Innovation Economics
Models for thinking about frontier technology in the real economy.
Most strategy decks reach for vibes. This section reaches for the economist's toolkit and applies it to the decisions you actually face: how much AI capex shifts your steady state, what GPU import dependency does to the rand, when adoption breaks the inflection. Move the sliders. Watch the curves shift. Thirteen artifacts, grouped into four families.
Family I
Macro growth foundations
The stage on which innovation operates. These models shape how boards think about productivity and long-run growth, even when no one names them.
Family II
Innovation-specific mechanics
How innovation actually happens — within firms, across networks, and over time.
-
QUADRANT · MK I × MK II
Schumpeter regimes
Mark I disruptors versus Mark II incumbent labs, plotted across sectors. Mistaking the regime is a category error. SA fintech is mostly Mark I; frontier AI is Mark II.
Open model
-
OVERLAP · UNI × INDUSTRY × GOV
Triple Helix
Etzkowitz and Leydesdorff: innovation emerges from the overlap of three institutional spheres. SA-anchored examples populate each region. The centre is the rarest and most productive intersection.
Open model
-
OVERLAY · COST × ADOPTION
Wright's Law × Rogers Diffusion
Where compute cost decline meets the adoption S-curve. The intersection at each year predicts the inflection where enterprise adoption accelerates.
Open model
-
COMPARATIVE · S-CURVES
Adoption Curves
Four overlaid technology adoption trajectories with annotated milestones. Read off the velocity at which infrastructure, AI, and platforms ship into operating models.
Open model
-
HISTORICAL · J-CURVE OVERLAY
GPT J-curve
Productivity vs years-since-introduction for electricity, IT, and AI on a common axis. The J shape is the point. AI is at year 4; the steep ramp begins after firms reorganise.
Open model
-
CITATION · PRODUCTIVITY
Productivity Deltas
Five canonical AI productivity studies, plotted side by side, with the methodology meta-row that decides whether you should believe them. The 14% bar and the 45% bar are not the same claim.
Open model
Family III
Open-economy frame
Why frontier-tech investment in South Africa is a structurally different problem from the same investment in the United States.
-
INTERACTIVE · OPEN-ECONOMY MODEL
IS-LM-FX for South Africa
AI capex shifts the IS curve right. GPU import dependency drags the BP curve down. Find the new equilibrium for the rand, the repo rate, and output.
Open model
-
SANKEY · CAPITAL FLOWS
Capital Flow Direction
Trailing twelve months of disclosed AI funding mapped from source region to destination bucket. Ribbon widths scale to dollar volume; three patterns are immediately legible.
Open model
Family IV
Poverty & inequality lenses
These should be layered on top of every model above — the diagnostic tools for what the growth and innovation models leave on the floor.
-
SCATTER · GOVERNANCE × VELOCITY
Innovation Phillips Curve
Deployment velocity vs governance overhead across sixteen jurisdictions plus four regional averages. The South Africa dot is the home-jurisdiction reading; the curve is the policy job.
Open model
-
TIME SERIES · PIKETTY
r > g, applied to AI capital
Twin time series: real global GDP growth versus return on a frontier-AI capital basket, 2010–2030. The shaded gap is the wealth-concentration mechanism, drawn for what it is.
Open model
-
MATRIX · ACEMOGLU – RESTREPO
Task displacement matrix
Eight task families plotted on displacement vs reinstatement. The long-run labour share depends on which dominates. AI threatens cognitive mid-skill tasks previous waves left untouched.
Open model
-
HEATMAP · SEN
Capability matrix
Ten AI deployments scored across six capability dimensions. The same tool can expand or contract substantive freedoms — the design choice is the policy.
Open model