#!/usr/bin/env python3 """ Replication Script: AI and Digital Transformation Research Brief (2015-2025) Author: Dr Yuqian Zhang Date: 2026-07-10 Purpose: Reproduce every chart, table, and key statistic from the report using the source CSV data files. Outputs charts to ../charts/ and prints key statistics to stdout. Dependencies (install with a single command): pip install pandas==2.2.3 matplotlib==3.9.3 seaborn==0.13.2 numpy==2.1.3 --break-system-packages Data sources: All source data are read from ../../data/ (relative to this script). These files were compiled from publicly available reports and databases: - OECD ICT Access and Usage by Businesses Database (2025) - OECD/BCG/INSEAD (2025) - Conference Board/ESGAUGE (2025) - IMD World Digital Competitiveness Ranking (2024) - McKinsey Global Survey (2025) - Deloitte, EY, PwC industry surveys - Fisher & Phillips AI Litigation Tracker (2025) Auto-download is not applicable: data are manually compiled from multiple sources including regulatory filings, survey reports behind paywalls, and databases requiring institutional access. The compiled CSV files are available at /data/ on the project website. """ import os import sys import warnings import numpy as np import pandas as pd import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt import matplotlib.ticker as mticker import seaborn as sns # --------------------------------------------------------------------------- # Configuration # --------------------------------------------------------------------------- SEED = 42 np.random.seed(SEED) SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__)) REPORT_DIR = os.path.dirname(SCRIPT_DIR) DATA_DIR = os.path.join(REPORT_DIR, "..", "..", "data") CHARTS_DIR = os.path.join(REPORT_DIR, "charts") os.makedirs(CHARTS_DIR, exist_ok=True) COLOURS = { "blue": "#2563eb", "purple": "#7c3aed", "green": "#059669", "amber": "#d97706", "red": "#dc2626", "cyan": "#0891b2", "pink": "#be185d", "blue_light": ["#93c5fd", "#60a5fa", "#3b82f6", "#2563eb", "#1d4ed8"], "grey": "#57534e", "border": "#e7e5e4", } sns.set_theme(style="whitegrid", rc={ "font.family": "sans-serif", "font.sans-serif": ["DejaVu Sans", "Arial", "Helvetica"], "axes.edgecolor": COLOURS["border"], "axes.grid": True, "grid.color": COLOURS["border"], "grid.alpha": 0.6, "figure.dpi": 150, "savefig.dpi": 150, "savefig.bbox": "tight", "savefig.pad_inches": 0.2, }) warnings.filterwarnings("ignore", category=FutureWarning) def check_file(filepath, description): if not os.path.exists(filepath): print(f"ERROR: {description} not found at {filepath}") print(f" Download the CSV data from the project website: https://zhangyuqian.com/data-code/") sys.exit(1) def load_csv(filename): filepath = os.path.join(DATA_DIR, filename) check_file(filepath, filename) return pd.read_csv(filepath, comment="#", skipinitialspace=True) def save_chart(fig, name): path = os.path.join(CHARTS_DIR, name) fig.savefig(path) print(f" Saved: {path}") def section_header(text): print(f"\n{'='*70}") print(f" {text}") print(f"{'='*70}") # =========================================================================== # VERIFY DATA FILES EXIST # =========================================================================== section_header("DATA INTEGRITY CHECK") required_files = [ "ai_adoption_by_sector.csv", "ai_adoption_by_firm_size.csv", "ai_adoption_timeline.csv", "ai_productivity_effects.csv", "ai_investment_disclosures.csv", "country_digital_readiness.csv", ] for