Lead AI Architect - Enterprise Transformation

Job Description

We are looking for a Lead AI Architect to own the technical architecture underpinning this transformation — from translating business and product requirements into implementable system designs, to defining the reusable platform capabilities, technology choices and engineering standards required to build and operate AI solutions at scale.

This is a hands-on architecture leadership role. You will work across the AI CoE, enterprise architecture, engineering teams, business functions and implementation partners to ensure individual AI solutions are built on a coherent, scalable and maintainable foundation rather than as disconnected point solutions.

The distinguishing challenge here is not building one impressive AI workflow. It is building the second, fifth and twentieth at a fraction of the cost of the first, on infrastructure that survives contact with real enterprise identity, real permissions, real data quality and real production incidents.

Where this role sits

Reports to: Head of the AI Centre of Excellence

Alignment: Dotted line to QAD engineering architecture; close partnership with Enterprise Architecture, Security & GRC, Data & Platform Engineering, and Product

Technical leadership of: The CoE engineering team (Senior AI Engineer and subsequent hires) and the engineering capacity provided by system integrators and cloud partners

Primary internal customers: The functional AI Engagement Specialists who own the workflow pipeline and PRDs for the business functions in scope

Team size: No direct reports at hire; technical authority across an internal and partner engineering group. Line management may follow as the CoE scales.

What you will own

Architecture and platform direction

• Own the enterprise AI reference architecture, defining the target-state architecture and the MVP or “golden path” required to begin delivering priority AI workflows while the broader platform evolves.

• Translate workflow requirements and PRDs into build-ready technical architecture — system boundaries, source systems, APIs and integrations, data requirements, identity and access, user and consumption layers, orchestration, and non-functional requirements. Closing the gap between a high-level PRD and a design detailed enough to generate development stories is an explicit, named responsibility of this role.

• Make and govern key architecture decisions across infrastructure, data, integration, semantic layer, AI and LLM services, agent orchestration, state and memory, application layers and deployment architecture — and record them, with rationale and revisit triggers, as durable decision records.

• Define what is reusable enterprise capability versus workflow-specific build, and where capability should be centralised versus federated, balancing scalability against speed to value.

• Lead platform and technology selection, evaluating options on functional fit, existing enterprise capability, build and migration effort, total cost of ownership, operating complexity, the skills required to sustain the platform, security, regulatory requirements and vendor lock-in.

• Draw the boundary between internal enterprise workflow and productisable capability. In a software company, some of what the CoE builds will be a candidate for the product. Design so that boundary remains crossable rather than discovering later that an internal tool cannot be productised without a rewrite.

• Continuously evolve the architecture as new use cases, products and capabilities emerge — avoiding premature complexity while ensuring near-term decisions do not constrain the longer-term platform.

Delivery and production readiness

• Drive the foundational platform build in parallel with workflow delivery, identifying the minimum non-negotiable capabilities required before solutions can safely reach production rather than waiting for the complete target platform.

• Establish production architecture and engineering standards covering development, test and production environments, CI/CD, logging, monitoring, AI and application observability, tracing, state and memory, human–agent handoffs, and operational support.

• Own the technical approach to AI operations — model and agent lifecycle management, evaluation, drift, reliability, observability, and cost and usage management.

• Own evaluation as an architectural concern. Define how quality and correctness are specified, measured, regression-tested and monitored, and make evaluation infrastructure a first-class platform capability rather than something each workflow reinvents.

• Design the human-in-the-loop and escalation model at the architecture level: where a human must approve, where an agent may act autonomously, how handoffs preserve context, and how reversal or remediation works when an agent gets it wrong.

• Own the cost architecture. Establish unit economics per workflow, model routing and tiering, caching strategy, token budgets, and the telemetry required for the business to see what each workflow costs to run.

Data, identity and integration

• Architect the data and integration foundations required by AI workflows — APIs, semantic and translation layers, metadata, data access patterns, and appropriate reuse of existing enterprise platforms.

• Solve identity, delegated authority and traceability for agents. Design how an agent acts on behalf of a user across multiple systems under least privilege, how roles and permissions propagate into retrieval and tool invocation, and how every action is attributable to a human or an agent in an auditable trail. This is among the hardest problems in the architecture and is squarely owned by this role.

• Design for multi-tenancy and data isolation where AI capability touches customer data or customer-facing surfaces.

Governance and technical leadership

• Set architecture guardrails for security and responsible AI, partnering with existing Security, GRC and enterprise architecture functions rather than recreating them inside the AI team.

• Provide technical leadership to the AI engineering team and implementation partners — reviewing designs, resolving trade-offs, and ensuring builds conform to the reference architecture. You remain the internal technical authority even where significant engineering capacity is supplied by SIs or cloud partners.

• Run architecture conformance review for partner-delivered work, with the standing and the willingness to reject a design that will not survive production.

Qualifications

What we are looking for

Essential

• 10+ years across software, platform, cloud, data or enterprise architecture, including meaningful experience architecting production AI/ML or GenAI systems — systems with real users, real failure modes and real operating cost, not pilots.

• Deep understanding of modern AI and agentic architectures: LLM platforms and model selection, agent orchestration, tool and API invocation, RAG and context architecture (chunking, hybrid retrieval, reranking, grounding, freshness), semantic layers, state and memory, evaluation and observability.

• Strong experience with at least one major cloud ecosystem — AWS and/or GCP — with the ability to evaluate services across platforms objectively rather than defending a favourite.

• Strong foundation in distributed systems and enterprise system engineering: APIs, integration patterns, authentication and authorisation, data architecture, application architecture, DevOps and CI/CD, and production operations.

• Experience designing architectures that span multiple enterprise applications and heterogeneous data sources, including systems you do not control.

• Ability to reason rigorously through build versus buy, reuse versus new capability, cost versus performance, and MVP versus target state — and to explain the reasoning to people who will be affected by it.

• Experience taking ambiguous business requirements through technical discovery into implementable designs.

…