AI in Family Business: Who Is Exposed and How to Use It

A second-generation owner runs an Iberian distribution company with revenue in the tens of millions. On a Sunday he reads that 86% of large family businesses now use AI in some form. His company has one pilot: a chatbot on the website that answers delivery questions, built by an agency, measured by nobody. His family’s wealth sits in four banks, two holding companies and a spreadsheet his sister’s accountant updates once a year. He is worried about AI in family business in general and in his family business in particular. Will it take his margins? His customers? His children’s interest in the company?

The worry is reasonable. It is also half of the picture. The founder’s AI problem has two balance sheets. The first is the operating company, which AI may disrupt. The second is the family’s own wealth, which AI and the technology around it can finally make visible, comparable and quick to decide on. Most of what is written about AI looks only at the first.

Our view is simple, and it is the spine of this piece: technology is on our side. Put it at your right hand and use it. We run our own work that way, and it is the reason a small, independent firm can see a family’s whole balance sheet faster than most larger competitors. What follows is the evidence for both halves: which sectors are exposed and how much, what family businesses and family offices are actually doing, and what a calm, ninety-day response looks like.

AI in family business: two balance sheets, one founder

Family businesses are not new to technological change. Most of the Iberian companies we see have survived at least one: the move from ledgers to ERP, from fax to email, from wholesale catalogues to e-commerce. What is different this time is speed and reach. Earlier waves of automation went after routine physical and clerical tasks. The current wave reaches into work that used to be safe because it needed judgement, language or expertise: drafting, analysing, answering, summarising, coding.

That matters for a founder in two ways at once.

The operating company. If the company sells knowledge, services or customer contact, a large share of its cost base sits in tasks that AI can now do faster or cheaper. That is a threat if competitors adopt first and an opportunity if the company does. If the company makes or moves physical things, the exposure is smaller and slower, but the office around the physical work, the quoting, planning, purchasing, forecasting and accounts, changes on the same timetable as everyone else’s.

The family balance sheet. The founder’s personal and family wealth is usually less organised than the company. The company has an ERP, a CFO and an auditor. The family has bank statements in different formats, a few holding companies, property held in different names, perhaps a private equity commitment or two, and nobody whose job it is to see all of it at once. This is the half of the problem where technology helps most and costs least, because the work is mostly data: collecting, reconciling, mapping, comparing.

Put the two together and the founder’s question changes. It stops being “will AI hurt us?” and becomes “where does AI change our company, and how do we use the same technology to see and decide on everything we own?” The first question is about exposure. The second is about posture. We think the right posture is the one we apply to ourselves: technology at your right hand, people making the decisions.

Which industries will AI disrupt the most?

Knowledge-intensive services face the most disruption: software and IT services, banking and insurance, professional services such as legal and accounting, marketing and media, and customer operations. McKinsey places about 75% of generative AI’s potential value in customer operations, marketing and sales, software engineering and R&D. Physical, site-based sectors are less exposed and change later.

Three independent sources point the same way. McKinsey’s June 2023 analysis of 63 generative AI use cases found that four business functions, customer operations, marketing and sales, software engineering, and research and development, account for roughly 75% of the total annual value, and named banking, high tech and life sciences among the industries with the biggest impact as a share of revenue. The IMF’s January 2024 analysis observed that, unlike earlier automation, AI reaches high-skilled jobs, which is why advanced economies are more exposed than emerging ones. And the OECD, when it tested for AI effects on labour markets in its June 2026 Economic Outlook, grouped the AI-exposed sectors as information technology, finance and insurance, real estate, professional, scientific and technical services, and management of companies.

Adoption data tells the same story from the other side. Eurostat’s 2025 figures show 62.5% of EU enterprises in information and communication using at least one AI technology, 40.4% in professional, scientific and technical activities, 24.8% in real estate and 10.8% in construction. The sectors that use AI most are the sectors whose work AI can most easily do.

