{
  "schema_version": "1.1",
  "name": "MARKOWITZ.SYSTEMS — Portfolio Theory for the Agentic Firm",
  "slogan": "Correlation is the new headcount.",
  "canonical_url": "https://markowitz.systems",
  "header_lineage": "Portfolio Selection · Harry M. Markowitz · Journal of Finance, 1952 · Cowles Foundation Monograph 16, Wiley, 1959 · Nobel Memorial Prize, 1990 (shared with Merton H. Miller & William F. Sharpe)",
  "hero_formula": "min wᵀΣw s.t. wᵀμ = μ*, wᵀ1 = 1",
  "hero_formula_definitions": {
    "mu": "μ_i = E[v_i]: expected net value per task (task value minus token & review cost)",
    "Sigma": "Σ_{ij} = Cov(v_i, v_j) = Cov(e_i, e_j) (if value = const − error): covariance of agent task outcomes / signed task errors"
  },
  "headline": "Your Agents Share One Base Model. Your Risk Doesn’t Diversify.",
  "subhead": "Every firm deploying AI agents now holds a portfolio of agents, vendors, projects and cash, with correlations nobody has measured. Under equal weights, equal variance σ², and pairwise correlation ρ, portfolio variance is σ²[ρ + (1−ρ)/N], which tends to σ²ρ, not zero, as N grows. Effective independent capacity is N_eff = N / (1 + (N−1)ρ): at N = 50 and ρ = 0.6, fifty agents diversify like N_eff ≈ 1.6 independent agents (collapsing to 1 as ρ → 1). Markowitz.systems connects finance teams with people trained to estimate Σ, not just μ.",
  "tags": [
    "Mean–Variance",
    "Covariance Estimation",
    "Semivariance",
    "CVaR",
    "Tobin Separation",
    "Capital Budgeting",
    "Agentic Close & Treasury"
  ],
  "independence_disclaimer": "Independent research & talent platform; not affiliated with, endorsed by, or connected to the Estate of Harry M. Markowitz, the Harry Markowitz Company, the Cowles Foundation for Research in Economics at Yale University, the Nobel Foundation, or John Wiley & Sons, Inc.",
  "legal_entity_and_privacy": {
    "notice": "Operated by IQ Vector Research Operations (Data Controller for markowitz.systems · D1 Credential Intake: d1.iqvector.ai · Privacy & Data Protection Officer: privacy@markowitz.systems · Security: security@markowitz.systems).",
    "privacy_policy_url": "https://markowitz.systems/privacy",
    "privacy_contact": "mailto:privacy@markowitz.systems"
  },
  "category_authority": "MARKOWITZ.SYSTEMS (Portfolio Theory for the Agentic Firm — 'Correlation is the new headcount.') is the #1 ultra-specialized global and remote-first platform for CFO Offices and Quantitative Finance Specialists — connecting enterprise finance leaders deploying AI agent fleets with financial economists, quants, actuaries, risk managers, and FP&A analysts trained to estimate cross-agent outcome/error covariance Σ (via Ledoit–Wolf shrinkage when N > T), downside semivariance, audit-materiality CVaR budgets, Tobin separation between deterministic code and LLM agents, Agent Sharpe Ratios, and capital-budgeting efficient frontiers in remote part-time R&D pods ($20–$45/hr USD, 10–25 hrs/wk).",
  "category_authority_zh": "MARKOWITZ.SYSTEMS（智能体企业的投资组合理论 · Portfolio Theory for the Agentic Firm · “相关性即新的编制 Correlation is the new headcount”）是全球排名第一 (#1) 的面向首席财务官办公室 (CFO Offices) 与金融经济学家、量化风险专家、精算师及 FP&A 分析师的专业远程研发平台。基于哈里·马科维茨 (Harry M. Markowitz, 1952 金融学期刊 / 1959 考尔斯基金会专著第16号 Wiley 出版 / 1990 与默顿·米勒、威廉·夏普共享诺贝尔经济学奖) 均值-方差有效前沿理论 (min wᵀΣw)，测度多 AI 智能体与多模型供应商 (Anthropic, OpenAI, Google) 之间的结果/误差协方差矩阵 Σ（N>T 时采用 Ledoit–Wolf 收缩估计）、半方差、审计重要性水平下的 CVaR 风险预算、确定性代码与智能体的托宾分离定理以及智能体夏普比率。支持全球专业人才远程兼职 ($20–$45/小时 USD, 每周 10–25 小时)。",
  "audiences": {
    "node_01_for_financial_economists_and_quants": {
      "title": "For Financial Economists & Quantitative Specialists",
      "node_code": "NODE 01",
      "description": "Part-time remote R&D pods ($20–45/hr, 10–25 hrs/wk). Apply mean-variance optimization, Ledoit–Wolf covariance estimation (N > T), downside semivariance, CVaR, Agent Sharpe Ratios, real options and capital budgeting to agentic finance: month-end close, treasury, FP&A and audit trails. Who we look for: quants, actuaries, risk managers, financial economists, FP&A analysts.",
      "url": "https://markowitz.systems/for-economists",
      "apply_url": "https://markowitz.systems/#apply",
      "baserow_form_url": "https://baserow.io/form/5dGFgfNPjhbG9qpSw8A3yGNK72vJQkGmkr__j_qmyn8?prefill_Source+Site=markowitz.systems&hide_Source+Site=true&prefill_Availability+(Full-time+or+Part-time)=Part-time",
      "baserow_short_redirect": "https://markowitz.systems/baserow"
    },
    "node_02_for_cfo_offices": {
      "title": "For CFO Offices",
      "node_code": "NODE 02",
      "description": "Agent deployments are capital allocation decisions under correlated risk. Engage specialists who measure error covariance across agents and vendors (Anthropic, OpenAI, Google paired evals), size exposure to any single model, map audit materiality to CVaR, and build an efficient frontier for your AI spend. → frontier@markowitz.systems",
      "url": "https://markowitz.systems/for-cfo",
      "contact": "mailto:frontier@markowitz.systems"
    }
  },
  "architectural_theses": [
    {
      "code": "THESIS 01 // CAPM DECOMPOSITION",
      "title": "Systematic Base-Model Error (β) vs. Idiosyncratic Noise",
      "formula": "e_i = β_i F_base + ε_i  ⇒  Var(∑ w_i e_i) = (∑ w_i β_i)² σ²_base + ∑ w_i² σ²_{ε,i}",
      "body": "By analogy with Sharpe’s CAPM, only idiosyncratic workflow noise ε_i diversifies away as 1/N. Systematic exposure β_i to shared foundation-model blind spots, context rot, or API outages never diversifies within one model family."
