AITG Framework

For: Private Equity investors (due diligence, IC memo preparation, portfolio monitoring) and Corporate management teams (self-assessment, transformation roadmap, board reporting)

The AI Transformation Gap Index (AITG): An Empirical Framework for Measuring AI Transformation Opportunity, Disruption Risk, and Value Creation at the Industry and Firm Level

Barr, D. (2026). Working Paper. Contact: dean@dsconsult.ai

---

Overview

This repository provides full reproducibility for the AITG framework paper. It contains the scoring models, Monte Carlo sensitivity analysis, stress tests, and an Excel companion workbook builder.

File	Description
aitg_monte_carlo.py	Monte Carlo sensitivity analysis (Layers 1-4) + custom firm scoring
aitg_stress_test.py	Deterministic stress tests (13 assertions)
build_aitg_excel_revised_fixed.py	Excel companion workbook builder

The paper and LaTeX source are available on arXiv (cs.AI).

---

Quick Start

Requirements

• Python 3.10+
• openpyxl (only needed for Excel workbook builder)

pip install openpyxl

No other dependencies. The Monte Carlo and stress test scripts use only the Python standard library.

Run Tests

# Monte Carlo unit tests (10 tests)
python aitg_monte_carlo.py --test

# Stress tests (13 assertions)
python aitg_stress_test.py

# Backtest: Spearman rank correlation (rho_s = 0.818)
python aitg_monte_carlo.py --backtest

Run Full 14-Company Report

python aitg_monte_carlo.py --all-firms

Outputs base-case VCB pipeline, Layer 4 Monte Carlo P10/P50/P90, and dimension breakdowns for all 14 firms in the paper.

---

Scoring a Custom Firm

There are two ways to score a firm not in the paper's 14-company dataset.

Option A: From Pre-Computed Dimension Scores (--score)

If you already have the 6 AITG dimension scores (0-10 scale) and 5 IFS factors (0-1 scale):

python aitg_monte_carlo.py --score \
    --name "Target Corp" \
    --dims "5.0,4.5,3.5,4.0,3.0,4.5" \
    --ifs "0.70,0.65,0.80,0.75,0.85" \
    --revenue 3.4 \
    --s-star 3.3 \
    --iass-star 9.38 \
    --exit-mult 10.0

Parameters:

Flag	Description
--dims	6 comma-separated scores: DIM, PAC, WAR, DAR, APR, OAC (0-10)
--ifs	5 comma-separated factors: OCC, DR, VTR, CRS, REG (0-1)
--revenue	Firm revenue in $B
--s-star	Industry critical-scale threshold in $B
--iass-star	Industry IASS* frontier score (0-10)
--exit-mult	Exit multiple (default: 10.0)
--name	Firm name for report header

Option B: From the 25-Question Management Survey (--survey)

If you are conducting PE due diligence or corporate self-assessment using the 25-question survey instrument (documented in ESM Appendix), enter the 29 raw responses (24 dimension questions + 5 IFS sub-factor ratings, each scored 0-4):

python aitg_monte_carlo.py --survey \
    "2,2,1,2,2,1,1,2,2,1,1,2,2,1,1,2,1,1,1,1,2,2,1,2,2,2,3,2,3" \
    --revenue 3.4 \
    --s-star 3.3 \
    --iass-star 9.38 \
    --name "Target Corp"

Survey response mapping:

Response	Score	Meaning
0	1	Minimal / no capability
1	3	Early stage
2	5	Moderate / cross-functional
3	7	Advanced (evidence required)
4	9	Industry-leading (evidence required)

29 values in order:

• Values 1-4: DIM (Data Infrastructure Maturity) — Q1-Q4
• Values 5-8: PAC (Process Automation Coverage) — Q5-Q8
• Values 9-12: WAR (Workforce AI Augmentation Rate) — Q9-Q12
• Values 13-16: DAR (Decision Automation Rate) — Q13-Q16
• Values 17-20: APR (AI Product/Revenue Integration) — Q17-Q20
• Values 21-24: OAC (Organizational AI Capability) — Q21-Q24
• Values 25-29: IFS sub-factors — OCC, DR, VTR, CRS, REG (Q25a-e)

The full 25-question survey text with response anchors is in the ESM (Section: "25-Question Management Survey Instrument").

Output

Both modes produce:

• 6 dimension scores and AITG_raw
• G_eff (effective gap), IR (implementation readiness)
• t_hat, t_base, t50 timing parameters (months)
• Base-case delta-EV ($B)
• Monte Carlo P10/P50/P90 (10,000 draws)
• Value Density and investment tier (Tier 1/2/3)

---

Excel Companion Workbook

The Excel builder creates a workbook with a survey input sheet that automates the same pipeline:

python build_aitg_excel_revised_fixed.py \
    --template "AITG_Excel_Companion_v1.xlsx" \
    --out "AITG_Excel_Companion_revised.xlsx"

Worksheets

Sheet	Function
SURVEY INPUT	Enter 25 survey responses (0-4); auto-computes dimension scores and IFS factors
COMPANY SCORER	6-dimension AITG rubric, AITG_raw, G_eff, wave zone, t_hat, Phi_f
IASS CALC	6-dimension geometric mean with paper Table 2 weights
t_hat LOOKUP	Pre-solved t_hat for AITG in [0.1, 9.9] at 0.1 steps
IFS CALC	5-factor IFS with endogenous t50,f timing
VCB MODEL	7 value pools, delta-EV, Value Density, investment tier

Workflow: Fill out SURVEY INPUT (column C, responses 0-4) -> scores auto-flow to COMPANY SCORER and IFS CALC -> VCB MODEL computes delta-EV and tier.

The Excel model is a deterministic base-case approximation. For Monte Carlo distributions (P10/P50/P90), use the Python script.

---

Monte Carlo Layers

python aitg_monte_carlo.py              # Full report (all 4 layers)
python aitg_monte_carlo.py --layer 2    # Single layer only
python aitg_monte_carlo.py --seed 42    # Custom RNG seed

Layer	Description	Key Output
1	IASS Weight Sensitivity	Rank stability R_s under +/-5% weight perturbation
2	Sobol First-Order Decomposition	Variance attribution (exit_mult 50%, capture 44%, IFS 5%)
3	Parameter Range Sweeps	Deterministic one-at-a-time sensitivity
4	Full VCB Monte Carlo	P10/P50/P90 of delta-EV (10,000 draws)

---

Key Results (from paper)

Metric	Value
Spearman backtest rho_s	0.818 (p < 0.004)
Sobol: exit multiple share	50%
Sobol: capture rate share	44%
Layer 1 rank stability R_s	0.19
Pearson r(G_eff, VD_mid)	0.22 (Wide-Gap Fallacy)

---

Citation

@techreport{barr2026aitg,
  author    = {Barr, Dean},
  title     = {The {AI} Transformation Gap Index ({AITG}): An Empirical Framework
               for Measuring {AI} Transformation Opportunity, Disruption Risk,
               and Value Creation at the Industry and Firm Level},
  year      = {2026},
  month     = {February},
  type      = {Working Paper},
  institution = {DS Consulting},
  note      = {Contact: dean@dsconsult.ai}
}

---

License

Code: MIT License. Paper and ESM: All rights reserved.