Andrew Templeton
Group CTO of a $1B/yr retail portfolio. Fifteen years operating.
AI is a portfolio decision, not a tools decision.
Most operators ship pilots. P&L alpha comes from pricing AI like an asset class. The allocation discipline is invariant; AI just expanded the menu.
15 yr operating · 3 vehicles (PE · advisory · consulting) · ~$1B/yr revenue managed
The System
Five steps, applied to every operating decision. Allocate is the keystone - everything else loads from there.
“The most coherent operating philosophy I have encountered for applying AI to a business without losing rigor or control.”
James Garvey, CEO
Your business is a directed graph →
Your P&L is a graph of nodes and edges - every cost center, revenue line, and causal link between them. Walk it and you find the mispriced edges where value is leaking. Three frameworks show you how.
Price the opportunity in dollars, not vibes →
Every operating opportunity has a return distribution, an error cost profile, and a Sharpe ratio - you just haven't calculated them yet. Five interactive tools do the math before you write the code.
Construct the portfolio →
Once you've priced the instruments, construct the portfolio - rank by risk-adjusted return, map your tolerances, handle correlations between bets. The same efficient frontier math your CFO uses for physical capital, applied to the CTO's roadmap.
Systems that converge by construction →
The goal is convergence by construction - design the system so improvement is a mathematical property, not an aspiration. Ratcheted quality gates and graduated autonomy compress execution risk into something you can manage.
Build the appreciating asset →
Models depreciate through distribution shift. Verifiers, labeled corpora, and institutional rubrics appreciate through operating use - one of the only capital asset classes that does. The framework tells you which side deserves your next dollar.
The version before it gets cleaned up
The frameworks above are the polished output. What I send is the work underneath: the buy-build-borrow calls, the margin math, and the case where a framework broke against a real P&L and had to be rebuilt. You get the operating judgment, not the press release. A few times a quarter, when one is worth the read.
Writing
All essays →Frameworks
Concept graphs
All of this is one graph. The curated layer up top is hand-built; the dense graphs below are the working substrate. They cross-reference. Eventually they live in one searchable space.
Curated
Dense
Larger working substrates. Each node is a lesson; edges are prerequisites and cross-graph references. Less polished than the curated layer - this is how I keep my own map.
The working notes, ongoing
New frameworks and the real operating decisions behind them: the margin math, the calls, the rebuilds. A few times a quarter, when one is worth the read.