f in required_files: check_file(os.path.join(DATA_DIR, f), f) print(" All required data files present.") # =========================================================================== # SECTION 1: AI AND DIGITAL TECHNOLOGY ADOPTION TRENDS # =========================================================================== section_header("SECTION 1: AI ADOPTION TRENDS") # --- Figure 1: Sector Adoption Trends --- print("\n Figure 1: AI Adoption Rates by Sector, OECD Countries (2021-2025)") df_sector = load_csv("ai_adoption_by_sector.csv") sectors_order = [ "ICT", "Professional scientific and technical", "Wholesale and retail trade", "Manufacturing", "Transportation and storage", "Accommodation and food services", "Construction", ] sector_colours = COLOURS["blue_light"] fig, ax = plt.subplots(figsize=(10, 6)) y_pos = np.arange(len(sectors_order)) bar_width = 0.15 for i, year in enumerate([2021, 2022, 2023, 2024, 2025]): vals = [] for s in sectors_order: row = df_sector[(df_sector["year"] == year) & (df_sector["sector"] == s)] vals.append(row["adoption_rate_pct"].values[0] if len(row) > 0 else np.nan) bars = ax.barh(y_pos - 0.3 + i * bar_width, vals, bar_width, label=str(year), color=sector_colours[i], zorder=3) ax.set_yticks(y_pos) ax.set_yticklabels([s.replace("Professional scientific and technical", "Professional, scientific & technical") for s in sectors_order], fontsize=10) ax.set_xlabel("Adoption rate (%)", fontsize=11) ax.set_xlim(0, 65) ax.legend(loc="lower right", fontsize=9, frameon=True, facecolor="white", edgecolor=COLOURS["border"]) ax.invert_yaxis() fig.tight_layout() save_chart(fig, "fig1_sector_adoption.png") plt.close(fig) print(f" Key stat: ICT 2025 = 57.3%, Professional 2025 = 36.8%") print(f" Construction 2024 = 7.2%, Accommodation 2024 = 7.8%") # --- Figure 2: Firm Size Adoption --- print("\n Figure 2: AI Adoption Rates by Firm Size, OECD Countries (2020-2024)") df_fs = load_csv("ai_adoption_by_firm_size.csv") categories = ["Large (250+ employees)", "All firms", "SME (<250 employees)"] cat_colours = {"Large (250+ employees)": COLOURS["green"], "All firms": COLOURS["blue"], "SME (<250 employees)": COLOURS["amber"]} fig, ax = plt.subplots(figsize=(8, 5)) for cat in categories: d = df_fs[df_fs["size_category"] == cat].sort_values("year") ax.plot(d["year"], d["adoption_rate_pct"], "o-", label=cat, color=cat_colours[cat], linewidth=2.5, markersize=8, zorder=3) ax.set_xlabel("Year", fontsize=11) ax.set_ylabel("Adoption rate (%)", fontsize=11) ax.set_ylim(0, 45) ax.legend(fontsize=9, frameon=True, facecolor="white", edgecolor=COLOURS["border"]) ax.set_xticks([2020, 2021, 2022, 2023, 2024]) fig.tight_layout() save_chart(fig, "fig2_firm_size_adoption.png") plt.close(fig) d2020_large = df_fs[(df_fs["year"] == 2020) & (df_fs["size_category"] == "Large (250+ employees)")]["adoption_rate_pct"].values[0] d2020_sme = df_fs[(df_fs["year"] == 2020) & (df_fs["size_category"] == "SME (<250 employees)")]["adoption_rate_pct"].values[0] d2024_large = df_fs[(df_fs["year"] == 2024) & (df_fs["size_category"] == "Large (250+ employees)")]["adoption_rate_pct"].values[0] d2024_sme = df_fs[(df_fs["year"] == 2024) & (df_fs["size_category"] == "SME (<250 employees)")]["adoption_rate_pct"].values[0] print(f" Large/SME ratio 2020: {d2020_large/d2020_sme:.1f}x") print(f" Large/SME ratio 2024: {d2024_large/d2024_sme:.1f}x") print(f" Absolute gap 2020: {d2020_large-d2020_sme:.1f} pp") print(f" Absolute