In a recent episode of the Cambridge Executive Business Insights: Rethinking AI podcast hosted by Professor Jaideep Prabhu, guest Thomas Wenstrøm, Managing Director and Partner at BCG and Nordic Leader of the Tech and Digital Advantage practice, shares a candid assessment of the complex and nuanced reality of AI use in many organisations.
His message is direct: while AI’s value is real, it remains unevenly distributed. Only a small number of organisations consistently realises significant returns on their AI investments, while a much larger cohort languishes with little more than pilot schemes and unfulfilled hype. Here, we explore the forces shaping the uneven adoption of AI, the patterns that characterise successful transformations, and why a strategic, people-centred approach, rather than just more technology, is the true differentiator.
The promise and reality of AI: value creation or value mirage?
AI has long promised sweeping improvements to productivity, innovation and decision-making. Yet Thomas Wenstrøm observes that a stark disparity persists between those who truly harness AI and the many that struggle to exceed the demo phase.
Only really 5% of companies are generating significant return from their AI investments… while 60% are still creating little or no material impact from their AI effort despite substantial investments.
This unevenness is not a matter of access to technology. It is not even, necessarily, a matter of scale. Rather, it is a function of how organisations approach AI: purposefully, strategically, and with a willingness to reimagine not just systems but also the very processes, roles and behaviours that shape how work is done.
Moving beyond pilots
AI pilots and demos are seductive, offering visible wins and the excitement of early adoption. But, as Thomas Wenstrøm explains, they are fundamentally easier to celebrate than to scale.
Organisations fall into several common traps:
- Process inertia: they roll out proof-of-concept demos but do not redesign the underlying business processes required to embed AI into daily operations.
- Lack of true ownership: the initiatives remain in the domain of technology or innovation teams, rather than being championed by business leaders with direct responsibility for performance and growth.
- P&L disconnect: too few efforts are explicitly tied to metrics that matter: revenue, margin, productivity, customer outcomes.
The result is what Thomas Wenstrøm calls “letting a thousand flowers bloom”, a proliferation of low-impact pilots that fail to take root.
The anatomy of AI winners: five defining patterns
From his work with leading global organisations, Thomas identifies five critical practices that distinguish the 5% of companies realising meaningful value:
- Strategic alignment: AI initiatives serve the enterprise as a whole, not as siloed technology projects. They target areas of real strategic relevance, where success will be meaningful for the organisation.
- Focus on scale, not scatter: rather than dispersing efforts across numerous small pilots, leaders invest in a handful of transformative use cases, seeking to reshape key processes or invent new services.
- Centralised, business-led delivery: large-scale AI undertakings must be governed centrally and led by those who own the outcomes, to ensure technology and business are deeply integrated.
- Value-driven capability development: the necessary human and technical capabilities are built iteratively and in pursuit of specific value, not as an end in themselves.
- Relentless focus on process and adoption: most of the work (and the challenge) is not technological but organisational: getting people to adopt, adapt and thrive in AI-powered environments.
10% of the effort is about the algorithms, 20% is about the tech, and 70% is really about the people, process adoption and so on.
The evolution of enterprise AI: three distinct waves
AI in the enterprise has not been a single, linear story; it has evolved through phases, each unlocking new opportunities and raising new challenges.
- Traditional AIML: the “efficiency and personalisation” era
The first wave was dominated by classic machine learning applied to forecasting, pricing, promotions and customer personalisation, and was especially strong in sectors like consumer goods and telecoms. These use cases often targeted customer-facing applications (e.g., churn prediction, pricing optimisation), leveraging data-rich environments to drive both revenue and customisation.
2. Generative AI: accelerating productivity and procurement
The advent of generative AI introduced new possibilities, including the automation of document analysis, customer service interactions, and even creative tasks. Use cases included:
- Automated cross-checking of contracts, invoices and purchase orders to capture “value leakage”.
- Productivity gains in functions such as HR and finance, exemplified by organisations like IBM, which realised billions in impact from such transformations.
3. Agentic AI: reinventing workflows and processes
With more recent agentic AI developments, the focus is shifting to the reinvention of dense, interconnected workflows, moving from incremental improvements to end-to-end transformation. Now, entire functions such as customer service or supply chain management can be largely reimagined by AI-powered agents orchestrating tasks, data flows, and even human collaboration.
Use cases and patterns: from back office to front line
Although the “where” of AI impact differs by sector, certain patterns are clear:
- Support functions (HR, Finance, IT): up to 25% of AI’s value lies in reinventing support operations, with similar models applicable across industries.