The table below is our editorial map of exposure, built from those sources. “Exposure” means how much of a sector’s typical work AI can already perform or accelerate. It is not a forecast of job losses, and it is not a judgement on any company.

What this means for the Iberian family-business base

Portugal and Spain have a recognisable family-business economy: wine and food, textiles and footwear, tourism and hospitality, construction and materials, wholesale and distribution, and a growing tier of professional and technology services. Read against the table, most of these sit in the lower and medium bands. The bottle, the hotel room, the building and the delivery van are not going to be done by a language model.

That is reassuring and incomplete. In each of these companies, a large share of the administrative and commercial work is exactly the kind that AI does well: answering customers in four languages, preparing quotes and tenders, matching invoices to orders, forecasting demand, writing product descriptions, checking compliance documents. A winery is not exposed. Its export office is. A construction group is not exposed. Its estimating and procurement departments are. The competitive gap opens there, between companies whose office gets faster and companies whose office does not.

The professional-services and distribution businesses on the list are a different case. For a law or accounting firm, an agency, a contact-centre operator or a software house, the core product sits in the high-exposure band. These owners should read the next section closely, because the change touches pricing, not only cost.

How many businesses may be affected: the numbers, dated

The headline numbers are large, and they are often quoted without dates or sources. Here they are with both.

Exposure across the economy. The IMF estimated in January 2024 that almost 40% of global employment is exposed to AI. In advanced economies the figure is about 60%; in emerging markets about 40%; in low-income countries about 26%. The IMF was careful about what exposure means. Roughly half of the exposed jobs in advanced economies may benefit from AI, which raises productivity. For the other half, AI may perform key tasks now done by people, which can lower demand for that work.

Hours of work. The McKinsey Global Institute estimated in May 2024 that, by 2030, about 27% of the hours currently worked in Europe and 30% in the United States could be automated, with generative AI accelerating the shift. About 20% could be automated even without generative AI. The same study expects demand for health and other STEM professionals to grow, and demand for office support, customer service, sales, production work and food services to keep declining.

Jobs and output. Goldman Sachs Research estimated in 2023 that generative AI could expose the equivalent of 300 million full-time jobs to automation, and could raise global GDP by 7%, almost USD 7 trillion, while lifting productivity growth by 1.5 percentage points over ten years. Its economists found roughly two-thirds of US occupations exposed to some degree, and estimated that for exposed occupations between a quarter and a half of the workload could be replaced.

Value at stake. McKinsey’s June 2023 report put the potential annual value of generative AI at USD 2.6–4.4 trillion across the 63 use cases it analysed. For comparison, the report notes, the United Kingdom’s entire GDP in 2021 was USD 3.1 trillion.

Firms, not only jobs: the adoption gap

For an owner, the more useful numbers are about companies. Eurostat reported in December 2025 that 20.0% of EU enterprises with ten or more employees used AI in 2025, up from 13.5% in 2024. The spread is wide. Denmark led at 42.0%; Romania was lowest at 5.2%. In the Eurostat database, Spain stood at 20.3% in 2025, almost exactly the EU average, and Portugal at 11.5%, well below it.

Size matters more than country. In 2025, 55.0% of large EU enterprises (250 or more employees) used AI, against 30.4% of medium-sized and 17.0% of small enterprises. In Portugal, 49.2% of large firms used AI and 9.4% of small ones. Most family businesses in Iberia are small or medium-sized, so for most readers the honest position is this: your largest competitors and suppliers are already using AI, and most of your peers are not yet.

The counterweight: exposure is not displacement

The exposure figures describe what AI could do. They do not describe what has happened. The OECD’s June 2026 Economic Outlook looked for evidence and found “no signs of widespread labour displacement” from business adoption of AI at the industry level. Job vacancies in the industries most exposed to AI rose more than in other sectors over the year to April 2026 in most economies with data, with the United States the notable exception.