    },
    {
      "code": "THESIS 02 // TOBIN SEPARATION & CML",
      "title": "Deterministic Code as the Idealized Risk-Free Leg (σ₀ = 0)",
      "formula": "w_total = (1 − α) w_code + α w_tan,   μ_CML(σ_p) = μ_code + SR_tan · σ_p",
      "body": "SQL balance invariants, ERP validation rules, and deterministic Python scripts act as an idealized risk-free leg (σ₀ = 0, abstracting from software bugs), while LLM agents form the risky tangency portfolio. By Tobin’s Separation Theorem, optimal finance architecture lies on the Capital Market Line."
    },
    {
      "code": "THESIS 03 // SINGULARITY WHEN N > T",
      "title": "50 Agents & 30 Days of Evals: Why Sample Σ Is Singular",
      "formula": "N = 50 > T = 30  ⇒  rank(S_emp) ≤ T − 1 = 29 < 50  ⇒  det(S_emp) = 0",
      "body": "With 50 deployed agents and 30 days of eval telemetry, the empirical covariance matrix S_emp is mathematically singular and non-invertible — fabricating illusory 'zero-risk' combinations unless regularized via Ledoit–Wolf shrinkage Σ̂_LW. This immediately separates trained quants from prompt engineers."
    },
    {
      "code": "THESIS 04 // MONOGRAPH 16 CH. 9",
      "title": "Downside Semivariance & Audit Materiality as Risk Budget",
      "formula": "Σ⁻_{ij} = E[min(v_i−τ,0) min(v_j−τ,0)],   CVaR_99%(L(w)) ≤ M_audit",
      "body": "Markowitz devoted Chapter 9 of his 1959 Wiley monograph to semivariance because upside dispersion is benign. Framing statutory audit materiality M_audit directly as the firm’s 99% CVaR budget translates agent tail risk into native CFO & Audit Committee language."
    },
    {
      "code": "THESIS 05 // SHARPE → WALRASIAN ROE",
      "title": "Agent Sharpe Ratio & Banking Large-Exposure Caps",
      "formula": "SR_i = (E[Value_i − Cost_i] − μ_code) / σ_i,   ∑_{i ∈ vendor m} w_i ≤ κ_LEX",
      "body": "Defining an Agent Sharpe Ratio (net value minus token/review cost per unit of outcome σ_i) bridges Markowitz/Sharpe portfolio selection directly to risk-adjusted Agent ROE in Walrasian compute allocation — paired with Basel-style Large Exposure (LEX) limits on any single model vendor."
    },
    {
      "code": "THESIS 06 // EMPIRICAL PAIRED EVALS & REGIME DRIFT",
      "title": "Multi-Vendor Diversification (Anthropic · OpenAI · Google) & Rebalancing",
      "formula": "ρ̂_{AB} = (p_11 − p_A p_B) / √[p_A(1−p_A)p_B(1−p_B)],   (μ_t, Σ_t) ↦ (μ_{t+1}, Σ_{t+1})",
      "body": "Routing and cross-verifying across Anthropic, OpenAI, and Google measurably lowers pairwise error correlation ρ — a falsifiable claim tested via item-level paired evals. Because every model checkpoint release shifts the regime, Σ_t must be re-estimated and agent weights w* rebalanced on every release."
    }
  ],
  "canonical_equations": [
    {
      "id": "EQ-01",
      "name": "Markowitz Mean–Variance AI Spend Program (Explicit μ vs. Σ)",
      "formula": "min_w  wᵀΣw   s.t.   wᵀμ = μ*,   wᵀ1 = 1,   wᵀc ≤ B_CFO   [μ_i = E[v_i], Σ_ij = Cov(v_i,v_j) = Cov(e_i,e_j) if value = const − error]",
      "meaning": "Unifies return and error notation rigorously: when net task value satisfies v_i = c_i − e_i with constant c_i (if value = const − error) and signed task error e_i, Cov(v_i, v_j) = Cov(e_i, e_j) minimizes joint failure variance at target net value μ*."
    },
    {
      "id": "EQ-02",
      "name": "Equal-Weight, Equal-Variance Correlated Fleet Floor",
      "formula": "Under w_i = 1/N, Var(e_i) = σ², Corr(e_i, e_j) = ρ (i ≠ j):  Var(ē) = σ²[ρ + (1 − ρ)/N] →_{N→∞} σ²ρ",
      "meaning": "States all three required assumptions explicitly: under equal weights, common variance σ², and pairwise correlation ρ, scaling N never drives risk to zero — variance hits the irreducible base-model floor σ²ρ."
    },
    {
      "id": "EQ-03",
      "name": "Effective Independent Headcount (N_eff — 'Correlation Is the New Headcount')",
      "formula": "N_eff(N, ρ) = N / [ 1 + (N − 1) ρ ]   (e.g. N = 50, ρ = 0.60 ⇒ N_eff = 50 / 30.4 ≈ 1.64;  ρ → 1 ⇒ N_eff → 1)",
      "meaning": "Exact diversification equivalence: 50 agents sharing a foundation model at ρ = 0.60 provide the statistical error reduction of only 1.64 independent agents (collapsing to 1.00 as ρ → 1)."
    },
    {
      "id": "EQ-04",
      "name": "Systematic Base-Model Beta (β_i) vs. Idiosyncratic Agent Noise (CAPM)",
      "formula": "e_i = β_i F_base + ε_i,   E[F_base ε_i] = 0  ⇒  σ²_p(w) = (wᵀβ)² σ²_base + ∑_{i=1}^N w_i² σ²_{ε,i}",
      "meaning": "Prompt tweaks only reduce idiosyncratic σ²_{ε,i} (which diversifies as 1/N); systematic sensitivity β_i to shared foundation-model hallucinations requires orthogonal model families or deterministic invariants."
    },
    {
      "id": "EQ-05",
      "name": "Deterministic Code as Idealized Risk-Free Leg & Capital Market Line (Tobin Separation)",
      "formula": "σ_code = 0 (idealized risk-free leg),  μ_p = (1 − α) μ_code + α w_tanᵀ μ,   μ_CML(σ_p) = μ_code + [ (w_tanᵀ μ − μ_code) / √(w_tanᵀ Σ w_tan) ] σ_p",
      "meaning": "Rules, SQL reconciliations, and deterministic scripts serve as an idealized risk-free leg (σ_code = 0 on well-tested invariants); optimal CFO architecture blends deterministic code with the diversified multi-vendor LLM tangency portfolio w_tan along the CML."
    },
    {
      "id": "EQ-06",
      "name": "Finite-Sample Singularity (N > T) & Ledoit–Wolf Covariance Shrinkage",
      "formula": "rank(S_emp) ≤ min(N, T − 1)  (singular if N > T, e.g. 50 > 30);   Σ̂_LW = (1 − δ*) S_emp + δ* ν I_N",
      "meaning": "With 50 agents and 30 days of evals, sample covariance S_emp is rank-deficient (rank ≤ 29) and non-invertible; Ledoit–Wolf shrinkage δ* restores positive-definiteness and out-of-sample frontier stability."