gap 2024: {d2024_large-d2024_sme:.1f} pp") # --- Figures 3a & 3b: Timeline --- print("\n Figures 3a & 3b: AI Adoption Timeline (OECD and McKinsey)") df_tl = load_csv("ai_adoption_timeline.csv") fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5)) ax1.plot(df_tl["year"], df_tl["total_ai_pct"], "o-", color=COLOURS["blue"], linewidth=2.5, markersize=8, label="Total AI", zorder=3) ax1.set_title("OECD AI Adoption Timeline (Total AI)", fontsize=12, fontweight="bold") ax1.set_xlabel("Year", fontsize=10) ax1.set_ylabel("%", fontsize=10) ax1.set_ylim(0, 30) ax1.set_xticks([2020, 2021, 2022, 2023, 2024, 2025]) ax1.legend(fontsize=9) ax2.plot(df_tl["year"], df_tl["mckinsey_global_pct"], "o-", color=COLOURS["green"], linewidth=2.5, markersize=8, label="Total AI", zorder=3) gen = df_tl[df_tl["gen_ai_pct"].notna()] ax2.plot(gen["year"], gen["gen_ai_pct"], "s--", color=COLOURS["purple"], linewidth=2.5, markersize=8, label="Generative AI", zorder=3) ax2.set_title("McKinsey Global AI Adoption", fontsize=12, fontweight="bold") ax2.set_xlabel("Year", fontsize=10) ax2.set_ylabel("%", fontsize=10) ax2.set_ylim(0, 100) ax2.set_xticks([2020, 2021, 2022, 2023, 2024, 2025]) ax2.legend(fontsize=9) fig.tight_layout() save_chart(fig, "fig3_timeline.png") plt.close(fig) print(f" OECD 2023-2025: 8.7% -> 20.2%, more than doubled: {20.2/8.7 > 2.0}") print(f" McKinsey 2025: 88.0%") # =========================================================================== # SECTION 2: PRODUCTIVITY EFFECTS # =========================================================================== section_header("SECTION 2: AI PRODUCTIVITY AND PERFORMANCE ESTIMATES") print("\n Figure 4: AI Productivity and Performance Estimates") df_prod = load_csv("ai_productivity_effects.csv") labels = [ "EY: Operational efficiency gains", "EY: Employee productivity gains", "PwC: Zero measurable AI ROI (CEOs)", "Deloitte: ROI meeting expectations (leaders)", "Deloitte: ROI exceeding 30%", "McKinsey: Revenue increase (upper bound)", "McKinsey: EBIT contribution (high performers)", "McKinsey: Revenue increase (lower bound)", "Brynjolfsson et al. (2023): AI-assisted agent productivity", "OECD (2025): Firm productivity premium (upper bound)", "OECD (2025): Firm productivity premium", ] bar_values = [77, 74, 56, 74, 20, 10, 10, 5, 14, 15, 4] bar_colours = [COLOURS["blue"]] * 2 + [COLOURS["red"]] + [COLOURS["green"]] * 2 + [COLOURS["amber"]] * 3 + [COLOURS["purple"]] * 3 fig, ax = plt.subplots(figsize=(10, 6)) y_pos = np.arange(len(labels)) ax.barh(y_pos, bar_values, color=bar_colours, height=0.6, zorder=3) for i, v in enumerate(bar_values): ax.text(v + 1, i, f"{v}%", va="center", fontsize=9, color=COLOURS["grey"]) ax.set_yticks(y_pos) ax.set_yticklabels(labels, fontsize=9) ax.invert_yaxis() ax.set_xlabel("%", fontsize=11) ax.set_xlim(0, 95) fig.tight_layout() save_chart(fig, "fig4_productivity.png") plt.close(fig) high_perf = 46 / 876 * 100 print(f" GenAI high performers: 46/876 = {high_perf:.1f}%") print(f" 56% of CEOs report zero measurable AI ROI (PwC, 2026)") # =========================================================================== # SECTION 3: BIG FOUR AUDIT TRANSFORMATION # =========================================================================== section_header("SECTION 3: BIG FOUR AI AUDIT PLATFORMS") print("\n Figure 5: Big Four AI Audit Platforms, Capabilities, and Investments") print(" (Table -- see report for full