That is the posture we recommend. Take the exposure seriously, because it is real and it compounds. Do not mistake it for a verdict. A company exposed to AI that adopts it carefully is in a better position than a company that is not exposed and ignores it. For the wider macro picture this year, including the energy shock the OECD built its forecasts around, see our reading of the global economy in 2026.

AI in family business today: what owners are actually doing

If the question is “how will AI affect my business?”, the most useful benchmark is what other family businesses are doing. Three surveys from the past two years give a consistent picture: wide use, shallow change.

Use is now the norm in large family businesses. Deloitte Private surveyed 1,587 family businesses with revenue of at least USD 100 million in 35 countries between March and June 2025, and published the results on 31 March 2026. Of these, 86% use AI actively or selectively: 44% in many areas of the business and 42% in select functions. Only 2% said they were neither using nor exploring it. The main uses were process efficiency (40%), risk management (39%), customer relationship management (39%) and customer experience (38%).

The returns reported were real. Respondents credited technology investment with increased productivity (68%), improved efficiency (67%) and enhanced decision-making (65%). The gaps were also real. 48% said their investment in the operational technology they need falls short, and 51% said inadequate technology adoption was a moderate or high risk to their growth over the next 12–24 months.

Ambition is modest. PwC’s 2025 Global Family Business Survey interviewed 1,325 family businesses in 62 countries and territories between April and June 2025. Just over three-fifths (61%) cited experimentation with AI as a growth opportunity. Only 3% said they were looking to reinvent their business. Most are adding AI to the existing model, not asking whether the model still holds.

PwC also reported, from family-business data in its 28th Annual Global CEO Survey, that 46% of public family businesses said generative AI had boosted both revenue and profitability. Around a third of non-family CEOs reported increased revenue (29%) or profitability (32%). Where family businesses commit, the evidence suggests they can do well.

The next generation is ahead of the board. PwC’s Global NextGen Survey 2024, based on 917 interviews in 63 territories, found that 73% of next-generation family members see generative AI as a powerful force for transformation. Yet 49% of family businesses had either prohibited AI or not started to explore it, only 7% had implemented it anywhere in the business, and only 14% had a person or team directly responsible for generative AI.

The pattern behind the numbers

Read together, the surveys describe a familiar family-business shape. Large family companies use AI widely, mostly for efficiency, and report good results. Smaller ones are slower. Few are asking the structural question. The next generation sees the opportunity more clearly than the people who control the budget, and responsibility for AI often sits with the head of IT, or with no one.

None of this is a failure of intelligence. It is a failure of ownership. AI touches every function, so it belongs to no function. The fix is governance, not software: one person accountable, one place where the board discusses it, and a clear role for the next generation. Our family governance and education work exists for exactly that kind of question, where the company, the board and the family meet.

How will AI affect my business, concretely?

For an owner, the honest answer comes from the work, not the sector label. List the main functions in the company and, for each, ask three questions. How much of the work is language, documents, numbers or customer contact? How much is repetitive and checkable? How much of the cost base does it carry? Where all three answers are “a lot”, the function will change within a few years, whether or not you start the change. Where the work is physical, relational or rare, it will change slowly, and the gain will come from the office around it.

Then look at your customers and suppliers, not only your own processes. A distribution company whose customers start ordering through automated purchasing systems has to be readable by those systems. A professional-services firm whose clients can draft a first version themselves has to sell what remains: judgement, accountability and the final signature. A tourism business whose guests plan trips with AI assistants needs its information to be accurate wherever those assistants look.

AI in the family office: from 13% to 22%

The second balance sheet is the family’s own. Here the data comes from the two largest annual surveys of family offices, and it shows a gap that most founders will recognise: family offices invest in AI more readily than they use it.

Use. Citi Wealth’s 2025 Global Family Office Report, summarised on Citi’s page of 11 May 2026, found that 22% of family offices have automated operational tasks or use AI for investment analysis, up from 13% in 2024. 16% use AI for investment performance reporting, more than double the previous year. The largest barrier, cited by 57%, is the lack of internal expertise. Citi’s summary is clear on the other condition: data privacy is non-negotiable for family offices, and tools that cannot guarantee security struggle to be adopted.

