BCG's "Build for the Future 2025" report, published in September 2025 under the title The Widening AI Value Gap, is the most granular read on AI maturity any of the major consultancies has put on the table. It surveyed 1,250 senior executives across nine industries and more than 25 sectors, and it scored each organisation against 41 foundational capabilities spanning strategy, technology, people, governance, and outcomes. The output is not a mood-of-the-market survey. It is a capability audit — and it draws a hard line between companies that build AI capability systematically and those that experiment without a structural plan.
The distribution is brutal. Only 5% qualify as "future-built" — BCG's top tier, the organisations systematically building cutting-edge AI capability across functions and consistently generating value. A further 35% are "scalers", beginning to turn capability into measurable return. The remaining 60% are laggards, reporting minimal revenue or cost gains and without the foundations to scale. The top 5% deliver 1.7 times the revenue growth of laggards, 3.6 times the three-year total shareholder return, 1.6 times the EBIT margin, and 2.7 times the return on invested capital. They also file 3.5 times more patents. These are not rounding-error differences. They are a structural performance gap, and BCG's own data shows it widening rather than closing.
The question for any DACH mid-market executive reading this is not whether the gap is real. It is whether your organisation is building the capabilities that close it — or accumulating tools that leave it exactly where it is.
The 60% problem
The most consequential number in BCG's data is not the 5% at the top. It is the 60% at the bottom. These are not companies that ignore AI. The laggard tier is dominated by organisations BCG would, in its four-stage maturity model, call emerging — they are aware of AI's potential, they are experimenting, many have run pilots, some have deployed individual tools. What they have not done is systematically build the capabilities that let AI operate at scale across the business. So the value never materialises.
This is where most DACH Mittelstand companies sit. They have ChatGPT licences. They have evaluated Copilot. A department head has run a proof of concept. The Geschäftsführung has discussed AI strategy in two consecutive board meetings. All of that is real activity — and none of it, on its own, moves an organisation out of the laggard 60%. Activity is not capability. BCG's framework is explicit that the leap from experimentation to value comes from interconnected systems, not from a portfolio of disconnected experiments.
The gap is not about technology acquisition. It is about capability construction. The 41 capabilities BCG assesses cut across data infrastructure, talent, governance, operating-model design, and cross-functional integration. Laggards may have a handful of these in place, typically in isolated pockets. Scalers have built them as a connected system that reinforces itself. This is exactly the distinction we draw between AI readiness and AI maturity: readiness answers whether you can deploy your first production workflow; maturity answers whether you have the infrastructure to deploy your tenth. BCG's data confirms that most companies have enough readiness to experiment and nowhere near enough maturity to scale — and that more tools will not close that gap.
What the future-built 5% actually do
BCG's analysis is specific about what separates the top tier, and it is not the choice of model or the size of the technology budget. It is how deeply AI is wired into how the business actually operates.
They redesign processes, not just tasks. Future-built companies do not bolt AI onto existing workflows — they restructure how work flows through the organisation. This converges almost exactly with McKinsey's independent finding in The State of AI in 2025: AI high performers are roughly 2.8 times more likely to fundamentally redesign workflows when deploying AI (55% versus 20% of everyone else), and workflow redesign is the single strongest predictor of enterprise-level AI impact McKinsey tested. Two studies, different consultancies, different methodologies, different samples — same conclusion. When findings converge like that, you stop treating it as one firm's framework and start treating it as a fact about how AI value is created.
They invest differently, not just more. Future-built firms plan to spend more than twice as much on AI as laggards, and BCG reports they allocate a markedly larger share of IT budget to it. But the headline figure matters less than the allocation. Laggards spend on tools and licences. Future-built firms spend on data infrastructure, talent development, and operating-model redesign — the parts that compound. The investment profile reflects a different theory of where AI value comes from, and for a Mittelstand business with a finite IT budget, that allocation discipline matters more than the absolute number ever will.
They concentrate on the core, not the periphery. BCG finds that roughly 70% of AI's value sits in core business functions — sales and marketing, manufacturing, supply chain, and pricing — while a support function like IT accounts for a far smaller share. Future-built companies point their AI programmes at these high-leverage domains rather than scattering experiments across back-office functions where the financial impact, even when the pilot "works", is immaterial. For the Mittelstand, this is the most actionable line in the entire report: the value is in the operating core, not in another internal-helpdesk chatbot.
