← Back to Resource Library Research  ·  AI & the Future of Work

AI, the Future of Work, and impact on Strategic Workforce Planning

A synthesis of what the evidence on AI evolution indicates, distilled from more than 250 episodes of Moonshots from Peter Diamandis and his Moonshot Mates. We set out to investigate three questions to guide us and our clients into the Singularity.

2.5 Reported skills half-life in 2026, against roughly 15 years in 2000
$1.6T AI infrastructure committed or contracted for 2026 to 2028 (on the record)
68% Median cost-per-task reduction observed, agent versus human
11% Organisations with a named owner of the AI workforce shift
Method

How this research was put together

250+

This is a synthesis, not an opinion piece. It draws on more than 250 episodes of Moonshots from Peter Diamandis and his Moonshot Mates, a long-running series of on-record conversations with frontier-lab leadership, investors, and innovators about where technology and work are heading.

We reviewed all of the material available, collected the recurring claims, data points, and forecasts, to distill the data needed to answer our three research questions. Figures cited are drawn from on-record statements and published forecasts; where a figure is a projection rather than a measurement, it is flagged as such.

Go to the presentations ↓
Scope

Three questions we set out to investigate

The research is organised around three questions that will guide us in recommending next steps.

Question 1

The impact of AI on Strategic Workforce Planning

  • How will work be organised in the future
  • How will the workforce be planned
  • What is the impact on SWP
Question 2

The timeframe for AI progression

  • On what horizon does capability actually advance
  • How fast is capital being committed
  • What physically constrains the pace
Question 3

The impact on the workforce itself

  • How does society prepare for it
  • What does the path toward general intelligence imply
  • What are the impacts, and how does it vary by geography
Findings

What the evidence indicates

Here, an analysis of the Moonshots material against the three questions. The figures below come from statements and published forecasts; projections are marked as projections, not measurements.

Question 1  ·  AI and Strategic Workforce Planning

Work is being re-sourced, not simply automated

The recurring pattern across the material is that roles tend to persist while the tasks inside them shift toward agents. Sourcing widens from two options (build, buy) toward four (build, buy, borrow, bot), which puts agents into the capacity model as a first-class line. Because skills decay faster, the material points to workforce planning moving from an annual exercise toward a shorter, recurring cadence, and most organisations have not yet assigned clear ownership of the shift.

The deeper change the sources describe is in what is scarce. Intelligence and software, once expensive and rationed, are becoming a commodity that is available on demand at a falling cost. When the capability to produce analysis, code, and content is abundant, the constraint moves elsewhere: to judgement, to the design of the work itself, and to how fast an organisation can put that capability to use. For Strategic Workforce Planning this reframes the core question from how many people are needed toward which capabilities to own, which to rent, and which to automate, and it shifts planning toward outcomes and capacity rather than headcount. It also raises the broader question the material returns to often: how education, institutions, and labour markets adjust when the supply of cognitive work expands faster than demand for it historically has.

37%Share of task volume shifting to agents by 2027Projection
4Sourcing options per role, up from 2
11%Organisations with a named owner of the shift
Question 2  ·  The timeframe for AI progression

The horizon is short, and the constraints are physical

On-record statements from frontier-lab leadership tend to cluster capability milestones within the second half of this decade, though estimates vary widely by source and should be read as forecasts rather than certainties. What is less speculative is the pace of capital and the physical limits around it: large sums are being committed years ahead, while energy and infrastructure are repeatedly cited as the binding constraint on how quickly that capital can be deployed.

A recurring caution in the material is that these thresholds tend to be clear only in hindsight. Because progress compounds, the point at which a capability crosses from novelty to norm is rarely obvious while it is happening, and is usually named after the fact. The practical conclusion the sources draw is to plan for the trend rather than for a date: treat the direction as reliable even where the timing is not, and assume capability tends to arrive before organisations are ready to absorb it.

$1.6TAI infrastructure committed or contracted, 2026 to 2028
2027 to 2029Range of capability and parity windows cited across sourcesProjection
EnergyMost-cited constraint on deployment pace
Question 3  ·  The impact on the workforce itself

Skills, tasks, and cost move fastest; geography sets the pace

At the task level, the material reports steep cost reductions where work is well defined, tapering toward zero where judgement dominates. At the skill level, the half-life of technical skills is reported to have fallen sharply, pushing value toward judgement, orchestration, and trust. The path toward more general capability is treated as directional rather than dated.

As routine cognitive work is commoditised, the sources describe value migrating to the capabilities that do not compress: judgement under uncertainty, taste, trust, and the orchestration of people and machines. The cost story compounds this, with inference and model costs reported to fall by large multiples year on year, so the economics favour organisations that move early. And because compute, energy, and talent concentrate in a few geographies, the same shift arrives on different timelines by country, which the material frames as a question of access and readiness rather than a single global event.

68%Median cost-per-task reduction, agent versus human
2.5 yrReported skills half-life in 2026, vs ~15 years in 2000
US / CN / EUThree regions on divergent compute, energy, and talent paths
Research outputs

Five decks, one per area of investigation

Each deck explores one part of the research in depth, drawn from the same source material. Read in order, they move from the global picture down to workforce-level detail.

Part 1  ·  8 slides

Geopolitical Lens

How compute, energy, and talent concentrate across the US, China, and the EU, and what that implies for the pace of change by country.

View Presentation →
Part 2  ·  9 slides

Capital Lens

Where AI capital is flowing, how fast it is being committed, and the gap between infrastructure spend and workforce investment.

View Presentation →
Part 3  ·  8 slides

Skills Atlas

How fast skills decay, which compound and which fade, and how tasks are affected across functions.

View Presentation →
Part 4  ·  8 slides

Agent Playbook

How work gets re-sourced across build, buy, borrow, and bot, with the cost economics examined task by task.

View Presentation →
Part 5  ·  8 slides  ·  Synthesis

Board Briefing

The synthesis: the headline numbers and their implications for workforce planning, gathered in one place.

View Presentation →
Where this leads us

From findings to focus

Research is the input, not the conclusion. We apply the same discipline to our own work that the evidence implies for any organisation. For every tool, framework, and service, one decision: start, increase, maintain, decrease, or stop.

Start
Begin
  • A quarterly workforce-planning cadence
  • The Chief of Work and Chief Digital Orchestrator roles
  • First agent fleets, with agents as a planned line in capacity
Increase
Do more
  • Agent and augmentation capability across OXYGEN AI, OXYGEN STAR, and OXYGEN PLAY
  • The bot dimension of TSCM: build, buy, borrow, bot
  • Job and task redesign: human, augment, automate
Maintain
Hold steady
  • The TSCM framework core, which the evidence still supports
  • SWP fundamentals and domain depth where they still compound
Decrease
Taper
  • Headcount-led capacity models
  • One-off, annual-only planning cycles
Stop
End
  • Treating AI as automation only, rather than re-sourcing
  • Over-investing in skills the evidence shows decay fastest
  • Planning the workforce without agents in the model

These are working positions, not fixed conclusions. We keep publishing the methodology and underlying summaries as the research is updated, so the basis for each one stays on the record.

Continue

From Research to Workforce Plan

The decks hold the detail and the sources. If you want to talk through what the evidence means for your organisation, our team is one call away.

View the Decks Get in Touch