Chapter Four

Talent, Operating Models, and the Agentic Future

^{Talent, Operating Models, and the Agentic Future}

Agentic AI turns automation from a tooling question into an operating model question.

If an AI agent can autonomously resolve incidents, summarise root causes, open tickets, validate data, trigger remediation, and brief stakeholders, then the organisation must ask: what do humans do the next day?

The honest answer has two parts, and stating only the comfortable one costs credibility.

For some roles, it will mean fewer people. For example, where work is genuinely routine and high-volume, agents will reduce headcount, and workforce plans that assume otherwise will not hold up under budget scrutiny.

For the organisation, the more consequential shift is in the mix of remaining work toward judgment, product ownership, engineering depth, data management, risk interpretation, and oversight of autonomous systems.

Two opposite errors should be avoided: treating AI purely as labour arbitrage, which hollows out the expertise needed to govern it; and promising that no role is affected, which erodes trust when reality diverges.

A defensible workforce plan names both effects, funds genuine redeployment for those who can move into higher-judgment work, and is candid about transition for those who cannot.

A Useful Mental Model

For many knowledge-work roles, humans will increasingly supervise, validate, and orchestrate AI-enabled workflows rather than execute every step manually.

The human does less routine e.g., “Human in the Loop” or “Human in the Lead” execution and more exception handling, ethical judgment, escalation, learning and accountability.

The Workforce Shifts are Already Visible

Operating Model Implications

Administrative and coordination work is compressed as agents create drafts, reconcile status, summarize meetings and prepare decisions
Technical depth matters more, not less. In-house teams must understand architectures well enough to challenge vendors and offshore partners
Business-side AI adoption often moves faster than IT-side modernization, creating a two-speed

enterprise and a backlog of governance questions
Location and presence may be revisited for critical AI, data and platform work where rapid collaboration, contextual learning and accountability matter

Product teams need embedded domain, technology, risk, and data skills rather than sequential handoffs between business and IT
Service ownership needs to expand from applications to agents, workflows, data products, prompts and model configurations
Outsourcing models should be rebalanced so internal teams retain enough expertise to audit quality, security, resilience and value
Talent planning should track work outcomes, not just roles, because AI changes tasks faster than job families

Deloitte frames agentic AI as a reimagination of how work is accomplished, requiring organizations to consider cost, workforce adoption and risk management.

McKinsey similarly highlights agentic capabilities such as workflow discovery, asset registries, observability, fine-grained authentication, evaluations and compliance controls. [3][6]

_{The future workforce is not humans versus AI. It is humans responsible for outcomes while AI becomes responsible for execution at scale}

Sumita Dutta, Global IT and Digital Head & CIO, Shell Downstream, Customer Operations, Shell

Are you planning to adopt A.I & Agentic capabilities in the next 12 months that will reduce headcount or shift talent toward higher-judgment activities?

Mostly reduce
Mostly shift
Both equally
Too early to tell