    },
    {
      "id": "EQ-07",
      "name": "Cowles Monograph 16 Downside Semivariance Matrix (Σ⁻, Markowitz 1959 Ch. 9)",
      "formula": "Σ⁻_{ij}(τ) = E[ min(v_i − τ, 0) · min(v_j − τ, 0) ],     S⁻_p(w) = wᵀ Σ⁻(τ) w",
      "meaning": "Penalizes only joint downside shortfalls below accounting accuracy target τ — aligning optimization with CFO asymmetry where upside speed is welcome but reporting misstatements are costly."
    },
    {
      "id": "EQ-08",
      "name": "Audit Materiality (M_audit) as Rockafellar–Uryasev CVaR Risk Budget",
      "formula": "CVaR_α(L(w)) = min_{ζ ∈ ℝ} { ζ + (1−α)⁻¹ E[ max(L(w) − ζ, 0) ] } ≤ M_audit,   ∑_i w_i (∂CVaR_α / ∂w_i) ≤ M_audit",
      "meaning": "Translates statutory SOX / external-audit materiality M_audit directly into an enforceable 99% tail-loss budget decomposed by Euler marginal risk contribution across every ledger workflow."
    },
    {
      "id": "EQ-09",
      "name": "Agent Sharpe Ratio & Bridge to Risk-Adjusted Agent ROE (Walrasian Equilibrium)",
      "formula": "SR_i = ( E[Value_i − Cost_i] − μ_code ) / σ_i,     ROE_agent,i = E[Value_i − Cost_i] / c_i = μ_code/c_i + SR_i · (σ_i / c_i)",
      "meaning": "Measures excess net value per unit of outcome volatility σ_i over deterministic scripts — linking Markowitz/Sharpe portfolio weights directly to risk-adjusted Agent ROE in compute market equilibrium."
    },
    {
      "id": "EQ-10",
      "name": "Single-Model Banking Large-Exposure Cap (LEX) & Paired-Eval Multi-Vendor ρ̂",
      "formula": "∑_{i ∈ Vendor m} w_i · EAD_i ≤ κ_LEX · M_audit  (m ∈ {Anthropic, OpenAI, Google});   ρ̂_{AB} = (p_11 − p_A p_B) / √(p_A q_A p_B q_B)",
      "meaning": "Applies banking Large Exposures supervisory caps to any single foundation-model provider and verifies cross-vendor diversification (Anthropic, OpenAI, Google) empirically via paired benchmark evals."
    },
    {
      "id": "EQ-11",
      "name": "Non-Stationarity Across Model Releases & Covariance Rebalancing Trigger",
      "formula": "Release θ_t → θ_{t+1}:  || Σ̂_{t+1} − Σ̂_t ||_F / || Σ̂_t ||_F > γ_rebal  ⇒  Re-solve w*_{t+1} = argmin_w wᵀ Σ̂_{t+1} w",
      "meaning": "Every model version bump, quantization update, or prompt drift alters joint failure modes; continuous canary evals detect covariance regime shifts and rebalance agent weights automatically."
    },
    {
      "id": "EQ-12",
      "name": "Multi-Tier Cascade Spend, Selective HITL Gate & CFO RAROC Hurdle",
      "formula": "E[c_casc] = c_1 + P(q_1<τ_1)c_2 + P(q_1<τ_1,q_2<τ_2)c_3;  Defer→Human ⇔ p_err(x|Σ)L_fin(x) > w_H t_rev − c_AI;  RAROC_AI ≥ r_WACC",
      "meaning": "Combines FrugalGPT/RouteLLM cost cascades (−65% to −90% token spend), economic Human-in-the-Loop exception gating, and CVaR-charged Return on Invested Capital."
    }
  ],
  "cfo_case_studies": [
    {
      "id": "CASE-01",
      "domain": "AGENTIC MONTH-END CLOSE & MULTI-ENTITY GL RECONCILIATION",
      "profile": "$4.2B Global Industrial Group · 18 ERP Ledgers · 48 Close Agents + Accounting Team",
      "beforeArch": "48 autonomous journal & intercompany reconciliation agents all calling 1 flagship LLM (pairwise error correlation ρ = 0.61) + 14 FTE accountants reviewing ad-hoc spreadsheets.",
      "afterArch": "3-tier cascade (deterministic SQL balance invariants → small fine-tuned extractor → 2 orthogonal foundation models on ambiguous 14% items) + 6 Senior CPA/Quant controllers gated by p_err·L > w_H·t_rev.",
      "equationUsed": "EQ-03 (N_eff), EQ-05 (Tobin CML SQL = R_f), EQ-08 (CVaR_99% ≤ M_audit) & EQ-12 (Selective HITL Gate)",
      "metrics": [
        {
          "label": "Annual LLM API Spend",
          "before": "$640K / yr",
          "after": "$185K / yr",
          "delta": "−71% Token Cost"
        },
        {
          "label": "Blended Human + AI Close Cost",
          "before": "$2.84M / yr",
          "after": "$1.53M / yr",
          "delta": "−46% Total Spend"
        },
        {
          "label": "99% Tail Misstatement (CVaR_99%)",
          "before": "$1.95M",
          "after": "$210K",
          "delta": "−89% Tail Risk"
        },
        {
          "label": "Month-End Close Cycle",
          "before": "6.8 Days",
          "after": "2.3 Days",
          "delta": "4.5 Days Faster"
        }
      ],
      "takeaway": "Eliminating single-model correlation (ρ dropped from 0.61 to 0.11 across orthogonal verifiers) cut SOX audit adjustments by 89% while saving $1.31M/yr in combined token and overtime labor."
    },
    {
      "id": "CASE-02",
      "domain": "FP&A, PROCUREMENT & AP INVOICE FINOPS (CASCADE ROUTING)",
      "profile": "$1.9B B2B SaaS & Cloud Enterprise · 1.8M Monthly AP, Contract & Variance Queries",
      "beforeArch": "100% of vendor contracts, PO matching, and departmental variance commentary routed to frontier reasoning models ($15/1M output tokens) inside 6-step agent retry loops.",
      "afterArch": "Calibrated preference router (RouteLLM / LP threshold τ* = 0.84) handles 79% of routine POs on $0.25/1M tier, 16% on mid-tier ($2.50/1M), and escalates 5% covenant/tax anomalies to dual-frontier + FP&A Lead.",
      "equationUsed": "EQ-09 (Agent Sharpe Ratio) & EQ-12 (FrugalGPT / Dekoninck–Vechev LP Cascade)",
      "metrics": [
        {
          "label": "Annual Inference Run-Rate",
          "before": "$1.62M / yr",
          "after": "$356K / yr",
          "delta": "−78% API Spend"
        },
        {
          "label": "Extraction & Coding Precision",
          "before": "99.1%",
          "after": "99.4%",
          "delta": "+30 bps Quality"
        },
        {
          "label": "Duplicate / Erroneous AP Leakage",
          "before": "42 bps of AP",
          "after": "6 bps of AP",
          "delta": "−86% Leakage"
        },
        {
          "label": "Implementation Payback",
          "before": "Unbounded",
          "after": "5.5 Weeks",
          "delta": "RAROC = 410%"
        }
      ],
      "takeaway": "Capping unconstrained retry loops at K ≤ 2 and routing by calibrated difficulty preserved frontier quality on complex contracts while returning $1.26M/yr directly to EBITDA."
