BCG’s 10-20-70 Rule in Practice: What a 500-Person Company’s AI Budget Actually Looks Like
Executive Summary
- BCG’s 10-20-70 framework says 10% of AI investment goes to algorithms, 20% to technology, and 70% to people and processes. The ratio is directionally correct — every major consulting firm and independent study confirms that people investment drives AI ROI — but most companies invert it, spending 60-70% on technology and scrambling to fund organizational change afterward.
- For a 500-person company spending $500K on AI in Year 1, the 70% “people” line item means $350,000 — and most CFOs cannot immediately see where that money goes. It goes to workflow redesign, role-specific training, change management staffing, measurement infrastructure, and the productivity dip that occurs during the 3-6 month adoption period. This document breaks down each line item with dollar ranges.
- The companies that follow the 10-20-70 ratio outperform those that do not by 2-3x on ROI. BCG’s own survey (n=1,803 C-level executives, January 2025) finds only 25% of companies report meaningful AI value. The gap maps directly to budget inversion: organizations spending heavily on tools without proportional people investment.
- Global AI spending rises 44% in 2026 while training budgets grow 5%. This is the inversion in real time. Fortune reports $500 billion in global AI spending against near-flat training investment — the precise mistake the 10-20-70 framework warns against.
- The framework’s real insight is not a budget formula — it is a diagnostic. If your AI budget looks like 60/30/10 (technology/people/algorithms), you are likely in the 75% that never see meaningful returns. Rebalancing is not about spending more. It is about spending differently.
The Framework: What BCG Actually Says
BCG introduced the 10-20-70 rule in 2020 and has reinforced it in every subsequent AI publication through 2026. The allocation:
10% — Algorithms and models. Selecting, configuring, and fine-tuning the AI models themselves. In 2020, this meant building custom ML models. In 2026, it means choosing between pre-built foundation models (GPT-4o, Claude, Gemini) and configuring them for your use case. Open-source models and cloud APIs have commoditized this layer. The 10% acknowledges that the technology itself is not the differentiator.
20% — Technology and data infrastructure. Data cleaning, integration, API configuration, security tooling, cloud infrastructure, and vendor platform setup. This is the plumbing that makes AI work inside your existing systems.
70% — People and processes. Workflow redesign, training, change management, role redefinition, governance, measurement systems, and organizational adaptation. This is where BCG says the value actually gets captured — or lost.
The framework is supported by converging evidence from multiple independent sources:
- MIT Sloan research confirms 70% of AI’s value depends on complementary investments in people and processes, not technology sophistication.
- McKinsey’s 2025 State of AI survey (n=~1,933 organizations) finds the single strongest predictor of EBIT impact from AI is workflow redesign — a people/process investment that only 21% of companies make.
- Bain research shows companies taking a “human-centric approach to workforce productivity” deliver 2.3x total shareholder returns vs. technology-first organizations (S&P Capital IQ analysis, 250 companies per quadrant, 2020-2024).
- EY’s 2025 Work Reimagined Survey (n=16,500) finds companies with strong talent foundations capture 40% more productivity from AI than technology-first investors. Only 28% of organizations get this right.
What Happens When Companies Invert the Ratio
Most mid-market companies allocate budgets in exactly the wrong proportions. The typical pattern:
| Budget Category | What BCG Recommends | What Most Companies Do |
|---|---|---|
| Algorithms/Models | 10% | 5-10% |
| Technology/Infrastructure | 20% | 55-65% |
| People/Processes | 70% | 25-35% |
The consequences are measurable:
- 72% of CIOs report their organizations are breaking even or losing money on AI investments (Gartner, n=506 CIOs and technology leaders, May 2025). The primary cause: deploying tools without redesigning workflows or investing in adoption.
- 74% of companies are stuck in pilot mode, testing tools in controlled settings but never delivering measurable ROI at scale (BCG AI Radar, January 2025, n=1,803 C-level executives).
- 60% of companies fail to define financial KPIs for AI — they spend on technology but have no measurement system to track whether it works (BCG, 2025).
- Productivity drops 15-25% during the 3-6 month adoption period at companies that deploy tools without adequate training and change management (SmartDev, 2025 analysis of SME AI implementations).
The Fortune article published today (March 17, 2026) crystallizes the inversion: global AI spending rises 44% in 2026 to $500 billion while training budgets grow just 5% and average learning time per employee actually falls from 47 to 40 hours annually. Microsoft’s Work Trend Index finds 73% of knowledge workers use AI at work, yet 60% have received no formal training.
The 70% Translated: Line-by-Line Budget for a 500-Person Company
The CFO’s question is always the same: “What does 70% on people actually mean in dollars?”
Below is a realistic Year 1 budget for a 500-person company allocating $500,000 to an AI initiative following the 10-20-70 framework. The total may vary based on scope, but the proportions hold.
Algorithms and Models — $50,000 (10%)
| Line Item | Cost Range | Notes |
|---|---|---|
| Model selection and evaluation | $5,000-$10,000 | Staff time testing 2-3 models against your use cases |
| API/license fees for models | $25,000-$35,000 | Foundation model access (GPT-4o, Claude, Gemini) |
| Fine-tuning or prompt engineering | $10,000-$15,000 | Customization for domain-specific tasks |
This layer is small because the models are commodities. A 500-person company is not training its own LLM. It is selecting from commercially available options and configuring them.
Technology and Data Infrastructure — $100,000 (20%)
| Line Item | Cost Range | Notes |
|---|---|---|
| Software licensing (user seats) | $36,000-$60,000 | $30-$50/seat/month for 100 initial users |
| Integration and configuration | $20,000-$40,000 | Connecting AI to existing systems (ERP, CRM, HRIS) |
| Data cleaning and preparation | $10,000-$20,000 | Often 2-3x more effort than planned (SmartDev, 2025) |
| Security and compliance tooling | $10,000-$15,000 | DLP, access controls, audit logging |
| Cloud infrastructure increment | $5,000-$10,000 | Additional compute/storage for AI workloads |
Note: If your company already uses Microsoft 365, Google Workspace, or Salesforce, platform-native AI add-ons (Copilot, Gemini, Agentforce) reduce integration costs by 50-70% vs. standalone tools (AI Smart Ventures, 2026).
People and Processes — $350,000 (70%)
This is the section most budgets either omit or radically undersize. It breaks into five categories.
Category 1: Workflow Redesign — $80,000-$120,000
| Line Item | Cost Range | Notes |
|---|---|---|
| Process mapping and audit | $15,000-$25,000 | Map 3-5 workflows as they actually operate, not as documented |
| External consulting (if needed) | $30,000-$60,000 | 4-6 week engagement for workflow redesign methodology |
| Internal staff time (dedicated) | $25,000-$40,000 | 2-3 process owners at 20-30% for 8-12 weeks |
| Pilot workflow documentation | $5,000-$10,000 | Standard operating procedures for redesigned workflows |
This is where McKinsey’s data is decisive: the 5.5% of organizations achieving 5%+ EBIT impact from AI are 3x more likely to have redesigned workflows. Every other dollar in this budget underperforms without this step.
Category 2: Training and Upskilling — $75,000-$100,000
| Line Item | Cost Range | Notes |
|---|---|---|
| Role-specific AI training (all 500) | $50,000-$75,000 | $100-$150/person for 4-8 hours of structured training |
| Advanced training for power users (50) | $10,000-$15,000 | $200-$300/person for deeper skill development |
| AI champion development (5-10) | $10,000-$15,000 | $1,000-$2,000/person for certification and coaching |
| Ongoing quarterly refreshers | $5,000-$10,000 | 2-4 hours/person/quarter to address skill decay |
The training multiplier is well-documented. BCG’s “AI at Work” research finds regular AI usage is “sharply higher for employees that receive at least five hours of training.” Deloitte’s TrustID data shows hands-on training increases AI trust 144%. Colgate-Palmolive trained 14,000 employees before granting AI Hub access — the result was 3,000+ employee-built AI assistants, not 14,000 unused licenses.
Category 3: Change Management — $60,000-$80,000
| Line Item | Cost Range | Notes |
|---|---|---|
| Change management lead (fractional/internal) | $25,000-$40,000 | 25-50% of a senior leader’s time for 6-12 months |
| Communication program | $10,000-$15,000 | Internal messaging, leadership talking points, FAQ |
| Resistance management | $10,000-$15,000 | Addressing AI anxiety, 1:1 coaching for resistant teams |
| Champion network coordination | $5,000-$10,000 | Supporting and incentivizing 5-10 internal champions |
Prosci’s benchmark data is the anchor here: projects with excellent change management succeed 88% of the time vs. 13% with poor change management (Prosci, data from 4,500+ projects). Companies that invest 8-12% of total project budget in change management achieve adoption rates 60-80% higher than those spending less than 5%.
Category 4: Measurement and Governance — $30,000-$50,000
| Line Item | Cost Range | Notes |
|---|---|---|
| Baseline measurement (before deployment) | $10,000-$15,000 | Establishing KPIs: cycle time, cost per transaction, quality |
| Dashboard development | $10,000-$20,000 | Tracking adoption, productivity, and financial impact |
| Governance framework implementation | $5,000-$10,000 | Acceptable use policy, review processes, escalation paths |
| Quarterly review process | $5,000-$10,000 | Structured assessment cadence with kill/continue criteria |
BCG finds 60% of companies fail to define any financial KPIs for AI. The CFO decision framework research in this repository shows AI projects with pre-approval financial success metrics achieve a 54% success rate vs. 12% without (Pertama Partners, n=2,400+ enterprise AI initiatives).
Category 5: Productivity Dip Absorption — $40,000-$60,000
| Line Item | Cost Range | Notes |
|---|---|---|
| Reduced output during adoption (3-6 months) | $30,000-$50,000 | 15-25% productivity loss across pilot teams |
| Help desk / support escalation | $5,000-$10,000 | Dedicated support for first 90 days |
| Iteration and course correction | $5,000-$10,000 | Budget for pivoting when first approach does not work |
This is the line item that most budgets miss entirely. SmartDev’s 2025 analysis of SME AI implementations documents a 15-25% productivity loss during the 3-6 month adoption period. Pretending it will not happen does not make it free — it just makes it unmanaged.
The Year 1 Budget Summary
| Category | BCG Ratio | Dollar Range | Key Line Items |
|---|---|---|---|
| Algorithms/Models | 10% | $50,000 | Model access, configuration, prompt engineering |
| Technology/Infrastructure | 20% | $100,000 | Licensing, integration, data prep, security |
| People/Processes | 70% | $350,000 | Workflow redesign, training, change management, measurement, productivity dip |
| Total | 100% | $500,000 |
Why This Ratio Is Hard to Follow — and Why It Matters Anyway
Three structural forces push companies away from the 10-20-70 ratio:
1. Technology costs are visible; people costs are hidden. A $50/seat/month license appears on a purchase order. The productivity dip during adoption appears nowhere — it looks like “people are slow this quarter.” The change management lead’s time appears in salary, not in the AI budget. Without deliberate effort to surface people costs, the budget naturally drifts toward technology.
2. Vendors sell technology, not organizational change. Every AI vendor’s sales team pitches a license fee and a deployment timeline. None of them pitch “you will also need $350,000 in workflow redesign and training.” The commercial incentive is to make the technology purchase feel like the whole investment.
3. CFOs budget in familiar categories. Software licensing is a known budget category. “Workflow redesign consulting” is not. “Productivity dip absorption” is not. The 70% requires CFOs to budget for things that do not fit neatly into existing line items, which is why many simply do not.
The organizations that overcome these forces are the ones BCG finds in the top 25% of AI value capture. BCG’s 2025 AI Radar data shows leading companies prioritize an average of 3.5 use cases (vs. 6.1 for laggards), allocate 80%+ of their AI investment to reshaping core functions, and generate 2.1x greater ROI than their peers.
The 10-20-70 framework is not prescriptive to the dollar. No CFO should treat it as an exact ratio. But it is a diagnostic: if your AI budget is 60% technology and 25% people, the research consistently shows you are in the 75% that do not see meaningful returns.
Key Data Points
- 10-20-70: BCG’s recommended AI investment ratio — 10% algorithms, 20% technology, 70% people/processes
- $350,000 of a $500,000 budget: What the “70% on people” looks like for a 500-person company
- 72%: CIOs reporting they are breaking even or losing money on AI (Gartner, n=506, May 2025)
- 74%: Companies stuck in pilot mode, never reaching measurable ROI at scale (BCG, n=1,803, January 2025)
- 44% vs. 5%: Global AI spending growth vs. training budget growth in 2026 (Fortune, March 2026)
- 88% vs. 13%: Project success rate with excellent vs. poor change management (Prosci, n=4,500+ projects)
- 60%: Knowledge workers using AI without formal training (Microsoft Work Trend Index, 2026)
- 40%: Additional productivity captured by companies investing in people first vs. technology first (EY, n=16,500)
- 2.3x: Total shareholder return premium for human-centric AI investors vs. technology-first (Bain, 250 companies per quadrant, 2020-2024)
- 2.1x: ROI advantage for companies that prioritize fewer use cases with deeper investment (BCG, 2025)
- 15-25%: Productivity loss during 3-6 month AI adoption period (SmartDev, 2025)
What This Means for Your Organization
The 10-20-70 framework answers the question that haunts every CFO evaluating an AI investment: “Where does the money actually go?”
The honest answer is uncomfortable. Most of the money goes to things that do not look like technology. It goes to a consultant spending six weeks mapping how your accounts payable process actually works before anyone configures an AI tool. It goes to a training program that takes 500 people offline for a day. It goes to a change management lead who spends six months in one-on-ones with resistant department heads. It goes to a measurement system that tells you whether the investment worked.
If you are a 500-person company budgeting $500,000 for AI, the question is not “which tool should we buy?” The question is “are we prepared to spend $350,000 on the people and process changes that make the tool worth buying?” If the answer is no, you should either reduce scope (one workflow, not five) or delay until the organization is ready. Deploying tools without the supporting investment is how 74% of companies get stuck in pilot mode permanently.
The practical starting point: take your current AI budget and audit the ratio. If technology exceeds 40% and people investment is below 50%, rebalance before adding new tools. The rebalancing does not require new money — it requires reallocating money you are already planning to spend from technology you will not fully use to people who will actually use it.
Sources
- BCG, “From Potential to Profit: Closing the AI Impact Gap” (January 2025, n=1,803 C-level executives, 19 markets, 12 industries) — https://www.bcg.com/publications/2025/closing-the-ai-impact-gap — Independent consulting survey. High credibility for strategic insights; note BCG sells AI transformation services.
- BCG, “The Widening AI Value Gap” (September 2025) — https://media-publications.bcg.com/The-Widening-AI-Value-Gap-Sept-2025.pdf — Follow-up to January survey. Same credibility caveats.
- McKinsey, “The State of AI” (March 2025, n=~1,933 organizations) — https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai — Largest recurring AI adoption survey. Strong methodology; note McKinsey sells AI implementation services.
- Gartner survey of CIOs (May 2025, n=506 CIOs and technology leaders) — via https://hrexecutive.com/new-research-from-gartner-phenom-and-others-reveals-ais-roi-problem-plus-industry-roundup-and-news/ — Independent analyst firm. High credibility for enterprise technology assessment.
- Gartner CHRO survey (December 2025, n=110 CHROs) — https://www.gartner.com/en/newsroom/press-releases/2026-3-16-gartner-identifies-top-change-management-trends-for-chros-in-age-of-ai — Small sample but targeted to senior HR leaders. Credible for directional insights.
- EY, “Work Reimagined Survey” (2025, n=16,500) — via change management methodologies research — Large independent survey. High credibility for workforce findings.
- Bain & Company, “Want More Out of Your AI Investments? Think People First” (2025, 250 companies per quadrant, S&P Capital IQ data, 2020-2024) — https://www.bain.com/insights/want-more-out-of-your-ai-investments-think-people-first/ — Rigorous financial analysis using public market data. High credibility; note Bain sells consulting services.
- Prosci change management benchmarking (n=4,500+ projects) — https://www.prosci.com/blog/cost-benefit-analysis-change-management — Industry standard for change management ROI data. High credibility.
- Microsoft Work Trend Index (2026) — via Fortune — Vendor-produced but widely cited. Large sample, transparent methodology.
- Fortune, “Companies are pouring billions into AI and cutting training budgets” (March 17, 2026) — https://fortune.com/2026/03/17/ai-economy-workplace-investment-human-potential-competitive-advantage/ — Independent business journalism. High credibility.
- SmartDev, “True Cost of Generative AI for SMEs: 5-Year Breakdown” (2025) — https://smartdev.com/gen-ai-implementation-cost-sme/ — Industry analysis with specific cost benchmarks. Moderate credibility; useful for practical cost ranges.
- AI Smart Ventures, “AI Implementation Cost: A Budget Guide for 2026” — https://aismartventures.com/posts/how-much-does-ai-implementation-cost-a-budget-guide-for-2026/ — Practitioner guide. Moderate credibility; useful for implementation cost ranges.
- Exceeds.ai, “What Is The 10-20-70 Rule For AI ROI?” — https://blog.exceeds.ai/10-20-70-rule-ai/ — Secondary analysis of BCG framework with case study data. Moderate credibility.
Created by Brandon Sneider | brandon@brandonsneider.com March 2026