Job Description

The Lead AI Engagement Specialist owns a portfolio of AI workflows within one or more business functions, end to end. You will identify where AI genuinely changes how the work is done, redesign that work, write the PRD, drive the build with the architect and engineers, land the adoption, and then manage the performance of the agents in production as a product you are accountable for.

This is a product and transformation role, not a business analysis role. The distinction matters and we mean it: you will own a roadmap, write specifications accurate enough to build from, make scope calls, and be measured on realised value rather than on delivered documentation. If a workflow does not clear its quality bar or nobody uses it, that is your problem to solve, not to report.

Where this role sits

Reports to: Head of the AI Centre of Excellence

Alignment: Dotted line to the leadership of the business function you cover

Works with: AI Engagement specialists, Lead AI Architect and Senior AI Engineer daily; external implementation partners; Security, GRC, Legal and Data teams as required

Owns: The prioritised workflow pipeline, PRDs, roadmap, adoption and live agent performance for one function. Lead & manage other AI Engagement specialists.

Accountable for: Realised, measured business value — not deployment count

What you will own

Opportunity identification and prioritisation

- Own the AI workflow pipeline for your function — a ranked, evidence-based view of where AI creates value, refreshed as the business and the technology move.

- Distinguish rigorously between work that AI genuinely changes and work that merely looks automatable in a demo — and be willing to argue that a well-sponsored idea is not worth building.

- Prioritise on value, feasibility, data readiness, risk and time to impact, in partnership with the architect.

Process discovery and workflow redesign





- Map the current state properly: volumes, cycle times, cost to serve, exception paths, decision points, systems and data touched, and where the work actually breaks — grounded in observation and data, not in what the process document claims.

- Redesign the work, not just the tool. Decide what the agent does, what the human does, where the handoff sits, what the escalation path is, and what changes in roles, policies, SOPs and controls as a consequence.

- Design the human-in-the-loop model with the architect: where review is mandatory, where an agent may act autonomously, and how a wrong action is caught and reversed.

Business case and value ownership

- Build the business case: baseline metrics, benefit hypothesis, quantified target, build and run cost, payback and sensitivity — and the honest version of it, including the case against building.

- Establish the baseline before launch. A benefit that cannot be measured against a pre-existing baseline will not be believed, and should not be.

- Own benefits realisation after go-live — report the real number, including when it is below target, and recommend iterate, rebuild or retire accordingly.

Product specification and roadmap

- Write the PRD: problem, users, scope and non-scope, target workflow, acceptance criteria, quality bar, data and system dependencies, non-functional expectations, risks, and success metrics — detailed enough that the architect and engineers can design and build from it without a translation layer.





- Own the roadmap for your function: sequencing, dependencies, release planning, and a defensible view of what comes next and why.

- Partner with delivery daily — make scope calls, resolve ambiguity, unblock, test, and accept. Be available to engineers at the pace they work.

Agent performance management

- Define the quality bar in measurable terms with engineering — what a correct output is, what an acceptable error rate is, and what failure mode is unacceptable at any rate.

- Build and curate the evaluation datasets for your workflows, drawing on real cases from your function.

- Review live agent output on an ongoing basis and run the feedback loop back into engineering.

- Monitor production performance — accuracy and quality, containment or deflection, escalation rate, latency, cost per transaction, and user satisfaction — and drive the iteration cycle.

- Treat a live agent as a product under active management, not a project that was completed.

Adoption and change

- Own adoption as an outcome you are measured on: enablement, training, communications, champions, policy and SOP updates, and management routines.

- Instrument usage and act on what it shows — including the uncomfortable finding that a well-built workflow is being avoided.

- Engage with resistance seriously. Where people are working around the agent, understand why before attempting to change the behaviour.

- Work with function leadership on role definition, capacity and performance expectations as work is redesigned.

Governance and CoE contribution

- Ensure workflows meet responsible AI, data handling, privacy and approval requirements, working with Security, GRC and Legal.

- Feed reusable patterns back into the CoE so that a solved problem in one function does not get re-solved from scratch in another.

📌 Lead AI Engagement Specialist (Pune)
🏢 QAD
📍 Pune