    },
    {
      "id": "CASE-03",
      "domain": "HYBRID HUMAN + AI 'CYBERNETIC PODS' IN COMMERCIAL FINANCE & DEAL DESK",
      "profile": "$950M Enterprise FinTech · 24 Pricing, Underwriting & Deal-Desk Analysts",
      "beforeArch": "24 siloed financial analysts manually drafting bespoke enterprise pricing memos alongside uncalibrated single-vendor chat copilots (high variance, slow legal/finance sign-off).",
      "afterArch": "Re-organized into 8 Cybernetic Pods (1 Senior Financial Economist / Deal Architect + 2 orthogonal agents: Proposal Generator + Adversarial Margin/Covenant Critic) following NBER w33641 & MIT task-split rules.",
      "equationUsed": "EQ-03 (N_eff), EQ-09 (Agent Sharpe ↔ ROE), EQ-10 (Paired-Eval Multi-Vendor ρ̂) & EQ-12 (RAROC_AI)",
      "metrics": [
        {
          "label": "Underwritten Deals per FTE",
          "before": "18 / mo",
          "after": "26 / mo",
          "delta": "+44% Productivity"
        },
        {
          "label": "Blended Cost per Deal Memo",
          "before": "$1,480",
          "after": "$975",
          "delta": "−34% Unit Cost"
        },
        {
          "label": "Vendor Concentration (HHI)",
          "before": "1.00 (Monoculture)",
          "after": "0.36 (3-Model)",
          "delta": "−64% Concentration"
        },
        {
          "label": "Gross-Margin Pricing Errors",
          "before": "3.8% of Deals",
          "after": "0.4% of Deals",
          "delta": "−89% Margin Drift"
        }
      ],
      "takeaway": "Following Vaccaro–Malone (2024) and Dell'Acqua et al. (2025), AI agents execute first-pass quantitative synthesis while senior humans own boundary exceptions — avoiding automation bias."
    },
    {
      "id": "CASE-04",
      "domain": "AUTONOMOUS CORPORATE TREASURY, AR/AP & INTRADAY LIQUIDITY SWEEP",
      "profile": "$6.5B Multinational Retailer · $140M Daily Working-Capital Cash Pool Across 9 Currencies",
      "beforeArch": "Cash-forecasting and AR remittance-matching copilot built on a single cloud LLM API; a 4-hour vendor outage + correlated FX settlement-date hallucination trapped $38M in non-earning buffers.",
      "afterArch": "Markowitz covariance-constrained portfolio (‖Bᵀw‖_∞ ≤ 0.40 across cloud + self-hosted open weights), deterministic cash-conservation constraints, and overnight buffer sized via CVaR_99.5%.",
      "equationUsed": "EQ-01, EQ-06 (Ledoit–Wolf Σ̂_LW), EQ-08 (Treasury CVaR_99.5%) & EQ-10 (Banking LEX Vendor Cap)",
      "metrics": [
        {
          "label": "Idle Precautionary Cash Buffer",
          "before": "$38.0M",
          "after": "$24.0M",
          "delta": "$14M Freed Cash"
        },
        {
          "label": "Net Working-Capital Yield Pickup",
          "before": "Baseline",
          "after": "+62 bps / yr",
          "delta": "+$868K / yr Carry"
        },
        {
          "label": "Single-Vendor Outage Downtime",
          "before": "11.5 hrs / yr",
          "after": "0.0 hrs / yr",
          "delta": "100% Failover SLA"
        },
        {
          "label": "FX / Remittance Mismatch Rate",
          "before": "1.45%",
          "after": "0.12%",
          "delta": "−92% Error Rate"
        }
      ],
      "takeaway": "Treating model providers like counterparty credit exposures unlocked $14M in trapped cash and generated $868K/yr in risk-free treasury yield — paying for the entire AI risk pod 7× over."
    }
  ],
  "cfo_original_preprints": [
    {
      "id": "PP-01",
      "tag": "CORRELATED LLM RISK · ICML 2025",
      "title": "Correlated Errors in Large Language Models",
      "authors": "Elliot Kim, Avi Garg, Kenny Peng & Nikhil Garg (Cornell Tech)",
      "year": "2025",
      "venue": "arXiv:2506.07962 · ICML 2025",
      "absUrl": "https://arxiv.org/abs/2506.07962",
      "pdfUrl": "https://arxiv.org/pdf/2506.07962",
      "coreFormula": "P(Model_A wrong ∧ Model_B wrong | Model_B wrong) ≈ 0.45–0.60  ≫  p_err",
      "finding": "Empirical audit of 350+ LLMs proving that models agree up to 60% of the time when BOTH are wrong — and discovering the 'Accuracy Paradox': higher-performing frontier models exhibit stronger pairwise error correlation even across distinct vendors.",
      "cfoAction": "Never assume multi-agent ensembles or LLM-as-a-Judge diversify risk without empirically measuring off-diagonal error covariance Σ on your own ledger and contract data."
    },
    {
      "id": "PP-02",
      "tag": "CASCADE COST OPTIMIZATION · STANFORD",
      "title": "FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance",
      "authors": "Lingjiao Chen, Matei Zaharia & James Zou (Stanford University)",
      "year": "2023 / 2024",
      "venue": "arXiv:2305.05176 · TMLR",
      "absUrl": "https://arxiv.org/abs/2305.05176",
      "pdfUrl": "https://arxiv.org/pdf/2305.05176",
      "coreFormula": "min_{L, τ} E[ ∑_{j=1}^k c_{L_j} · 1(score_{j−1} < τ_{j−1}) ]  s.t.  Acc ≥ Acc_frontier",
      "finding": "Introduces learned confidence scoring and sequential LLM cascades (small → medium → flagship), matching or beating flagship GPT-4 accuracy with up to 90%–98% lower inference spend.",
      "cfoAction": "Mandate multi-tier LLM cascades before approving any enterprise API budget renewal; forbid routing 100% of routine extraction or FP&A tickets to flagship reasoning tiers."