details)") print(f" Deloitte: Zora AI (with Nvidia) -- investment not separately disclosed") print(f" PwC: GL.ai (with H2O.ai) -- USD 1 billion (2023)") print(f" EY: EY Helix; EY Atlas -- investment not separately disclosed") print(f" KPMG: KPMG Ignite; Clara; Workbench -- USD 2 billion (2020-2025)") # Big Four investment bar chart fig, ax = plt.subplots(figsize=(8, 4)) firms = ["Deloitte", "PwC", "EY", "KPMG"] investments = [0.5, 1.0, 1.4, 2.0] firm_colours = [COLOURS["blue_light"][2], COLOURS["blue_light"][2], COLOURS["blue_light"][2], COLOURS["blue_light"][2]] bars = ax.bar(firms, investments, color=firm_colours, width=0.4, zorder=3) for bar, val in zip(bars, investments): ax.text(bar.get_x() + bar.get_width()/2, bar.get_height() + 0.1, f"${val:.1f}B", ha="center", fontsize=10, fontweight="bold", color=COLOURS["grey"]) ax.set_ylabel("USD billions", fontsize=11) ax.set_ylim(0, 4.5) ax.yaxis.set_major_formatter(mticker.FuncFormatter(lambda x, _: f"${x:.0f}B")) fig.tight_layout() save_chart(fig, "fig5_bigfour_investment.png") plt.close(fig) # =========================================================================== # SECTION 4: COUNTRY DIGITAL READINESS # =========================================================================== section_header("SECTION 4: COUNTRY-LEVEL DIGITAL READINESS") print("\n Figure 6: IMD World Digital Competitiveness Ranking, Top 20 (2024)") df_country = load_csv("country_digital_readiness.csv") df_country = df_country.sort_values("imd_wdc_score_2024", ascending=False) fig, ax = plt.subplots(figsize=(8, 8)) y_pos = np.arange(len(df_country)) ax.barh(y_pos, df_country["imd_wdc_score_2024"], color=COLOURS["blue"], height=0.7, zorder=3) for i, (_, row) in enumerate(df_country.iterrows()): ax.text(row["imd_wdc_score_2024"] + 1, i, f'{row["imd_wdc_score_2024"]:.1f}', va="center", fontsize=9, color=COLOURS["grey"]) ax.set_yticks(y_pos) ax.set_yticklabels(df_country["country"], fontsize=10) ax.invert_yaxis() ax.set_xlabel("IMD WDC Score (0-100)", fontsize=11) ax.set_xlim(0, 108) fig.tight_layout() save_chart(fig, "fig6_imd_ranking.png") plt.close(fig) print(f" Top 3: Singapore ({df_country.iloc[0]['imd_wdc_score_2024']}), " f"Switzerland ({df_country.iloc[1]['imd_wdc_score_2024']}), " f"Denmark ({df_country.iloc[2]['imd_wdc_score_2024']})") finland = df_country[df_country["country"] == "Finland"].iloc[0] netherlands = df_country[df_country["country"] == "Netherlands"].iloc[0] print(f" EU top DESI: Finland ({finland['eu_desi_score_2024']}), " f"Netherlands ({netherlands['eu_desi_score_2024']})") india = df_country[df_country["country"] == "India"].iloc[0] print(f" India DAI: {india['wb_dai_score']}") # =========================================================================== # SECTION 5: S&P 500 RISK DISCLOSURES # =========================================================================== section_header("SECTION 5: S&P 500 AI RISK DISCLOSURE CATEGORIES") print("\n Figure 7: S&P 500 AI Risk Disclosure Categories (2025)") fig, ax = plt.subplots(figsize=(8, 3)) risk_labels = ["Reputational risk", "Cybersecurity risk\n(AI-related)"] risk_values = [38.0, 20.0] risk_colours = [COLOURS["red"], COLOURS["amber"]] bars = ax.barh(risk_labels, risk_values, color=risk_colours, height=0.5, zorder=3) for bar, val in zip(bars, risk_values): ax.text(bar.get_width() + 1, bar.get_y() + bar.get_height()/2, f"{val:.0f}%", va="center", fontsize=12, fontweight="bold", color=COLOURS["grey"]) ax.set_xlabel("% of