AI and the future of work: What to do now
- The Professor
- 4 hours ago
- 6 min read
A clear guide for leaders who need signal, not noise. I show where roles will shrink, where human work stays resilient, and which new AI jobs to hire for next.
If you care about the future of work, here is the practical view. The AI impact on jobs is arriving in three waves at once. Some routine roles are already shrinking as tools reach production. Some hands-on work remains resilient because it lives in messy real-world settings that reward dexterity and trust. A new class of jobs is emerging as AI and future of work regulations evolve, encompassing governance, safety, and orchestration. Please plan for all three at the same time.
The near-term AI impact on jobs
Takeaway: expect vacancy freezes and smaller teams in routine service and clerical tasks first, not instant extinction of entire occupations.
Leaders should start with a simple truth. Replacement typically involves non-replacement of vacancies and gradual team compression. The places to watch over the next twelve months are high-volume, rules-based tasks where the return is already proven.
Tier one customer service for chat and email now deflects routine contacts. Medical transcription and clinical scribing are being integrated into ambient tools within electronic records. Accounts payable and receivable move toward straight-through processing as document AI matures. First-line IT helpdesks are losing resets and simple how-to guidance to virtual agents. The large platforms increasingly handle programmatic media buying and ad operations. Claims teams see drafting and triage support from models. Survey fieldwork loses phone bank hours as synthetic respondents are tested. Low complexity content writing loses value as basic text is generated and A/B tested. Reception-style scheduling moves to voice agents with natural turn-taking. General data entry and form capture shifts across sectors.
What matters now? Press pause on backfilling the most exposed tasks. Move experienced personnel to roles focused on exceptions, quality assurance, and case resolution. Track three key performance indicators (KPIs) every month in service operations, finance operations, and IT support: automation rate, first contact resolution, and audit outcomes. Publish them inside the organisation. Let the data lead the redeployment plan.
Where work stays human for the next decade
Takeaway: Roles in unstructured settings that rely on dexterity and trust have low exposure in the near term, though admin will be automated.
There are many jobs where AI will assist without touching the craft. If your day involves on-site work, requires fine motor control, or relies on care and rapport, your core task is resilient.
· Electricians
· Plumbers and pipe fitters
· Gardeners and horticultural growers
· Early years child care workers
· Hairdressers
· Cooks and chefs
· Physiotherapy assistants and technicians
· Health care assistants and other direct care workers
· Motor vehicle mechanics
· Cabinet makers and joiners
The pattern is consistent. Admin will change first. Quoting, rostering, inventory, notes, and simple diagnostics. The craft stays human. The paperwork does not.
New jobs created by AI and the future of work
Takeaway: regulation and rollouts drive hiring in governance, safety, data, and orchestration, with clear skill paths managers can learn.
Here is the simple logic behind the new jobs AI is creating. As systems transition from pilots to daily use, boards now face legal and brand risks associated with their design and operation. The rise in regulation and assurance creates demand for professionals who can translate high-level rules into practical, day-to-day applications. You need leaders to set policy and steer investment. You need auditors and safety testers to check evidence, run adversarial tests, and sign off on controls. You need incident responders who can contain model failures and explain decisions to customers and regulators. You also need product-facing roles. Retrieval and knowledge engineers make sure the correct facts reach the model. Orchestrators design how agents call tools, apply guardrails, and log what happened. Provenance specialists label media so people can trust what they see. Finally, you need teachers inside the firm. Coaches who help managers use copilots well, measure adoption, and change how teams work. In short, more AI means more governance, more assurance, and better orchestration.
Chief AI Officer or Head of AI Governance. Sets strategy, controls, and reporting to the board.
AI Auditor for ISO 42001. Tests the AI management system and supports conformity assessment.
AI Safety red team specialist. Design adversarial tests before and after release.
AI incident response lead. Builds playbooks for model incidents and coordinates legal and communications.
Synthetic data engineer. Generates privacy-friendly data and validates utility and bias.
RAG or knowledge engineer. Makes retrieval work at enterprise scale and curates ontologies and taxonomies.