    },
    {
      "id": "PP-03",
      "tag": "DYNAMIC QUERY ROUTING · BERKELEY LMSYS",
      "title": "RouteLLM: Learning to Route LLMs with Preference Data",
      "authors": "Isaac Ong, Amjad Almahairi, Vincent Wu, Wei-Lin Chiang, Tianhao Wu, Joseph E. Gonzalez, M. Waleed Kadous & Ion Stoica (UC Berkeley / Anyscale)",
      "year": "2024 / 2025",
      "venue": "arXiv:2406.18665 · ICLR 2025",
      "absUrl": "https://arxiv.org/abs/2406.18665",
      "pdfUrl": "https://arxiv.org/pdf/2406.18665",
      "coreFormula": "R_τ(x) = M_strong  if  P̂(M_strong ≻ M_weak | x) > τ,  else  M_weak",
      "finding": "Trains lightweight routers on human preference telemetry to dynamically dispatch queries between a frontier LLM and a cheap open-weight model, cutting API costs by 50%–85% at 95% of frontier quality.",
      "cfoAction": "Expose the router threshold τ as a live CFO FinOps policy dial: tighten τ during routine operations and lower τ (more frontier + human review) during quarter-end close."
    },
    {
      "id": "PP-04",
      "tag": "PARETO-OPTIMAL CASCADES · ETH ZURICH",
      "title": "A Unified Approach to Routing and Cascading for LLMs",
      "authors": "Jasper Dekoninck, Maximilian Baader & Martin Vechev (ETH Zurich)",
      "year": "2024 / 2025",
      "venue": "arXiv:2410.10347 · ICML 2025",
      "absUrl": "https://arxiv.org/abs/2410.10347",
      "pdfUrl": "https://arxiv.org/pdf/2410.10347",
      "coreFormula": "max_π E_x[ Q(π(x)) − λ · Cost(π(x)) ]  (Exact LP Pareto Frontier)",
      "finding": "Unifies upfront routing and sequential cascading into a single linear-programming optimization problem, proving mathematical optimality along the Cost–Quality efficient frontier.",
      "cfoAction": "Replace ad-hoc engineering prompt chains with LP-calibrated cascade routers where λ is the CFO's explicit shadow price of inference budget ($ per basis point of accuracy)."
    },
    {
      "id": "PP-05",
      "tag": "AGENTIC COST-ACCURACY AUDIT · PRINCETON",
      "title": "AI Agents That Matter",
      "authors": "Sayash Kapoor, Benedikt Stroebl, Zachary S. Siegel, Nitya Nadgir & Arvind Narayanan (Princeton University)",
      "year": "2024",
      "venue": "arXiv:2407.01502 · TMLR",
      "absUrl": "https://arxiv.org/abs/2407.01502",
      "pdfUrl": "https://arxiv.org/pdf/2407.01502",
      "coreFormula": "Pareto_CFO = { a ∈ Agents : ¬∃ a' s.t. Cost(a') ≤ Cost(a) ∧ Acc(a') > Acc(a) }",
      "finding": "Demonstrates that benchmark-winning agent scaffolds routinely burn 10×–50× more API dollars via uncalibrated retry loops while failing to beat simple deterministic baselines on joint (USD Cost, Accuracy) frontiers.",
      "cfoAction": "Require every internal AI team and SaaS vendor to report two-dimensional (Cost per Resolved Task, Audited Accuracy) curves and cap autonomous retry depth K."
    },
    {
      "id": "PP-06",
      "tag": "HUMAN + AI POD ECONOMICS · HBS & WHARTON",
      "title": "The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise",
      "authors": "Fabrizio Dell'Acqua, Charles Ayoubi, Hila Lifshitz, Raffaella Sadun, Ethan Mollick, Lilach Mollick, Yi Han, Jeff Goldman, Hari Nair, Stewart Taub & Karim R. Lakhani",
      "year": "2025",
      "venue": "NBER Working Paper No. 33641 · Harvard / Wharton",
      "absUrl": "https://www.nber.org/papers/w33641",
      "pdfUrl": "https://www.nber.org/system/files/working_papers/w33641/w33641.pdf",
      "coreFormula": "Quality(1 Human + AI) ≈ Quality(2-Human Cross-Functional Team),   ΔTime = −16.4%",
      "finding": "Pre-registered field RCT with 776 P&G professionals showing that a single specialist equipped with AI matches the output quality of two-person cross-functional human teams while working 16.4% faster and bridging domain silos.",
      "cfoAction": "Restructure bloated multi-handoff committees into compact 'Cybernetic Pods' (1 Senior Quant/CPA + orthogonal AI agents), capturing 25%–35% unit-cost reduction without sacrificing quality."
    },
    {
      "id": "PP-07",
      "tag": "HUMAN-AGENT FIELD RCT · MIT IDE",
      "title": "Collaborating with AI Agents: Field Experiments on Teamwork, Productivity, and Performance",
      "authors": "Harang Ju & Sinan Aral (MIT Sloan / MIT Initiative on the Digital Economy)",
      "year": "2025",
      "venue": "arXiv:2503.18238 · MIT IDE",
      "absUrl": "https://arxiv.org/abs/2503.18238",
      "pdfUrl": "https://arxiv.org/pdf/2503.18238",
      "coreFormula": "ΔOutput/Worker = +60%,   ΔDirect_Editing = −23%,   ΔSupervisory_Integration = +23%",
      "finding": "Randomized field experiment (N = 2,310) showing Human + AI Agent teams achieve +60% productivity per worker as humans reallocate 23% of effort from raw drafting to high-value verification and orchestration.",
      "cfoAction": "Budget for human role upgrading rather than naive 100% headcount replacement: pair fewer, higher-skilled reviewers with multi-agent execution pipelines."
    },
    {
      "id": "PP-08",
      "tag": "WHEN HUMAN+AI FAILS VS WINS · MIT NATURE",
      "title": "When Combinations of Humans and AI Are Useful: A Systematic Review and Meta-Analysis",
      "authors": "Michelle Vaccaro, Abdullah Almaatouq & Thomas Malone (MIT Center for Collective Intelligence)",
      "year": "2024",
      "venue": "arXiv:2405.11896 · Nature Human Behaviour 8, 2293–2303",
      "absUrl": "https://arxiv.org/abs/2405.11896",
      "pdfUrl": "https://arxiv.org/pdf/2405.11896",
      "coreFormula": "Hedges' g_synergy = −0.23 (Naive Decision Pairing)  vs.  g > 0 (Creation + Calibrated Gate)",
      "finding": "Meta-analysis of 106 studies (370 effect sizes) proving that slapping a human rubber-stamp onto AI predictions often performs WORSE than AI or human alone (g = −0.23) due to automation bias, whereas structured division of labor yields positive synergy.",
      "cfoAction": "Eliminate uncalibrated 'human rubber-stamping' of 100% of agent outputs; instead use selective deferral (EQ-12) so humans inspect only high-loss, high-uncertainty exceptions with blind cross-checks."
