looks like Review and challenge solution architectures end-to-end: LLM selection, RAG pipeline design, agentic workflow logic, evaluation frameworks, guardrails, cost modelling, and scalability Define and enforce the standards that separate a working POC from a deployable product — and have the difficult conversations when that line is being blurred Identify technical risk early — data quality issues, latency bottlenecks, hallucination exposure, integration complexity, model drift — before they become delivery problems Act as a technical quality layer across engagements: reviewing outputs, challenging assumptions, and ensuring what gets built actually solves the business problem Client Advisory & Solution Shaping Advise senior client stakeholders on what is technically feasible, what is not, and what the real path to production looks like — with the credibility that comes from having done it Shape and size AI solutions accurately: translating ambiguous client ambitions into concrete technical requirements, realistic delivery estimates, and honest risk assessments, understanding the existing vendor/product landscape and capability Distinguish clearly between what a POC proves, what it does not prove, and what needs to be true before scaling — and