As artificial intelligence investment surges across industries, real progress varies widely. Drawing on conversations with more than 50 European companies, we explore who is truly prepared for AI‑driven disruption and who risks being left behind.
“At every workshop you will see, in the backyard, a heap of old iron, a few wheels, a few levers, a few cranks, broken and eaten with rust. Twenty years ago, that was the pride of the city. People flocked in from the country to see the great invention; now it is superseded, its day is done. All the boasted science and philosophy of this day will soon be old”
Artificial intelligence (AI) is increasingly seen as an integral part of modern business strategy, capable of driving future efficiencies and productivity. But although investment in the technology itself is at unprecedented levels, what is industry uptake actually like?
Over the past six months the EMEA Credit Analysts team at Columbia Threadneedle Investments has conducted an AI engagement exercise. Our objective is to assess how well companies and industries might be able to adapt to an upcoming period of AI-driven change.
Methodology
Technologies come and go. In this instance, however, we are interested in the human behavior behind the technology. Why is a company deploying this technology? What’s management’s track record with technological change? Is the governance around it appropriate? What is the benefit to customers? How will it impact competition or barriers to entry?
To find out, we talked to more than 50 European companies under our coverage across communications, utilities, consumer, insurance, transportation, banking and real estate sectors. Staff representation was from management, investor relations, chief data officers and heads of AI adoption.
We asked a list of tailored questions – written in collaboration with our internal AI experts (see appendix) – designed to assess how well our companies are likely to be able to handle and adapt to change. We think engagement is crucial to understanding a business. We want to sit down face to face with leaders and challenge their thinking. Businesses will publicly say good things about what they are doing, but are they aware of what they don’t know, and would they really say anything negative about the issues they are facing? We only find out by asking the difficult questions.
Once we had spoken to the companies, we collated all the information from the exercise and used our own AI tools to help us spot the trends. This exercise is a great example of how we integrate company engagement into our investment process, showcasing our research intensity and collaboration, as well as our deep understanding of technology from a fundamental and investment perspective.
Our findings: leaders, laggards, followers
Leaders: early adopters
What makes these companies truly formidable is their proprietary data. Relx and Experian sit on treasure troves of trusted, high-quality information that make their AI output genuinely better than their competitors. Having good proprietary data means the answers generated by the large language models, or LLMs, are better. It is a moat that is nearly impossible to replicate.
Elsewhere, telecoms and utility companies are “quiet winners”. AI tools mean grids and networks can be analysed in real time, helping to reduce both fault resolution times and manual labour costs. Customer service becomes better and cheaper as AI-powered agents handle customer enquiries. All of which means these companies will be able to run more efficiently with better margins. Management at Swisscom, for example, says AI is the biggest transformation in their history.
As a result of this adoption, we expect these headcount-heavy industries to cut jobs. Here we are looking for companies to upskill their remaining employees and create a culture of commerciality around the technology. A strong management team is essential to long-term planning and value creation. A good sign that this is happening is if AI investment is structurally separated from cost-optimisation mandates, or if it comes under a fully accountable strategy overseen by the CEO. If not, it generally means AI initiatives are reported through operations, meaning they are all about cost. Although AI experimentation requires tolerance for failure (we applaud the approach at United Utilities where management empowers talented individuals to be creative), companies need to deploy AI safely to avoid reputational or financial damage.
Laggards: investment yet to feed through
Banks are at the start of the AI investment cycle and are spending large sums in a bid to capitalise on AI. CaixaBank, for example, has committed €5 billion through its Cosmos Plan and BBVA has deployed ChatGPT to 11,000 employees globally with an impressive 90% adoption rate (and employee time savings of two to three hours a week).
For most banks there is a pause between initial investment and tangible financial targets and subsequent results that impact profitability. Santander, however, stands out in this regard. It expects AI to contribute to pre-tax profitability, citing €1 billion per annum by 2028, focussed 70% on costs and 30% on revenues. It is targetting transformational cost-efficiency improvements, a good part of which will be driven by AI, and has already reported tangible progress in 2026.
However, the banking sector’s Achilles heel is universal: data infrastructure. AI capabilities depend on accessible, clean, well-governed data. Every bank we talked to raised the issue of legacy systems, data fragmentation and poor data quality constraining AI deployment. What is more, many banks are heavily dependent on the same vendors: Microsoft, Google and OpenAI. If all these institutions rely on the same technology for vital processes, could it increase systemic risk, and would that raise eyebrows at the regulators? One major bank identifies “operational and third-party dependency” and vendor lock-in as key risks.