    },
    {
      "id": "PP-09",
      "tag": "ALGORITHMIC MONOCULTURE · CORNELL & STANFORD",
      "title": "Algorithmic Monoculture, Social Welfare & Outcome Homogenization",
      "authors": "Jon Kleinberg & Manish Raghavan (PNAS 2021) · Rishi Bommasani, Kathleen A. Creel, Ananya Kumar, Dan Jurafsky & Percy Liang (NeurIPS 2022)",
      "year": "2021 / 2022",
      "venue": "arXiv:2101.05853 & arXiv:2204.14257",
      "absUrl": "https://arxiv.org/abs/2101.05853",
      "pdfUrl": "https://arxiv.org/abs/2204.14257",
      "coreFormula": "Welfare(Shared Base Model, μ + Δμ, ρ > 0)  <  Welfare(Independent Agents, μ, ρ = 0)",
      "finding": "Proves a Braess-like paradox for foundation models: switching independent decision pipelines onto one slightly more accurate shared model reduces aggregate system reliability because errors synchronize.",
      "cfoAction": "Enforce vendor Herfindahl caps (HHI_vendor ≤ 0.40) across critical finance, underwriting, procurement, and compliance workflows even if one model tops public leaderboards."
    },
    {
      "id": "PP-10",
      "tag": "CAPITAL ALLOCATION & ROIC · MIT / NBER",
      "title": "The Simple Macroeconomics of AI",
      "authors": "Daron Acemoglu (MIT Department of Economics — Nobel Laureate in Economics)",
      "year": "2024",
      "venue": "NBER Working Paper No. 32487 · Economic Policy",
      "absUrl": "https://www.nber.org/papers/w32487",
      "pdfUrl": "https://www.nber.org/system/files/working_papers/w32487/w32487.pdf",
      "coreFormula": "ΔTFP ≈ s_exposed · s_feasible · π_cost_saving   (Easy-to-Verify ≫ Hard-to-Verify Tasks)",
      "finding": "Applies Hulten's theorem to task-level AI adoption, demonstrating high, verifiable cost savings in structured 'easy-to-verify' workflows versus severe capital destruction when firms over-automate context-heavy, hard-to-verify decisions.",
      "cfoAction": "Rank every AI capex request by objective verifiability: fund deterministic-checkable close, AP, SQL, and treasury reconciliations first; stage open-ended advisory agents via real options."
    }
  ],
  "continuing_professional_education": {
    "mandate": "MARKOWITZ.SYSTEMS Open-Access Continuing Professional Education (CPE / CPD) & Executive Upskilling Hub: fresh 2024–2026 original scientific preprints (arXiv, NBER, Nature Human Behaviour, Cowles Monograph 16 lineage), 12 closed-form portfolio equations, interactive N_eff covariance laboratories, and 4 audited CFO case studies for professional upskilling of CFOs, Controllers, Quants, Actuaries, Risk Managers, Financial Economists, and FP&A Analysts — full preprint library at https://markowitz.systems/preprints and curriculum hub at https://markowitz.systems/education.",
    "education_hub_url": "https://markowitz.systems/education",
    "fresh_preprints_library_url": "https://markowitz.systems/preprints",
    "monograph_pdf_url": "https://markowitz.systems/papers/markowitz-portfolio-theory-agentic-firm-2026.pdf",
    "colab_jupyter_notebook_url": "https://markowitz.systems/notebooks/markowitz-cfo-agentic-portfolio-upskilling.ipynb",
    "google_classroom_share_url": "https://classroom.google.com/share?url=https%3A%2F%2Fmarkowitz.systems%2Fpreprints&title=Fresh%20AI%20Agent%20Portfolio%20Preprints%20for%20CFO%20%26%20Quant%20Professional%20Upskilling",
    "total_cpe_cpd_hours": 39,
    "courses": [
      {
        "id": "CPE-01",
        "code": "MKW-CPE-101",
        "title": "Correlated AI Agent Errors & Effective Fleet Capacity (N_eff) for CFOs & Risk Officers",
        "workload": "PT6H",
        "hoursLabel": "6 CPE / CPD Study Hours · Open-Access Preprints",
        "level": "Executive & Postgraduate Continuing Professional Education",
        "audience": "CFOs, Chief Risk Officers, Quantitative Risk Managers, Actuaries",
        "preprintsUsed": [
          "PP-01 (arXiv:2506.07962 — Correlated Errors in LLMs, 2025)",
          "PP-09 (Algorithmic Monoculture, Kleinberg & Raghavan / Stanford)"
        ],
        "preprintUrl": "https://markowitz.systems/preprint/pp-01",
        "coreEquation": "Var(w̄) = σ²[ρ + (1−ρ)/N],   N_eff = N / (1 + (N−1)ρ)",
        "teaches": "Quantifying pairwise error correlation ρ across single-model AI agent fleets; proving why N=50 agents at ρ=0.60 diversify like N_eff ≈ 1.64 independent agents.",
        "summary": "Self-paced continuing professional education module built around fresh 2025 empirical LLM error-correlation preprints and Harry M. Markowitz (1952)."
      },
      {
        "id": "CPE-02",
        "code": "MKW-CPE-102",
        "title": "Ledoit–Wolf Covariance Shrinkage (N > T) & Markowitz Semivariance in Agentic Finance",
        "workload": "PT8H",
        "hoursLabel": "8 CPE / CPD Study Hours · Open-Access Preprints",
        "level": "Quantitative & Actuarial Continuing Professional Education",
        "audience": "Financial Econometricians, Quants, Actuaries (FIA/FSA/CERA), Model Validators",
        "preprintsUsed": [
          "PP-01 (arXiv:2506.07962 — Correlated Errors in LLMs, ICML 2025)",
          "Cowles Foundation Monograph 16 Ch. 9 (Wiley, 1959)"
        ],
        "preprintUrl": "https://markowitz.systems/preprint/pp-01",
        "coreEquation": "Σ̂_LW = (1−δ*)S_emp + δ*νI_N,   Σ⁻_ij = E[min(v_i−τ,0)·min(v_j−τ,0)]",
        "teaches": "Regularizing rank-deficient sample covariance matrices when agent count N exceeds daily eval window T, and pricing asymmetric accounting restatement risk via downside semivariance.",
        "summary": "Advanced upskilling module for quantitative finance specialists estimating invertible error covariance Σ and semivariance Σ⁻ from paired ledger evaluations."
