The world’s leading strategy firms – BCG, McKinsey, and Bain – are being paid billions to help companies navigate AI transformation. When three firms that rarely agree on anything reach the same conclusion, it’s worth paying attention.
Their shared verdict: most organisations are solving the wrong problem.
They’re treating AI adoption as a technology challenge when it’s really an organisational change challenge. The firms differ on terminology and emphasis, but the underlying diagnosis is consistent and it has direct implications for how leaders should be spending their time and budget.
BCG: You’re Spending Your Money in the Wrong Place
BCG’s most important contribution to this debate is the 10-20-70 Rule, a framework that exposes where most AI investments go wrong.
Successful AI transformation breaks down as follows:
- 10% on algorithms and AI models
- 20% on technology infrastructure and data pipelines
- 70% on people, process redesign, and cultural change
Most organisations invert this. They spend heavily on the 10% and evaluating models, running vendor comparisons, negotiating licences while treating the 70% as an afterthought. BCG research confirms that organisations that invest deliberately across all three layers triple their chances of capturing the full value of AI.
BCG also identifies three distinct ways organisations can actually create value with AI:
- Deploy – immediate productivity improvements through tools like automation or coding assistants
- Reshape – re-engineering functional areas, such as end-to-end supply chain or marketing operations
- Invent – building entirely new revenue streams or business models that AI makes possible
Most organisations are stuck at Deploy. The firms generating outsized returns are moving toward Reshape and Invent.
McKinsey: The Gap Between High Performers and Everyone Else
McKinsey’s research into what separates AI high performers from the rest identifies six dimensions where they consistently pull ahead:
- Strategy – AI investments are tied directly to specific, high-value business outcomes, not broad digital transformation goals
- Talent – their workforce is “bilingual”, meaning people who understand both the business problem and what AI can and cannot do
- Operating Model – they’ve moved beyond fragmented pilots toward an integrated hub-and-spoke model that allows successful approaches to scale
- Technology – modular, robust infrastructure that doesn’t require rebuilding every time a new use case emerges
- Data – focused on the data that matters for specific competitive advantages, not enterprise-wide data cleaning exercises that consume years and deliver little
- Adoption – workflows have been redesigned so AI is embedded in how work actually happens, not available as an optional tool
The sixth dimension is where most organisations underinvest. A capability sitting unused in a system is not an AI implementation, it’s an expensive experiment.
Bain: Stop Collecting Small Wins
Bain identifies what they call the “micro-productivity trap”: organisations that deploy dozens of AI tools, accumulate small efficiency gains across the business, and then discover that none of it adds up to meaningful bottom-line impact.
Their counter is a more disciplined approach:
Zero-based process design. Rather than layering AI onto existing workflows, define where you want to end up first, what Bain calls a “Point of Arrival”, then work backwards to design the process. This is harder than incremental improvement and produces dramatically different results.
Fewer, bigger bets. Focus on four to five “battleground domains” where AI can deliver a decisive advantage, rather than spreading effort across the organisation. Concentration beats diversification here.
Prepare for agentic AI. AI is moving from tools that respond to prompts toward autonomous agents that plan and execute multi-step workflows with limited human direction. Organisations that haven’t thought through how they’ll govern and oversee this are building toward a blindspot.
What All Three Actually Agree On
Strip away the proprietary terminology and the consensus is clearer than it looks.
Responsible AI is infrastructure, not compliance. Governance, transparency, and human oversight aren’t legal requirements to be satisfied but the foundation that lets organisations move faster with confidence. Audit trails, human-in-the-loop processes for high-stakes decisions, and bias testing are about building systems that can be trusted at scale.
Domain transformation beats tool deployment. Every firm recommends moving from scattered AI tool adoption to targeting entire domains such as a complete software development lifecycle, the full customer engagement journey, for AI-enabled redesign. The unit of transformation should be a business outcome, not a tool.
The workforce gap is real and it’s your problem to solve. AI literacy is becoming a baseline expectation across roles, not a specialist skill. Organisations that treat upskilling as a training department issue rather than a leadership priority are creating a capability debt that compounds over time.
Data strategy is business strategy. The firms that are winning with AI are the ones that have identified the specific data that creates a competitive advantage in specific contexts, and have invested in making that data reliable.
The Organisational Change Implication for Leaders
The frameworks differ in structure. The conclusion is the same.
AI management is 70 to 80 percent an organisational change problem. The technology question, for example which model, which platform, which vendor, is the smallest part of the challenge, and it’s the part most organisations spend the most time on.
The firms getting outsized returns from AI aren’t doing anything exotic with the technology. They’ve made different decisions about people, processes, governance, and focus. Those decisions compound over time. Organisations that continue treating AI as primarily a technology procurement exercise are making a choice, even if they don’t recognise it as one.
The standard sequence; people, process, technology, in that order, remains the right one. AI doesn’t change the sequence. It raises the stakes for getting it wrong.
References
[1] Bain & Company. Unsticking Your AI Transformation.
[2] Boston Consulting Group. (2024, December 12). The Leader’s Guide to Transforming with AI.
[3] McKinsey & Company. (2025, November 5). The State of AI: Global Survey 2025.
[4] McKinsey & Company. Responsible AI Principles.
[5] Boston Consulting Group. Responsible AI.