Augmentation changes the workforce question
Jobs and Skills Australia’s national Generative AI Capacity Study concludes that generative AI is more likely to augment human work than replace it. Adoption remains early and uneven, and the impact will vary by occupation and industry. That shifts the executive question from “How many jobs can AI remove?” to “Which work should change, and what capability will make that change productive?”
Augmentation is not achieved by distributing tools and waiting for efficiency. Tasks must be redesigned, data and permissions aligned, quality checked and workers supported as judgement moves to new points in the process. This is organisational design as much as technology implementation.
Do not confuse model access with capability
A technically strong AI engineer can build integrations and evaluation systems. They cannot, alone, determine which exceptions matter in insurance, what a clinician must verify, why a logistics planner overrides a forecast or which client context changes a professional judgement. That knowledge sits in the operating business.
The differentiating team combines technical leverage with domain depth. It makes implicit rules visible, identifies where AI assistance is useful and preserves human control where consequences or ambiguity are high.
Start with the bottleneck
Before opening a role, map the path from idea to adopted outcome. If teams cannot access reliable data, prioritise data engineering or ownership. If prototypes do not solve a meaningful user problem, strengthen AI product leadership. If solutions stall at deployment, build machine-learning or platform engineering. If technically sound tools are ignored, the gap may be workflow design and change.
This diagnosis prevents an expensive pattern: hiring a senior AI specialist into an organisation that has not provided a problem, product owner, data path or route to production.
Build a capability pod around an outcome
For an initial use case, a small cross-functional pod is often more effective than an isolated centre of excellence. It might include a domain lead, product manager, data or analytics engineer, AI or software engineer and access to security, privacy and change expertise.
The pod should own a measurable workflow outcome and remain close enough to users to learn. Central standards and reusable platforms still matter, but value is created where capability meets a real operating problem.
Look for boundary-crossing evidence
Hybrid roles are valuable when they contain genuine depth, not a collection of fashionable labels. An AI product leader should be able to define evaluation and make commercial trade-offs. A data engineer working on AI should understand retrieval, observability and privacy implications. A domain lead should be willing to turn experience into explicit tests and decisions.
Assessment should explore moments when candidates crossed boundaries: translated a business problem for engineers, challenged an unsafe assumption, changed a workflow after user evidence or brought a system from pilot into normal operations.
Plan to develop the existing workforce
If augmentation is the dominant pattern, capability cannot be supplied by hiring alone. Existing employees hold the context AI systems need, and managers will shape whether new ways of working are trusted. Organisations should define role-specific literacy, safe-use expectations and opportunities to practise with real work.
Specialists then become multipliers. They establish platforms, patterns and coaching that allow other teams to use AI responsibly. This is a more scalable mandate than becoming the internal team that builds every automation itself.
Hire for the organisation you are becoming
The best first hire is the person who unlocks the next stage of maturity. That may be a Head of AI in a platform-ready enterprise, but it may equally be a data product leader, analytics engineer, responsible-AI lead or domain-focused product manager. Title prestige is a poor substitute for sequencing.
Begin with the work, identify the bottleneck and assemble the smallest complete team capable of changing an outcome. In an augmented workforce, human judgement is not the residue left after automation. It is a capability to design deliberately.