Investment. UBS’s Global Family Office Report 2026, published on 28 May 2026 and based on 307 family offices with an average family net worth of USD 2.7 billion, found that 65% are already invested in AI across the value chain, from data centre infrastructure to software platforms and semiconductor producers. That is what other families do with their portfolios. Whether and how much a family allocates to AI as an investment theme is a matter for its regulated managers and its investment policy; we have written separately on the difference between innovation and experimentation when a theme becomes fashionable.

So roughly two-thirds of large family offices own AI as an asset, and roughly one in five uses it to run the office. The constraint is not money. It is expertise, data and trust.

AI in wealth management: what it is good at, and what it is not

The work of overseeing a family’s wealth breaks down into tasks that suit technology very well and tasks that do not.

Technology is good at collecting statements from several custodians, normalising their formats, reconciling positions and cash, classifying assets, calculating exposures and performance on a consistent basis, flagging what changed since last month, tracking deadlines, and drafting summaries of long documents. These tasks are repetitive, checkable and data-heavy. Done by hand, they take weeks and are where most errors live. Our glossary entry on consolidated reporting explains why this one task does more for a family’s decisions than almost anything else.

Technology is not good at deciding what the family wants, weighing a sibling’s needs against a parent’s, judging whether a manager is honest, choosing between two reasonable structures, or telling a founder that a cherished investment should go. Those are judgements. They need a person who is accountable for them.

The family offices that use AI well keep that line clear. The ones that struggle either avoid technology altogether and drown in reconciliation, or buy a platform and expect it to make decisions it cannot make.

Put technology at your right hand: what it changes for UHNW families

For ultra-high-net-worth (UHNW) families, the case for technology is not novelty. It is four specific gains, each with a mechanism.

Efficiency. In a family with four banks, the first week of every quarter used to go on collecting PDFs, re-keying positions and chasing missing statements. With feeds or structured imports from each custodian, a reporting platform and a defined data model, that work becomes a scheduled process. People look only at the exceptions: a position that does not reconcile, a fee that changed, a cash balance that moved without explanation. The saving is not just time. It is the errors that disappear because nobody re-types a number.

Productivity. The point of saving time is to spend it on something better. When the data arrives clean, the advisor can spend the week asking the questions that matter. Are two banks holding the same bond fund under different names? Is the private equity programme drawing capital faster than the family’s liquidity plan assumed? Which manager has underperformed its benchmark after fees for three years? These questions were always possible. They were rarely asked, because the people who could ask them were busy reconciling.

Reach. Wealthy families are increasingly spread across countries, languages and legal systems. A single family may hold a Portuguese holding company, Spanish property, a UK investment account and a child studying in the United States. Technology lets one small team keep a single, current view across all of it, produce the same report in English, Portuguese and Spanish, and keep track of filing dates in each jurisdiction. Reach used to require a large office in each country. Much of it now requires a good data model and discipline.

Decision quality. The largest gain is the least visible. A family with one consolidated view decides in days what a family with four bank statements decides in months, or never. When the balance sheet is current and complete, a proposal can be tested before the meeting: what happens to liquidity if the company pays no dividend for two years, how concentrated the family is in one sector once the operating company is counted, what a property sale does to the tax position in each country, subject to the family’s tax advisors. The meeting is spent choosing, not discovering.

Why this lets a small firm outrun larger competitors

There is a structural point here. Much of the traditional wealth-advisory industry was built on headcount: more analysts to reconcile, more assistants to chase documents, more offices to cover more countries. Headcount is expensive and slow to change. Technology does not remove the need for judgement, but it removes much of the manual work that headcount was hired to do.