They are building for agents. BCG reports that AI agents already account for 17% of total AI value in 2025, rising to a projected 29% by 2028, and that a third of future-built firms already use agents compared with 12% of scalers and almost none of laggards. The top tier is designing workflows now that accommodate agentic AI — systems that take multi-step actions autonomously rather than only responding to a prompt. That is the trajectory from Level 02 to Level 03 in our Three Levels framework: from AI as specialist to AI as operator. The companies building for it today are not chasing a trend; they are positioning for where two-thirds of net-new AI value will land.
Why 41 capabilities, and why that should worry you
The number 41 is not decorative. It reflects the actual breadth of organisational change required to move from emerging to scaling, spanning strategy and ambition, data and technology infrastructure, talent and skills, governance and risk, operating model, and innovation culture. No single one of them is the unlock; the maturity comes from how they connect.
Most Mittelstand companies have never mapped their own capabilities against a framework anywhere near this granular. They know, in broad strokes, that they need "better data" or "more AI talent." What they lack is a structured view of which specific capabilities are present, which are half-built, and which are simply absent. Without that map, investment decisions run on intuition — and intuition is precisely what put 60% of the market in the laggard tier.
The parallel to the six-dimension diagnostic in the AI Operating System methodology is direct. Our diagnostic assesses workflow readiness, data accessibility, decision authority, compliance posture, team capacity, and operating-model clarity. Those dimensions map onto BCG's capability categories — not one-to-one, but as a deliberately compressed subset built for the DACH mid-market. BCG's 41-capability assessment is designed for global enterprises with dedicated transformation teams and the budget to run them. The six-dimension diagnostic is designed for a Geschäftsführer who needs to know where to start this quarter.
Why the gap compounds
BCG's data shows a self-reinforcing dynamic: leading companies generate measurable AI value, reinvest the profit into more capability, talent, and technology, and pull further ahead. The capability advantage shortens deployment cycles, which generate more operational data, which improves AI performance, which justifies the next round of investment. It is the same compounding mechanism we describe in The Compounding Cost of AI Inaction — only here it is observed at industry scale rather than inside a single company.
For the laggard 60%, the window to catch up is narrowing — and not because the technology is getting harder. The technology is getting cheaper and more accessible by the quarter. The window narrows because the capability gap compounds. Every quarter a future-built competitor runs AI workflows in production, it accumulates operational learning that cannot be bought: which data pipelines break under load, which governance structures actually hold, which delegation models the team trusts enough to use. That knowledge is organisational, not technical, and the only way to build it is to run the thing in production and learn from what goes wrong.
The implication for the Mittelstand is unambiguous. The move from emerging to scaling is made by starting now — with one workflow, in the operating core — and building capability through deployment rather than through another planning cycle. The companies that will be scalers in two years are not the ones writing AI strategy decks today. They are the ones deploying their first production workflow this quarter and learning from it.
From BCG's framework to your next step
BCG's data validates the principle the AI Operating System is built on: maturity is a function of capabilities, not tools. You do not mature by buying more AI licences. You mature by building the organisational infrastructure — data pipelines, governance models, delegation rules, review cycles, operating-model clarity — that lets AI operate reliably at scale.
The practical question for any Mittelstand company reading this research is narrower than the 41-capability assessment makes it sound: which capabilities do we have, which are we missing, and which one do we build first? The answer is not the same for every business. It depends on your industry, your current data infrastructure, your regulatory environment, and the specific workflows where AI can create operating leverage. Most companies in the laggard tier do not need an enterprise-grade 41-capability audit. They need a focused diagnostic that names the two or three gaps blocking the move from experiment to production.
That the consulting industry takes this market seriously is no longer in question — BCG reported $14.4 billion in revenue for 2025, with 25% of it (roughly $3.6 billion) tied to AI-related work, the first disclosure of its kind from a Big Three firm. The opportunity is real and the spend is real. The only open question is whether your organisation builds capability systematically or keeps experimenting without a structural plan.
A Fit Call maps your organisation against BCG's maturity logic and names the two or three capability gaps standing between you and your first scaling workflow — before another year of pilots leaves you in the 60%.
References: BCG, "The Widening AI Value Gap — Build for the Future 2025," September 2025 (media-publications.bcg.com; press release); McKinsey, "The State of AI in 2025: Agents, Innovation, and Transformation," November 2025 (mckinsey.com); BCG, "BCG Reports $14.4 Billion in Revenue," April 2026 (prnewswire.com).