How AI is rebundling work: Redesigning jobs around human judgement, context, and relationships

Digital transformation digitised what exists: Channels, workflows, and dashboards. Intelligence transformation rewires how decisions are made.

It upgrades the enterprise nervous system—from people reading reports to AI agents sensing, reasoning, and acting in loops, says Anees Merchant, EVP and Global Head of Innovation, IP, and Analytics Consulting at C5i in an exclusive interaction with Bizz Buzz

How is “intelligence transformation” different from digital transformation?

Digital transformation digitised what exists: channels, workflows, and dashboards. Intelligence transformation rewires how decisions made. It upgrades the enterprise nervous system: from people reading reports to AI agents sensing, reasoning, and acting in loops.

Unit of value shifts from “app” to decision velocity (time from signal → action). Architecture shifts from systems-of-record to systems-of-reasoning (retrieval, tools, agents, governance). Metrics shift from adoption and SLA to lift (win rate, yield, fraud loss, revenue per visit). Organisation design shifts from functions to human+AI teams (ops copilots, risk sentinels, code robots).

Samples: A lender that underwrites thin-file borrowers using cash-flow AI; a steel mill that lets an optimisation agent set furnace parameters; a retailer that prices by cohort every hour.

Digital made us paperless; intelligence makes us judgment-rich, compounding advantage with every interaction.

What role will AI play in shaping India’s business growth over the next decade?

Three flywheels:

Demand: Vernacular copilots built on Digital Public Infrastructure (Aadhaar, UPI, ONDC, AA) will unlock millions of first-time users and MSMEs, enabling credit, compliance, and commerce in their own language.

Supply: India can become the world’s AI engineering and governance hub. Capabilities in safety, evaluation, data operations, and agent orchestration could be exported just as IT services once were.

Product: “Bharat-first, world-ready” products—agri advisory agents, logistics optimizers, and health triage copilots—built on cost-effective edge devices.

Strategic bets for India Inc: Treat proprietary data as retrieval capital; build corpora moats.

Move from pilots to production agents tied to P&L (collections, underwriting, sales).

Invest in compute + talent locally; don’t rent your advantage.

Co-create with the State on DPI, standards, safety sandboxes, and sectoral datasets.

If we get this right, AI doesn’t just lift GDP; it formalises the informal, compresses distance, and turns every employee into a builder.

What are the Five AI-driven shifts to expect in 2026?

Agentic workflows beat apps: Teams run playbooks where AI calls tools, vendors, and APIs end- to-end. Small, specialized models rise: Domain-tuned models using private retrieval outperform giant general models for critical tasks.

Trust becomes a board metric: Model risk, provenance, and safety incidents are reported in the same manner as cyber breaches.

Synthetic data goes mainstream: It is used to reduce bias, stress-test edge cases, and protect privacy in regulated sectors.

On-device and edge inference expands: Field operations, retail, and manufacturing run AI offline with automatic sync for audit trails.

Signals are already visible: Contact-centre copilots close tickets autonomously; CFO agents reconcile ledgers; quality-control cameras grade products without the cloud; banks ship model cards with their products. 2026 is the year we stop “prompting demos” and start shipping dependable agents tied to outcomes.

Why AI will transform rather than eliminate jobs?

Jobs are bundles of tasks: Every job consists of multiple tasks. Artificial intelligence (AI) separates routine, repetitive tasks from higher-value activities. It then enables organisations to reorganise work around human strengths such as judgment, contextual understanding, creativity, and relationship-building.

The automation paradox: When technology reduces the cost of completing a task, demand for that task often increases. As AI makes work faster and cheaper, organisations typically see: An increase in customer interactions; Greater demand for analysis and insights; Faster operational cycles. Rather than reducing workload, automation can expand overall activity and output.

The rise of “centaur” workflows”: The most effective operating model combines human oversight with AI execution. In these “centaur” workflows: Humans define intent, constraints, and ethical boundaries; AI executes tasks, drafts outputs, analyses data, and monitors processes. This collaborative approach improves speed while maintaining accountability and judgment.

Emerging roles: As work is redesigned, new roles are taking shape, including: Agent Orchestrator; Model Risk Analyst; Data Product Owner; AI Quality Assurance (QA) and Red Team Specialist; Workflow Designer. These roles focus on managing, governing, and optimising AI-enabled systems.

Practical examples: Work transformation is already visible across industries: Sales professionals spend more time on customer discovery while AI handles research and preparation; Claims officers focus on complex or exceptional cases as AI processes routine claims and clears backlogs; Software teams move from writing repetitive code to supervising and governing AI-driven development tools.

Organisations that proactively redesign work are more likely to achieve sustainable growth and competitive advantage.

Elimination happens when organisations refuse to redesign work. Transformation occurs when leaders measure time to insight, errors avoided, and revenue per employee—and reinvest productivity gains into better service and new lines of business.

How can organisations build trust in “black box” AI?

Turn “black box” into glass governance: Risk-tier use cases: different controls for a marketing copilot vs. a credit model.

Data lineage & consent: Track provenance, purpose, and retention; enable audit and user redress.

Explainability where it matters: reason codes, counterfactuals, and uncertainty; aligned to domain needs.

Human-on-the-loop: Escalate on low confidence, novelty, or fairness triggers.

Evaluation as a product: Pre-deployment red teaming; post-deployment drift, bias, and safety monitoring with kill-switches.

Transparent artifacts, including model cards, playbooks, and incident reports, are shareable with regulators and customers.

Sample: A lender shipping adverse-action notices generated from retrieval-grounded explanations; a hospital copilot that shows sources and forces second-checks for high-risk orders. Trust isn’t a feature; it’s an operating system for AI.

Why is understanding AI as critical as reading and writing in the future of work?

Reading and writing enable us to think at scale; AI fluency allows us to build at scale. Every worker will direct a personal stack of agents. Without AI literacy, you become a passenger in systems you can’t question.

Core skills: Problem framing (from vague goal → evaluable task); Tool thinking (what data/tools an agent needs); Skeptical verification (check claims, detect confabulations); Workflow design (guardrails, handoffs, metrics); Data hygiene (privacy, consent, security).

This isn’t about “prompting tricks.” It’s about agency; turning ideas into repeatable workflows. The new divide won’t be coders versus non-coders; it will be directors of intelligence versus consumers of intelligence. Like spreadsheets in the 80s, those who learn early compound their advantage; those who don’t outsource their judgment.