That changes who can compete. A small, independent firm that runs its work on good platforms and clean data can see a family’s whole balance sheet as quickly as a large institution, often more quickly, because it carries no legacy systems and no product to sell. That is not a claim about being better than anyone. It is a mechanism. The cost and speed advantage that used to come from size now comes from how the work is organised. For a founder, the lesson applies to the company as much as to the family: an organised small team with the right tools is not at a disadvantage against a large one that has not changed how it works.

How PWA works with technology, and where it stops

We describe here what we actually do, in plain verbs. PWA — Private Wealth Advisory designs and runs the operating layer around private wealth, and technology is how we keep that layer accurate and current.

We consolidate. We build one view of the family’s wealth across every custodian, entity and currency, on a reporting platform selected for that family. We have implemented Addepar, Maestro, Asseta and smaller specialist tools, and we have replaced expensive deployments with simpler ones when that made more sense. The choice depends on the complexity of the structure, the asset mix and how much the family wants to operate itself. Our page on consolidated reporting and back office sets out the full scope.

We map. We keep an entity map current: which company owns which asset, who signs, which jurisdiction governs, what is filed when. Alongside it sits a filings and renewals calendar, so that annual accounts, tax filings, KYC refreshes and capital calls are seen before they are due.

We organise documents. Statements, structure documents, contracts and correspondence go into an encrypted, permissioned document store with clear retention rules. Repetitive steps, such as filing incoming statements, refreshing KYC packs or preparing periodic reports, are automated where that is safe, so the family’s time and ours go to judgement, not paperwork.

We prepare decisions. When a decision is coming, the analysis is prepared before the meeting: the current position, the options, the scenarios and their effect on liquidity, concentration and the structure. Tooling prepares the decision memo. People decide. Our investment oversight work uses the same discipline to review managers, fees and risk across every bank.

We work in three languages. Our reports and memos are prepared in English, Portuguese and Spanish, so that every family member and every local advisor reads the same thing.

This is not new to us. Pedro Souto, who founded PWA, was an algorithmic trader and spent his career building financial technology inside institutions: proprietary trading systems at DV Trading, investment analytics across asset and wealth management at FRC Group, and the regulatory analytics infrastructure at Banco de Portugal. The habits are the same ones a trading desk teaches: test the model, check the data, assume that assumptions fail.

The family office technology stack: a layer, not a product

Founders often ask which software they should buy. The better question is what the family office technology stack needs to do. In our experience it has five parts: a reporting layer that consolidates custodians and entities; a document layer that holds the records; an identity and security layer that protects people, devices and access; a workflow layer that handles the recurring tasks; and an analysis layer that turns the data into decision memos. The products that fill each layer change. The layers do not.

Vendors in this market range from large platforms built for institutional family offices to lighter tools aimed at families with simpler structures. We name none here as a recommendation, because the right choice depends on the family. We select and implement platforms as an independent integrator. We do not sell or resell them, and we are paid by the family, and only by the family.

Where it stops

Technology runs inside rules we set before any tool touches a family’s data.

• People decide. No tool chooses an investment, a structure or a manager. Tools collect, reconcile, compare and draft. A person reads, challenges and signs off.
• The family holds the keys. Credentials and signing rights stay with the family. Where we need operational access, it is scoped, auditable, owned by the family and revocable at any time.
• Data stays private. Family data is handled under the GDPR and the family’s own access rules. We do not put confidential family information into tools that cannot guarantee how it is stored and used.
• We do not cross the regulated line. PWA does not manage money, hold mandates, take commissions or give regulated investment, tax or legal advice. Where a question needs a licensed manager, tax advisor or lawyer, we prepare the file and coordinate them.

A 90-day plan for founders worried about AI

Worry is useful only if it turns into a plan. Here is the one we would put in front of a founder who owns both an exposed company and a family balance sheet that nobody sees whole. It runs on three tracks in parallel and takes about ninety days.

