AI for Portfolio Construction: Claude Tools for Asset Allocation, Optimization, and Risk Budgeting

Portfolio Construction and AI

Portfolio construction translates an investment thesis into a managed set of positions with defined risk characteristics. It spans asset allocation, security selection, position sizing, factor exposure management, and ongoing rebalancing. Claude with ClaudeFinLab handles the quantitative analysis — optimization math, factor decomposition, risk attribution — and the qualitative judgment — position rationale, concentration limits, scenario analysis.

Asset Allocation

• "Build a strategic asset allocation for a $50M endowment with a 5% annual spending rate: investment horizon 30+ years, real return target 4.5% after spending. Proposed SAA: equities 55% (US large 25%, international developed 15%, EM 10%, small cap 5%), fixed income 20% (investment grade corps 10%, TIPs 5%, HY 5%), alternatives 25% (private equity 10%, real assets 8%, hedge funds 7%). Expected return: 7.8% nominal, 5.3% real. Sharpe ratio vs a 60/40 portfolio?"
• "Tactical asset allocation shift: given current macro environment — Fed in cutting cycle, recession probability 30%, credit spreads at historical 40th percentile, equity P/E 19x (fair value 17-19x range) — propose a tactical overlay on the SAA. Overweights/underweights by asset class, magnitude, and the trigger that would reverse each position."
• "Black-Litterman portfolio: I want to express 3 views — (1) US equities will outperform international by 3% (confidence 70%); (2) HY will outperform investment grade by 150bps (confidence 60%); (3) EM equities neutral (no view). Starting weights: market cap weights. Compute the BL posterior expected returns and the implied optimal portfolio weights."

Mean-Variance Optimization

• "Mean-variance optimization for a 10-asset portfolio. Expected returns: [US eq 8.5%, int'l eq 7.8%, EM eq 9.2%, US IG bonds 4.8%, US HY 6.4%, TIPs 4.2%, commodities 5.8%, real estate 7.2%, private equity 11.4%, cash 4.9%]. Correlations and volatilities provided. Compute the efficient frontier: minimum variance portfolio, maximum Sharpe ratio portfolio, and the portfolio targeting 7.5% return. What is the Sharpe ratio at each target return?"
• "Constraints-aware optimization: same 10-asset portfolio but with constraints: (1) No single asset class >30%; (2) Total alternatives (PE + real estate + commodities) ≤ 35%; (3) IG bonds + TIPs ≥ 15% (liquidity reserve); (4) No short positions. How do these constraints shift the optimal portfolio vs unconstrained? What return/risk is sacrificed for each constraint?"
• "Resampled efficient frontier: standard mean-variance is sensitive to input assumptions. Run 500 Monte Carlo simulations of expected returns (±1 standard deviation on each input) and average the optimal weights across simulations. The resampled portfolio is more robust to estimation error. Compare: concentrated MVO portfolio (3-4 large positions) vs resampled portfolio (more diversified). Which is preferable and why?"

Factor Exposure Analysis

• "Decompose this equity portfolio's factor exposures using a 5-factor model (Fama-French 5: Mkt, SMB, HML, RMW, CMA): portfolio beta to each factor — Mkt 1.08, SMB 0.42 (small cap tilt), HML 0.28 (value tilt), RMW 0.18 (profitability tilt), CMA −0.12 (investment tilt). Expected factor return contribution at long-run factor premia: MKT 5.5%, SMB 2.0%, HML 3.0%, RMW 3.8%, CMA 3.5%. Portfolio expected alpha from factor exposures vs pure market exposure?"
• "Factor concentration risk: the portfolio has 68% of active risk in the value factor (HML). Stress test: value underperforms by 4 standard deviations (as in 2020 value drawdown). Portfolio return impact: 0.28 beta × 4σ value drawdown. Is this an acceptable concentration? What trade reduces value exposure while minimizing impact on expected return?"

Risk Budgeting and Risk Parity

• "Risk parity allocation for 4 asset classes: US equities (vol 16%), bonds (vol 6%), commodities (vol 18%), gold (vol 14%). Target: equal risk contribution from each. Step 1 — uncorrelated approximation: weight ∝ 1/vol: equities 1/16=6.25%, bonds 1/6=16.7%, commodities 1/18=5.56%, gold 1/14=7.14%. Normalize: equities 20.9%, bonds 55.8%, commodities 18.6%, gold 23.9% → doesn't sum to 100% because correlation matters. Solve the true risk parity weights iteratively with the correlation matrix."
• "Risk budget allocation: I want to allocate portfolio risk as: equities 50% of risk budget, fixed income 20%, alternatives 30%. Portfolio volatility target: 10%. Compute: maximum equity position size (equity vol 16%, correlation to rest of portfolio 0.6) → marginal risk contribution → required weight to hit 50% of 10% = 5% risk from equities."

Rebalancing Analysis

• "Rebalancing decision: target weights vs current weights after 6 months of drift. [Table: asset class, target %, current %, drift bps]. Trigger rules: rebalance when any asset class drifts >500bps from target. Which positions breach the threshold? Compute the minimum trades to restore targets. Factor in: $50M portfolio × 25bps trading cost on each trade. Total rebalancing cost vs risk benefit of returning to target."
• "Tax-aware rebalancing: the equity portfolio is overweight by $4.2M. Instead of selling appreciated stock (long-term gains tax 23.8%), consider: (1) redirect dividends and new cash contributions to underweight assets for 6 months ($400K/month cash inflow), (2) harvest tax losses in underperforming positions to offset gains. Estimate the tax cost of immediate rebalancing vs gradual drift-and-redirect approach. When does the risk cost of drift exceed the tax benefit of delay?"

Performance Attribution

• "Brinson-Hood-Beebower attribution: portfolio return +12.4% vs benchmark +9.8% (active return +2.6%). Decompose into: (1) Allocation effect — overweight equities (+5% vs benchmark 60% → 65%) during period when equities outperformed bonds by 8%; (2) Selection effect — within equities, stock picking generated +1.8% vs benchmark equity return; (3) Interaction effect. Sum to 2.6%. Which source of active return was the largest contributor?"
• "Factor attribution of active return: active return 2.6%. Factor decomposition: value tilt contributed +1.2% (value outperformed during the period), small cap tilt +0.8%, sector over/underweight contributed +0.4%, stock-specific +0.2%. Is this skill (stock selection) or factor beta (passive exposure to value/small)? What does this imply about manager alpha?"

Related Articles

• Quantitative Finance AI
• Portfolio VaR Analysis
• Wealth Management AI