AI Macroeconomic Scenarios: Productivity, Inflation & Labor

An AI macroeconomic scenario is not a point forecast. It is a traceable chain from assumptions about adoption, capability, investment, energy, and labor adjustment to possible outcomes for productivity, inflation, employment, and financial stability.

That distinction matters. AI can raise productive capacity while creating near-term investment demand, sector bottlenecks, job transitions, and asset-price risk. The direction and timing of each channel depend on assumptions that should be visible and testable.

This guide provides an assumption register, three-scenario workbook, transmission matrix, and falsification triggers. It is designed for research and planning, not investment advice.

Start with five transmission channels

Channel	Initial mechanism	Variables to monitor
Productivity	Tasks become faster or higher quality; new tasks become feasible	Output per hour, total factor productivity, adoption by task and firm
Labor	Tasks are automated, complemented, redesigned, or moved	Hours, vacancies, wages, participation, occupational transitions
Prices	Unit costs may fall while investment and constrained inputs may rise	Core inflation, service prices, energy and semiconductor prices, margins
Investment and rates	Compute, power, data centers, software, and training require capital	Capex, credit growth, real rates, depreciation, energy capacity
Financial stability	Expectations and concentrated exposures reprice	Valuations, leverage, collateral concentration, funding and default indicators

The BIS Annual Economic Report 2026 analyzes AI’s potential implications for productivity, inflation dynamics, and the natural rate of interest while stressing uncertainty. The IMF’s 2026 scenario work similarly treats macro outcomes as conditional on adoption, productivity gains, and adjustment.

Build an assumption register before a narrative

Use one row per assumption:

Field	Example question
Assumption ID	Can another analyst cite this exact premise?
Variable	Adoption, capability, capital cost, energy, labor mobility, diffusion lag
Direction/range	What changes, and over which horizon?
Source and date	Which primary source supports the premise?
Mechanism	Through which transmission channel does it matter?
Dependencies	Which other assumptions must also hold?
Observable proxy	What current data could confirm or weaken it?
Falsification trigger	What evidence would make us revise it?
Owner/review date	Who monitors it, and when?

Do not let a model fill unsupported numbers. It may organize sourced assumptions, test consistency, and identify missing links; a human owns the evidence and ranges.

Three useful AI macro scenarios

The scenarios below specify directions, not invented forecasts.

Scenario A: diffusion with complementarity

Assumptions:

• AI capability improves gradually and diffuses across many firms;
• process redesign and worker training accompany software deployment;
• energy and compute capacity expand without a persistent bottleneck;
• competition passes some cost reductions into prices.

Expected chain: business investment rises first; productivity improves with a lag; real income and capacity expand; labor shifts toward complementary tasks; disinflationary supply effects may emerge after near-term investment pressure.

Scenario B: concentrated gains and slow adjustment

Assumptions:

• frontier capability advances, but adoption remains concentrated in large firms and selected occupations;
• organizational change, data quality, and skills constrain diffusion;
• displaced tasks and new roles do not match quickly;
• market power allows some productivity gains to remain in margins.

Expected chain: measured productivity improves unevenly; wage and regional dispersion widen; aggregate inflation effects remain ambiguous; concentrated valuations and credit exposures grow.

The IMF’s analysis of AI and the economics of adjustment highlights that the path of worker and firm adjustment can matter as much as the technology’s technical potential.

Scenario C: investment overshoot and correction

Assumptions:

• expectations move faster than profitable use cases;
• compute, energy, and data-center investment is debt- or valuation-sensitive;
• realized revenue and productivity lag capital deployment;
• a demand, financing, regulatory, or technology shock changes expectations.

Expected chain: investment and constrained-input prices rise; asset and collateral concentration increase; weaker projects face refinancing or impairment pressure; capex and related demand correct.

This scenario does not predict a bubble. It tests balance-sheet and concentration risk. Our GPU-backed lending due-diligence framework provides an asset-level evidence contract for one such exposure.