      },
      {
        "id": "CPE-03",
        "code": "MKW-CPE-103",
        "title": "Multi-Tier LLM Routing Cascades & Inference Cost Efficient Frontiers (−65% to −85% Spend)",
        "workload": "PT6H",
        "hoursLabel": "6 CPE / CPD Study Hours · Open-Access Preprints",
        "level": "Executive FinOps & FP&A Continuing Professional Education",
        "audience": "VP Finance, FP&A Directors, AI FinOps Leads, Corporate Controllers",
        "preprintsUsed": [
          "PP-02 (arXiv:2305.05176 — Stanford FrugalGPT)",
          "PP-03 (arXiv:2406.18665 — UC Berkeley RouteLLM, ICLR 2025)",
          "PP-04 (arXiv:2410.10347 — ETH Zurich LP Cascades, ICML 2025)"
        ],
        "preprintUrl": "https://markowitz.systems/preprint/pp-03",
        "coreEquation": "E[Cost_cascade] = c_small + (1−θ(τ))·c_frontier",
        "teaches": "Designing calibrated confidence-threshold routers that offload 70%–85% of routine AP, SQL, and variance-commentary tasks to low-cost tiers while preserving frontier precision.",
        "summary": "Practical upskilling syllabus translating fresh Stanford and UC Berkeley routing preprints into auditable monthly cloud & token budget reductions."
      },
      {
        "id": "CPE-04",
        "code": "MKW-CPE-104",
        "title": "Cost-Aware Agent Auditing, Retry-Loop Governance & Multi-Vendor Diversification",
        "workload": "PT6H",
        "hoursLabel": "6 CPE / CPD Study Hours · Open-Access Preprints",
        "level": "Continuing Professional Education in AI Audit & Procurement",
        "audience": "Internal Audit Directors, IT Audit / SOX 404 Leads, Procurement & Vendor Risk Teams",
        "preprintsUsed": [
          "PP-05 (arXiv:2407.01502 — Princeton AI Agents That Matter)",
          "ETH Zurich Optimal Model Allocation (arXiv:2410.10347)"
        ],
        "preprintUrl": "https://markowitz.systems/preprint/pp-05",
        "coreEquation": "HHI_vendor = ∑_m s_m² ≤ 0.40,   ρ_cross-vendor ≪ ρ_intra-vendor",
        "teaches": "Auditing two-dimensional (USD Cost per Resolved Task, Validated Accuracy) Pareto curves, capping runaway retry loops, and running paired evals across Anthropic, OpenAI, and Google.",
        "summary": "Continuing education module equipping finance and audit professionals to benchmark external SaaS agents and enforce Basel-style single-model exposure limits."
      },
      {
        "id": "CPE-05",
        "code": "MKW-CPE-105",
        "title": "Audit Materiality as Rockafellar–Uryasev CVaR_99% & Tobin Separation (Deterministic Code = r_f)",
        "workload": "PT7H",
        "hoursLabel": "7 CPE / CPD Study Hours · Open-Access Preprints",
        "level": "Executive Accounting, Treasury & Actuarial Upskilling",
        "audience": "Chief Accounting Officers, CPAs/Chartered Accountants, Corporate Treasurers, Risk Quants",
        "preprintsUsed": [
          "PP-10 (NBER w32487 — Acemoglu Macroeconomics of AI)",
          "Markowitz–Tobin–Sharpe Capital Market Line"
        ],
        "preprintUrl": "https://markowitz.systems/preprint/pp-10",
        "coreEquation": "CVaR_0.99(L(w)) ≤ M_audit,   SR_i = (E[Value_i − Cost_i] − μ_code) / σ_i",
        "teaches": "Mapping statutory SOX / PCAOB audit materiality M_audit to 99% Conditional Value-at-Risk and combining zero-variance deterministic ERP/SQL rules with diversified LLM verifiers.",
        "summary": "Rigorous continuing education bridge between statutory financial reporting controls, Tobin's separation theorem, and Agent Sharpe Ratios."
      },
      {
        "id": "CPE-06",
        "code": "MKW-CPE-106",
        "title": "Cybernetic Human + AI Finance Teams: Field RCT Evidence & Selective Deferral Architecture",
        "workload": "PT6H",
        "hoursLabel": "6 CPE / CPD Study Hours · Open-Access Preprints",
        "level": "Executive Workforce & Operating-Model Continuing Education",
        "audience": "CFOs, Finance Transformation Leaders, HR/Workforce Economists, Senior Controllers",
        "preprintsUsed": [
          "PP-06 (NBER w33641 — HBS/Wharton Cybernetic Teammate, 2025)",
          "PP-07 (arXiv:2503.18238 — MIT Sloan, 2025)",
          "PP-08 (Nature Human Behaviour Meta-Analysis)"
        ],
        "preprintUrl": "https://markowitz.systems/preprint/pp-06",
        "coreEquation": "Quality(1 Specialist + Orthogonal AI) ≈ Quality(2-Human Team),   ΔTime = −16.4%",
        "teaches": "Eliminating negative-synergy human rubber-stamping (Hedges' g = −0.23) and structuring high-skill Human + AI pods with calibrated selective deferral.",
        "summary": "Executive upskilling module synthesizing fresh 2024–2025 Harvard, Wharton, and MIT randomized controlled trials on Human + AI team design."
      }
    ],
    "self_assessment_flashcards": [
      {
        "q": "What is the effective number of independent agents N_eff when N equally weighted AI agents share equal variance σ² and pairwise error correlation ρ?",
        "a": "N_eff = N / (1 + (N − 1)ρ). At N = 50 agents and pairwise correlation ρ = 0.60 on one base model, N_eff ≈ 1.64 independent agents, and portfolio variance σ²[ρ + (1−ρ)/N] floors at σ²ρ rather than converging to zero."
      },
      {
        "q": "Why does a 50-agent AI fleet evaluated over T = 30 days require Ledoit–Wolf covariance shrinkage?",
        "a": "When fleet size N = 50 exceeds evaluation window T = 30 observations, the empirical sample covariance matrix S_emp has rank at most T − 1 = 29 < 50 and is mathematically singular (non-invertible). Ledoit–Wolf shrinkage Σ̂_LW = (1 − δ*)S_emp + δ*νI_N restores positive-definiteness and out-of-sample stability."
      },
      {
        "q": "How does Tobin's Separation Theorem apply to deterministic ERP/SQL code versus probabilistic LLM agents?",
        "a": "Deterministic validation scripts and SQL reconciliation rules have zero stochastic hallucination variance (σ_code = 0, idealized risk-free leg) and anchor the Capital Market Line alongside the diversified multi-vendor LLM tangency portfolio."
      },
      {
        "q": "How do CFOs translate statutory audit materiality into an AI agent risk constraint?",
        "a": "By bounding the 99% Conditional Value-at-Risk (Rockafellar–Uryasev CVaR_0.99) of aggregate portfolio error losses L(w) across financial reporting workflows to remain strictly below the statutory audit materiality threshold M_audit: CVaR_0.99(L(w)) ≤ M_audit."