Explainability or interpretability engineer. Produces documentation and methods for transparency duties.
Content provenance and deepfake forensics lead. Implements content credentials and verification.
AI agent orchestrator or automation architect. Designs multi-agent systems, tool use, guardrails, and observability.
AI coach and adoption trainer. Builds prompt practice, playbooks, and team capability with measurement
What to do now
Takeaway: treat replace, reshape, and recruit as a single programme with an owner, targets, and reviews.
Map your top twenty tasks by volume and risk across service, finance, IT, claims, and content. For each task, decide to replace, reshape, or augment. Replace means scale automation and stop backfilling. Reshape means redesign the workflow with a human review step and clear escalation. Augment means keeping the craft, but removing paperwork and reworking it. Assign an owner for each decision and publish monthly numbers.
In parallel, recruit for governance and retrieval. Your governance lead sets policy, model lifecycle controls, and incident response. Your retrieval lead reduces time to value in daily work. Book a short capability programme for managers on safe and productive use of copilots, on incident playbooks, and on the basics of ISO 42001 controls. This is not a one-off. Refresh the programme each quarter as tools move.
Evidence block
Methods in short. I synthesised three published videos and their source packs. I coded thirty roles into three groups. Likely to shrink within twelve months. Resilient for five to ten years. Newly created by regulation and rollouts. I scored near-term displacement as high or medium-high and placed roles on a two-by-two that maps adoption momentum against task structure. Where sources diverged, I took the more conservative position.
Wrap up
Action checklist
Map twenty high-volume tasks and mark each as replace, reshape, or augment
Freeze backfills for the five tasks with the strongest business case and redeploy people to focus on quality and exceptions.
Hire one governance lead and one retrieval or knowledge engineer, then add a fractional coach.
Establish three key performance indicators (KPIs) for service and finance operations and publish them monthly.
Develop a concise program on ISO 42001 controls and incident response tailored for managers.
From the professor’s desk
I speak to teams every week who want a straightforward answer on AI and the future of work. There isn't one. The smart move is to hold three ideas at once. Replace where the numbers are clear and the risks are controlled. Reshape where human judgment and care matter. Recruit for the new work around governance and orchestration. That is how you protect people, deliver value, and keep your organisation out of tomorrow morning's news.
References
1. Allstate. Two thousand twenty five. Turns out AI is more empathetic than Allstate’s insurance reps. The Wall Street Journal. https://www.wsj.com/articles/turns-out-ai-is-more-empathetic-than-allstates-insurance-reps-cf5f7c98
2. Brookings Institution. Two thousand twenty five. Is generative AI a job killer Evidence from the freelance market. https://www.brookings.edu/articles/is-generative-ai-a-job-killer-evidence-from-the-freelance-market
3. Business Insider. Two thousand twenty five. AI will augment high skilled jobs but hit clerical work hardest. https://www.businessinsider.com/gen-ai-automate-clerical-work-augment-high-skill-roles-report-2025-8
4. International Labour Organization. Two thousand twenty five. Generative AI and jobs. A refined global index of occupational exposure. https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure
5. McKinsey and Company. Two thousand twenty five. The state of AI. Global survey. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
6. NHS England. Two thousand twenty five. Guidance on the use of AI enabled ambient scribing products in health and care settings. https://www.england.nhs.uk/long-read/guidance-on-the-use-of-ai-enabled-ambient-scribing-products-in-health-and-care-settings
7. OpenAI. Two thousand twenty four. Klarna’s AI assistant case study. https://openai.com/index/klarna
8. The Guardian. Two thousand twenty five. The death of creativity, AI job fears in advertising. https://www.theguardian.com/technology/2025/jun/09/ai-advertising-industry-google-facebook-meta-ads
9. The Wall Street Journal. Two thousand twenty five. AI will soon dominate ad buying. https://www.wsj.com/articles/ai-will-soon-dominate-ad-buying-whether-marketers-like-it-or-not
10. World Economic Forum. Two thousand twenty five. The future of jobs report twenty five. https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf