Boston Consulting Group’s research on scaling AI across large enterprises identified a persistent pattern: organizations that succeed with AI invest fundamentally differently from those that stall after initial pilots. BCG codified this finding as the 10-20-70 rule, which states that only 10% of the effort in successful AI transformation involves algorithms and models, 20% involves data and technology infrastructure, and 70% involves business process transformation and people change management.

The 10-20-70 Breakdown

10% - Algorithms and Models

The AI models themselves represent the smallest share of the effort required to achieve business impact. Whether an organization uses a pre-trained foundation model, fine-tunes an open-source model, or trains a custom model from scratch, the algorithm selection and model development phase is typically the most well-understood and best-resourced part of the journey. Data science teams and ML engineers generally know how to build models. The challenge is rarely the model itself.

20% - Data and Technology

Data infrastructure, integration, and technology platforms account for roughly 20% of the scaling effort. This includes data pipelines, feature stores, model serving infrastructure, monitoring systems, and integration with existing enterprise systems. Organizations often underestimate the complexity of connecting AI systems to legacy data sources, maintaining data quality at scale, and building the MLOps infrastructure needed for production deployment.

70% - Business Transformation

The largest share of the effort involves changing how people work, how decisions are made, and how business processes operate. This includes redesigning workflows to incorporate AI outputs, retraining employees to work alongside AI systems, adjusting incentive structures to encourage adoption, managing organizational resistance, and restructuring roles and responsibilities. This is where most AI scaling efforts fail: not because the model does not work, but because the organization does not change to use it effectively.

Why Most Organizations Get This Backwards

BCG found that organizations that struggle with AI scaling typically allocate resources in the opposite proportion: they invest heavily in algorithms and technology while treating organizational change as an afterthought. The result is technically functional AI systems that sit unused because the people and processes around them have not changed.

A common failure pattern is the “proof of concept graveyard,” where organizations successfully demonstrate AI capabilities in isolated experiments but cannot move them into production because the business processes, governance structures, and employee skills needed to operationalize the AI do not exist.

Applying the Framework

Organizations using the 10-20-70 rule typically restructure their AI programs around several practices:

Cross-functional teams. Rather than building AI in a centralized data science lab and handing it to business units, successful organizations embed AI teams within business functions. This ensures that model development is driven by real business problems and that the people who will use the AI are involved from the start.

Change management as a first-class activity. The 70% allocation means dedicating named individuals and budgeted resources to training, communication, process redesign, and stakeholder management. This is not an add-on to the AI project; it is the majority of the AI project.

Executive sponsorship with operational involvement. Successful AI scaling requires executive sponsors who do more than approve budgets. They actively participate in removing organizational barriers, resolving conflicts between AI initiatives and existing processes, and modeling the behavioral changes they expect from their teams.

Iterative deployment. Rather than attempting organization-wide rollouts, successful organizations deploy AI capabilities incrementally, learning from each deployment cycle and adjusting their change management approach based on what they observe.

Evidence Base

BCG’s research draws on analysis of over 1,000 AI initiatives across industries. Their findings consistently show that organizations that achieve significant financial returns from AI are those that treat it as a business transformation program with a technology component, not a technology project with a business component.