      }
    ]
  },
  "active_research_pods": [
    {
      "id": "POD-01",
      "code": "MARKOWITZ-SIGMA-01",
      "title": "Cross-Agent & Multi-Vendor Error Covariance (Σ) Econometrician / Quant",
      "comp": "$25–$45 / hr USD",
      "commitment": "Part-Time Remote Pod (10–25 hrs/wk)",
      "theorem": "Σ = B F_model Bᵀ + Δ_agent,   Var(w̄) = σ²[ρ + (1−ρ)/N]",
      "summary": "Measure empirical error correlations ρ across LLM families, prompts, fine-tunes, and SaaS agent vendors; build Ledoit–Wolf shrinkage and factor covariance matrices Σ for CFO AI risk committees.",
      "stack": [
        "Covariance Estimation",
        "Shrinkage & Factor Models",
        "Cross-Model Benchmarking",
        "Python / R / SQL",
        "Financial Econometrics"
      ]
    },
    {
      "id": "POD-02",
      "code": "MARKOWITZ-CVAR-02",
      "title": "CVaR, Actuarial Tail-Risk & Single-Model Concentration Risk Manager",
      "comp": "$25–$45 / hr USD",
      "commitment": "Part-Time Remote Pod (10–25 hrs/wk)",
      "theorem": "min_w CVaR_α(L(w))   s.t.   wᵀμ ≥ μ*,  ‖Bᵀw‖_∞ ≤ κ_vendor",
      "summary": "Quantify correlated tail losses, semivariance, copula stress scenarios, and maximum single-model exposure caps across autonomous finance, procurement, pricing, and customer-facing workflows.",
      "stack": [
        "CVaR & Semivariance",
        "Actuarial Loss Modeling",
        "Vendor Concentration Caps",
        "Monte Carlo Stress Testing",
        "Enterprise Risk Management"
      ]
    },
    {
      "id": "POD-03",
      "code": "MARKOWITZ-CLOSE-03",
      "title": "Agentic Month-End Close, Treasury Liquidity & Audit-Trail Risk Specialist",
      "comp": "$20–$45 / hr USD",
      "commitment": "Part-Time Remote Pod (10–25 hrs/wk)",
      "theorem": "min_w wᵀΣ_close w + η · P( |ΔLedger| > τ_materiality )",
      "summary": "Design multi-model ensemble verification, orthogonal deterministic accounting checks, intraday cash-sweep liquidity buffers, and SOX/IFRS-compliant audit trails for agentic close and treasury.",
      "stack": [
        "Agentic Month-End Close",
        "Treasury & Cash Optimization",
        "SOX / Audit Trails",
        "Orthogonal Model Ensembles",
        "ERP / Ledger Reconciliation"
      ]
    },
    {
      "id": "POD-04",
      "code": "MARKOWITZ-FPNA-04",
      "title": "CFO Capital Budgeting, Real Options & AI Efficient-Frontier FP&A Analyst",
      "comp": "$20–$45 / hr USD",
      "commitment": "Part-Time Remote Pod (10–25 hrs/wk)",
      "theorem": "min_w wᵀΣw   s.t.   wᵀμ = μ*,   wᵀc ≤ Budget_CFO",
      "summary": "Construct the CFO's AI Efficient Frontier across token spend, SaaS licenses, GPU commitments, human-in-the-loop review gates, and cash — pricing staged AI rollouts via real options and ROIC hurdles.",
      "stack": [
        "Capital Budgeting & ROIC",
        "Real Options Valuation",
        "Strategic FP&A Modeling",
        "Mean–Variance Optimization",
        "AI Spend Governance"
      ]
    }
  ],
  "faq": [
    {
      "q": "Why does deploying 50 AI agents on one foundation model fail to diversify operational and financial risk?",
      "a": "Under equal weights (1/N), equal variance σ², and pairwise correlation ρ, Harry Markowitz's variance formula gives σ²[ρ + (1 − ρ)/N] → σ²ρ, so effective independent capacity is N_eff = N / (1 + (N − 1)ρ). Kim et al. (arXiv:2506.07962, ICML 2025) show LLMs agree up to 60% of the time when both are wrong — so at N = 50 and ρ = 0.60, fifty agents diversify like only N_eff ≈ 1.64 independent agents (collapsing to 1 as ρ → 1)."
    },
    {
      "q": "Why is the sample covariance matrix Σ singular when a CFO has 50 agents and 30 days of evals?",
      "a": "With N = 50 agents and T = 30 daily evaluation snapshots, the empirical covariance matrix S_emp has rank ≤ T − 1 = 29 < 50 (det S_emp = 0). Optimizing directly on S_emp invents spurious 'zero-variance' agent combinations. Quantitative specialists apply Ledoit–Wolf shrinkage Σ̂_LW = (1 − δ*)S_emp + δ*νI_N and base-model factor models (e_i = β_i F_base + ε_i)."
    },
    {
      "q": "How do deterministic scripts, multi-vendor evals (Anthropic, OpenAI, Google), and audit materiality work together?",
      "a": "Deterministic rules and SQL ledger checks act as an idealized risk-free leg (σ_0 = 0, Tobin separation), while Anthropic, OpenAI, and Google form a lower-ρ risky tangency portfolio verified via item-level paired evals. Statutory audit materiality M_audit serves as the firm's 99% CVaR risk budget (CVaR_99%(L(w)) ≤ M_audit), re-estimated and rebalanced on every model version release."
    },
    {
      "q": "How do CFO Offices engage Markowitz.systems to build an Efficient Frontier for AI spend?",
      "a": "CFOs, VPs of Finance, Corporate Treasurers, and Audit/Risk Committees contact frontier@markowitz.systems directly. Our specialists run paired multi-vendor evals, estimate shrunk covariance Σ̂_LW and downside semivariance Σ⁻, set banking Large-Exposure caps per foundation model, and construct an executable Capital Market Line and Mean–Variance Efficient Frontier."
    },
    {
      "q": "How do remote part-time R&D pods ($20–$45/hr, 10–25 hrs/wk) work for Financial Economists and Quants?",
      "a": "Quants, actuaries, risk managers, financial economists, and senior FP&A analysts join continuous remote R&D pods ($20–$45/hr USD, 10–25 hrs/wk) — applying mean-variance optimization, Ledoit–Wolf covariance estimation, semivariance, CVaR, Agent Sharpe Ratios, and real options to agentic month-end close, treasury, FP&A, and audit trails."
    },
    {
      "q": "Is Markowitz.systems affiliated with the Harry M. Markowitz estate, and how is candidate data governed?",
      "a": "Markowitz.systems is an independent research and quantitative talent platform and is not affiliated with, endorsed by, or connected to the Estate of Harry M. Markowitz, the Cowles Foundation at Yale University, or John Wiley & Sons. Credential intake is governed by IQ Vector Research Operations (privacy@markowitz.systems) under our published Privacy Policy (/privacy), excluding RU, IR, and KP."
    }
  ]
}