- Core business (Sales, Product, Customer Experience): the remaining 75% is highly sector- and company-specific, depending on business models and strategic priorities.
Organisations that succeed focus on two to six areas, combining both support and core business transformation, tailored to where the stakes are highest for their growth or efficiency.
Thomas offered the following examples:
- IBM: reinvented HR and finance with AI, unlocking billions in value.
- AWS: increased software engineer productivity through AI-powered development tools and robust change management.
- L’Oréal: built a virtual beauty advisor, enabling direct consumer engagement and redefining their established business model.
Deploy, reshape, invent
In charting the course for AI adoption, Thomas Wenstrøm recommends a “portfolio approach”, which can be distilled into three modes of transformation:
- Deploy: roll out off-the-shelf AI tools (think: code generation, chat interfaces) to boost individual productivity organisation-wide. This generally accounts for 10-20% of the effort but is necessary for broad AI fluency.
- Reshape: reimagine entire business functions by embedding AI into the core of how work is done. This is where the majority (80%) of investment and change should focus.
- Invent: pioneer entirely new offerings or revenue streams based on AI’s unique capabilities, that is, in areas where established processes do not exist.
Placing disproportionate emphasis on “deploy” i.e. simply hoping widespread use of AI tools will organically scale up to real business outcomes, often leads to disappointment. True transformation requires reshaping the core processes and responsibilities that define the company.
Why most of the work is human: overcoming the organisational barrier
Perhaps the greatest illusion in AI implementation is that technology is the hard part. The evidence says otherwise.
In practice, much of the challenge lies in changing how people work, learn, and interact with technology:
- Change management: convincing the workforce to adopt new ways is slow, messy and often invisible compared to the rollout of shiny new systems.
- Role reconfiguration: roles, incentives, and routines must evolve alongside AI, requiring robust retraining, leadership, and patience.
- Integration, not automation: the real breakthroughs occur when AI’s strengths are paired with human judgement and oversight, rather than simply replacing people.
The outcomes are much stronger when you do both: when you have the intelligence of the machines and human judgement on top.
Where fear of displacement or “data isn’t good enough” narratives persist, engagement and inclusion are critical. In Thomas’ experience, once teams are actively involved in AI initiatives, anxiety is quickly supplanted by enthusiasm and engagement.
Responsible, trusted, and ethical AI: no longer optional
With greater power comes heightened scrutiny, both internally and externally.
- Responsible AI: companies are taking the questions of bias, hallucination, and ethical oversight seriously. But while risks exist, these should not be a blanket excuse for inertia; organisations are innovating ways to advance responsibly and transparently.
- Trust and the human connection: for many, there remains a reluctance to insert AI “front-of-house” with customers, especially where nuanced judgement or relationships are paramount, such as in customer service. Trust in the technology (internally and externally) must be earned through successful pilots, transparent processes, and human oversight.
Sector and regional perspectives: the view from the Nordics
From a Nordic perspective, Thomas Wenstrøm sees both promise and peril. While digital maturity and talent are strong, and expectations for AI value are strikingly high, investment remains largely at the “deploy” level, incremental, broad and often decoupled from core business reinvention.
Unless more focus is placed on strategic, high-impact transformations, a “value bubble” may emerge in the region, where the opportunity is anticipated but never fully realised in practice.
The pace of change: diffusion and the coming inflection
Does this mean the AI revolution will slow? This is unlikely. Thomas Wenstrøm is optimistic, momentum is clearly accelerating. In the past year alone, more organisations are shifting to serious, programmatic approaches, with the number of CEOs personally steering the AI agenda having doubled. The barriers are real, rooted in organisational inertia, but the urgency is only increasing.
This will be a journey for the next years… But I do expect that there will be a quite significant increase [in value realisation] over the next couple of years.
Recommendations for leaders
If there is one insight to take from Thomas Wenstrøm, it is this: meaningful AI transformation is not about chasing every new tool, nor about mere technology accumulation. It is about deliberate, high-stakes bets anchored to core business imperatives, orchestrated from the top, and implemented with relentless attention to people, process and ownership.
The short-term imperative for leadership:
- Identify two to six areas where success can be tied to economic value.
- Drive adoption and redesign, not in isolation, but in deep partnership between business leadership and technology.
- Invest in change management and workforce capability as heavily as you do in data and algorithms.
From technological hype to purposeful impact
The AI revolution in business will not be won by those who gather the most pilots or amass the largest tech stacks. Instead, it will be led by organisations willing to confront the hard work of transformation: realigning strategy, empowering leaders, investing in adoption, and ultimately realising that the most valuable intelligence is shared between people and machines, ambition and discipline, vision and execution.