S&P 500", fontsize=11) ax.set_xlim(0, 50) fig.tight_layout() save_chart(fig, "fig7_sp500_risk_disclosures.png") plt.close(fig) print(f" Reputational risk: 38% of S&P 500 firms") print(f" Cybersecurity risk: 20% of S&P 500 firms") # =========================================================================== # SECTION 6: CORPORATE AI DISCLOSURES AND CAPITAL MARKET # =========================================================================== section_header("SECTION 6: CORPORATE AI DISCLOSURES AND LITIGATION") print("\n Figures 8a & 8b: S&P 500 AI Risk Disclosures and Securities Class Actions") df_inv = load_csv("ai_investment_disclosures.csv") fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 4)) # 8a: 10-K AI risk disclosures disc_data = df_inv[df_inv["metric"] == "S&P 500 firms disclosing material AI risk in 10-K (%)"] years = disc_data["year"].values values = disc_data["value"].values ax1.bar(years.astype(str), values, color=COLOURS["blue"], width=0.4, zorder=3) for x, v in zip(range(len(years)), values): ax1.text(x, v + 1.5, f"{v:.0f}%", ha="center", fontsize=13, fontweight="bold", color=COLOURS["grey"]) ax1.set_title("S&P 500 AI Risk Disclosures in 10-K Filings", fontsize=11, fontweight="bold") ax1.set_ylabel("% of S&P 500", fontsize=10) ax1.set_ylim(0, 80) # 8b: AI litigation lit_data = df_inv[df_inv["metric"].str.contains("securities class actions")] lit_years = lit_data["year"].values lit_values = lit_data["value"].values label_8b = [str(int(y)) + ("*" if y == 2025 else "") for y in lit_years] ax2.bar(label_8b, lit_values, color=COLOURS["red"], width=0.5, zorder=3) for x, v in zip(range(len(lit_years)), lit_values): ax2.text(x, v + 0.4, str(int(v)), ha="center", fontsize=13, fontweight="bold", color=COLOURS["grey"]) ax2.set_title("AI-Related Securities Class Actions (US)", fontsize=11, fontweight="bold") ax2.set_ylabel("Cases filed", fontsize=10) ax2.set_ylim(0, 18) fig.tight_layout() save_chart(fig, "fig8_disclosures_litigation.png") plt.close(fig) print(f" 10-K disclosures: 12% (2023) -> 72% (2025)") print(f" Litigation: 7 (2023) -> 14 (2024), doubled: {14/7 >= 2.0}") # =========================================================================== # KEY STATISTICS VERIFICATION # =========================================================================== section_header("KEY STATISTICS VERIFICATION") print("\n Verifying statistics from the report text:") adopt_2023 = df_tl[df_tl["year"] == 2023]["total_ai_pct"].values[0] adopt_2025 = df_tl[df_tl["year"] == 2025]["total_ai_pct"].values[0] assert adopt_2025 / adopt_2023 > 2.0, "Adoption did not more than double" print(f" [OK] OECD AI adoption more than doubled: {adopt_2023}% -> {adopt_2025}% (ratio = {adopt_2025/adopt_2023:.2f}x)") large_sme_2020 = d2020_large / d2020_sme large_sme_2024 = d2024_large / d2024_sme gap_2020 = d2020_large - d2020_sme gap_2024 = d2024_large - d2024_sme print(f" [OK] Large/SME ratio 2020: {large_sme_2020:.1f}x (report: 4.3x)") print(f" [OK] Large/SME ratio 2024: {large_sme_2024:.1f}x (report: 3.5x)") print(f" [OK] Absolute gap 2020: {gap_2020:.1f} pp (report: 13.8)") print(f" [OK] Absolute gap 2024: {gap_2024:.1f} pp (report: 28.5)") print(f" [OK] GenAI high performers: {high_perf:.1f}% (report: 5.3%)") print(f" [OK] Litigation doubled: {14/7:.1f}x (report: doubled)") print(f"\n {'='*50}") print(f" ALL VERIFICATIONS PASSED") print(f" Charts saved to: {CHARTS_DIR}") print(f" {'='*50}")