The winners here have invested in data platform investments and cloud migration and have a chief data officer role within the reporting structure. Standard Chartered and Barclays are also on the right track here.
Followers
Real estate and mining bring up the rear, albeit for different reasons.
Real estate companies are adopting a measured, infrastructure-first approach to AI. Many possess extensive proprietary datasets on their assets and tenants, some of which already support in-house analytics capabilities. While mainstream AI tools have been adopted across various functions – including leasing, valuation, tenant engagement and asset acquisition – the real estate outlook varies by subsector. For office landlords, AI presents a potential threat if widespread white-collar job losses materialise. This particularly affects tenants with substantial backoffice operations vulnerable to automation. Conversely, datacentre operators view AI as an opportunity that could drive incremental leasing demand. Meanwhile, shopping centre owners see limited threat given the inherently physical nature of retail experiences, especially in a landscape where e-commerce is already well established.
Importantly, real estate companies are fundamentally asset-intensive rather than labour-intensive, resulting in already-high operating margins. Consequently, the urgency to capture benefits from headcount reduction or productivity enhancement is less pronounced than in more labour-dependent sectors.
Mining faces even steeper barriers. The physical nature of these operations require extensive safety trials before any large-scale deployment of AI. Vale stands as the exception, with more than 1,000 AI models deployed in areas such as autonomous haulage and drilling since 2017. It also plans to have 90 autonomous trucks by 2028, which are set to deliver 15% productivity gains.
Efficiencies not transformation
Perhaps the most telling discovery across all the sectors is the “no headcount reduction paradox”. Despite claims of improved efficiency, the majority of companies are not cutting jobs. Most banks we spoke to said it was too soon for such a move, but the cost base of a bank is 60% people – so there is plenty of room. Real estate firms emphasised the need for staff reallocation, while mining businesses said cutting staff was not their main objective. Even tech companies expressed concerns about job destruction.
This suggests three things: either a genuine commitment to human-centric transformation; organisational resistance to change; or an overstating of efficiency gains. When companies suggest employees are saving several hours a week but don’t adjust their staffing, where has that productivity gone? Our sense is that headcount reductions are coming but concrete plans are still a few quarters away.
Investment outcomes
For investors, the message is clear: look to the companies with management strength that are embracing AI and can make money from the technology, not those still figuring out how to use it. Certain players within information services and the telecoms and utility sectors have the maturity, data advantages and revenue traction to which banks can only aspire.
Within banks, Santander and BBVA lead the pack with genuine ambition and scale. The real test comes next year when the massive investments reveal to what extent they will provide competitive differentiation and the level of return on investment.
The bottom line
The AI revolution is real and happening right now. The winners are the ones who started running years before everyone else even heard the starting gun.
Over the long term it is always the human behavior behind the technology that drives the performance of the company. We are looking for companies with mature data infrastructure and governance, with AI-literate management teams encouraging upskilling and innovation over pure cost cutting. Based on our engagement, those companies are starting to edge ahead.
As ever, the tools will continue to change, but the disciplines of initially adoption and secondly adaptation, remain constant. The companies that recognise this – and more importantly act on it – won’t just survive the next wave of innovation but define it.
Appendix
Our AI questions:
Governance
- Who has overall responsibility for AI within the company. Is it at C-suite or lower level?
- Who has overall responsibility for the governance process around AI? How do you balance governance with pursuing opportunities?
- How are senior managers incentivised to embrace AI?
- Which part of the organisation is driving the change? Is it at the user level or centralised with IT?
- How are you ensuring that you have the right talent mix at all levels of the company to ensure that you can grow the use of AI while managing the key risks?
Opportunities
- How are you making sure you don’t fall behind peers? Are you building a differentiator or are you just trying to catch up?
- Can you provide specific examples of doing things more efficiently (faster and/or cheaper) or better? (Are you doing tasks which otherwise would not have been possible?)
- Have you been able to reduce headcount as a direct result of AI adoption?
- Are you changing your operational business model or your business strategy due to AI? Could you please provide specific examples?
- How do you prioritise AI use cases? Is it driven by users, or by executing a big picture plan driven by management? How do you promote shared learning within the organisation?
Competition
- Which AI startups/products should investors be aware of, and what do startups need to do to gain market share? How much of a threat are they to your business?
Data
- From a workflow perspective, getting data available to easily load into the AI is very important. It determines whether automation is possible or not. What challenges have you faced around data?
Risks
What do you see as the key risks to the business AI poses?