Copyable scenario workbook

```text SCENARIO NAME: HORIZON:

1. CORE ASSUMPTIONS

• Capability:
• Adoption/diffusion:
• Capital and financing:
• Energy/compute supply:
• Labor mobility/training:
• Competition/policy:

2. TRANSMISSION CHAIN Assumption -> firm behavior -> labor/investment response -> costs/prices -> income/demand -> financial feedback

3. INDICATORS Leading: Coincident: Lagging:

4. DISTRIBUTION Winning sectors/tasks: Exposed sectors/tasks: Regional/firm-size differences:

5. FALSIFICATION Evidence that weakens this scenario: Evidence that strengthens another scenario: Review date and owner: ```

Require every statement in sections 1 and 3 to carry a source, publication date, observation period, and unit.

Red-team the causal chain

For each scenario, ask:

1. Does higher AI investment increase demand before it increases capacity?
2. Are measured productivity gains true efficiency or a change in quality, hours, or composition?
3. Who owns the complementary capital and captures the gains?
4. Can workers and firms move across tasks, sectors, and regions within the stated horizon?
5. Does market power change pass-through from lower unit cost to consumer prices?
6. Are energy, chips, data, or financing binding constraints?
7. Could the same indicator support two different mechanisms?
8. What feedback runs from valuations and credit conditions back to investment?

An AI system should generate counterarguments against the supplied chain, not choose the “most likely” macro future from general knowledge.

Falsification triggers by channel

Scenario claim	Strengthening observation	Weakening observation
Broad productivity diffusion	Gains spread across firm sizes and sectors	Gains remain concentrated despite adoption spend
Labor complementarity	Hours and wages rise in AI-exposed complementary tasks	Persistent displacement without transition
Disinflationary supply effect	Unit costs fall and competition passes savings through	Margins rise while consumer prices do not fall
Investment bottleneck	Capex and constrained-input prices rise together	Capacity expands without sustained price pressure
Financial concentration risk	Leverage and collateral exposure cluster	Exposure is diversified and funded conservatively

The trigger must name a data series, threshold logic, and review cadence in the actual workbook. Avoid universal thresholds divorced from history and measurement error.

Data and model controls

• Freeze the source packet and record retrieval dates.
• Separate observed data from scenario assumptions.
• Preserve units, seasonal adjustment, revisions, and publication lags.
• Do not combine nominal and real series without an explicit deflator.
• Test alternative causal chains and policy responses.
• Record model, prompt, tool, and reviewer versions.
• Label generated prose as synthesis, not evidence.

Use the financial-analysis control framework for claim-level sourcing and calculation checks. If scenarios inform an executive decision, measure review and correction cost using the cost-per-verified-decision framework.

Frequently asked questions

How could AI affect productivity?

AI may automate tasks, complement workers, improve decisions, and enable new products. Aggregate gains depend on adoption, process redesign, skills, competition, and diffusion across firms.

Is AI inflationary or deflationary?

Both channels are plausible. Investment and constrained inputs can add near-term pressure, while higher capacity and lower unit costs can reduce pressure if savings pass through.

Will AI cause unemployment?

AI changes tasks and labor demand, but outcomes depend on complementarity, new work, mobility, training, institutions, and the speed of adjustment. A single forecast hides these dependencies.

What is an AI macroeconomic scenario?

It is a conditional chain linking explicit AI adoption and capacity assumptions to possible productivity, labor, price, investment, and financial outcomes.

Which indicators should an AI macro scenario track?

Track adoption, investment, output per hour, wages and hours by task or sector, vacancies, prices and margins, energy/compute capacity, credit, leverage, and valuations.

How should AI be used in macro scenario analysis?

Use it to structure sourced evidence, test consistency, and generate counterarguments. Do not use unsourced model memory as data or let the model select policy or investments autonomously.