Human investment committees struggle with disagreement because conflict often leads to HR interventions or departing talent. In contrast, AI agents can critique each other's proposals and vote transparently without personal friction. This allows asset allocators to capture the informational benefits of dissent at scale.
AI implementation in asset management will likely preserve top executive committees and entry-level analysts while disrupting middle tiers. Mid-level employees must adapt to new workflows or face team consolidation. Those who remain will pivot toward managing technology and addressing higher-level strategic problems.
Allowing AI agents to update their own code requires strict guardrails and evaluation beyond surface-level performance. Allocators must assess whether an agent's economic reasoning was sound during underlying market conditions, not just whether a trade generated high returns. A meta-agent and investment committee must oversee these code modifications.
Federal taxes took roughly 160 to 170 basis points per year from taxable US equity returns over the last 30 years. Compounded over three decades, this drag eliminates approximately 36% of a portfolio's terminal value. Historically, this modern period represents a relatively low tax environment compared to prior decades.
Private credit returns are taxed as ordinary income, interrupting the compounding process every period. Consequently, a 10% private credit return translates to roughly 7% in equity-equivalent terms. Investors need a 14% to 15% pre-tax yield in private credit to match an ordinary 10% equity return after taxes.
Holding equities across taxable accounts, Roth IRAs, and traditional IRAs produces fundamentally different after-tax cash flows. Because tax rules differ, an optimal asset allocation cannot treat equities identically across all accounts. Allocators must model each asset dynamically within its specific account structure to maximize after-tax wealth.
Timing factor allocations is notoriously challenging, so portfolios should avoid relying on a single positioning methodology. Instead, allocators can diversify across a market similarity model, a factor momentum model, and a valuation dispersion model. Combining these models allows modest, humble tilts while controlling active tracking error risk.