Track one: the company

1. Run an exposure review by function. List the company’s functions and, for each, score how much of the work is language, documents, data or customer contact, how repetitive it is and how much cost it carries. Do the same for your three largest customers and suppliers: how will their use of AI change what they need from you?
2. Name one owner for AI. One person, reporting to the chief executive, with a budget and a mandate to run two or three measured pilots. Not the head of IT by default. PwC found only 14% of family businesses have such a person.
3. Start an AI inventory. The EU AI Act, Regulation (EU) 2024/1689, sets obligations according to how an AI system is used and what risk it carries. You cannot meet obligations you have not mapped. List every AI tool the company uses, including the ones employees adopted on their own, what data each touches and who is responsible for it. Your lawyers can then tell you what applies.
4. Put AI on the board agenda. One standing item a quarter: what was tried, what it saved or earned, what it risked. If no board member understands the technology, add one who does, or bring in an outside voice for that item.

Track two: the family

1. Build the consolidated view first. Before any analysis tool can help, the family needs one current, reconciled picture of everything it owns: every bank, every entity, every property and commitment. This is the step most families skip and the one that makes every later decision faster. Our note on the minimum viable family office describes how small this can start.
2. Set data and access rules. Decide who in the family sees what, where documents live, who holds which credentials and what happens when an advisor or employee leaves. Families are attacked at the household level, not through the bank.
3. Give the next generation a real role. The surveys show the next generation is more confident about AI than the board. Give one of them a defined project, such as the AI inventory in the company or the reporting design for the family, with a deadline and a report to the family council.

Track three: the advisors

1. Ask each advisor what they automate. Your banks, accountants, lawyers and managers are all adopting AI. Ask them, in writing, which parts of your work they now do with technology, how they protect your data while doing it, and whether the time saved shows up in what you pay.
2. Ask who sees the whole. Each advisor sees their part. Ask which of them is responsible for the consolidated picture. If the answer is nobody, that is the gap to close first.
3. Get an outside read. Before buying a platform or hiring a family office team, have someone independent look at how the wealth is organised today and what should change first. That is what a written second opinion is for.

At the end of ninety days the founder should have an exposure map of the company with two or three pilots running, a named owner and a first inventory for the AI Act, one consolidated view of the family’s wealth or a clear plan to build it, and written answers from every advisor. None of this needs a transformation programme. It needs ownership, data and a calendar.

Questions we hear

How will AI affect my business?

It depends on how much of your work is knowledge work. Customer service, marketing, software, finance and administration change first; production, construction and site work change later, mostly through planning and the back office. Deloitte found 86% of large family businesses already use AI, so the question is less whether to use it than where first.

Which industries will AI disrupt the most?

Knowledge-intensive services: software and IT, banking and insurance, professional services such as legal and accounting, marketing and media, and customer operations. McKinsey puts about 75% of generative AI’s value in four functions, and Eurostat counts 62.5% of EU information and communication firms already using AI, against 10.8% in construction.

What industries will not be affected by AI?

None will be untouched. Physical, site-based work such as construction, agriculture, hospitality and much of manufacturing is less exposed, because the core task is done by hands and machines on location. The change arrives through the office around that work: quoting, scheduling, purchasing, forecasting, compliance and the accounts.

How are family offices using AI?

Mostly for operations and reporting, still on a minority basis. Citi’s 2025 survey found 22% of family offices automating operational tasks or using AI for investment analysis, up from 13% a year earlier, and 16% using it for performance reporting. The largest barrier, cited by 57%, is the lack of internal expertise.

Does PWA use AI, and does it manage my money?

PWA uses technology to consolidate statements, keep entity maps and calendars current and prepare decision memos, and people make every judgement. It selects and implements family office platforms as an independent integrator and resells none. It does not manage money, hold mandates, take commissions or give regulated investment, tax or legal advice.

Where should a family business start with AI?

With one owner, one function and one clean data set. In the company, name a person accountable for AI and start where work is repetitive and measurable. In the family, build one consolidated view of the wealth first, because no tool can analyse statements that sit in four banks and a spreadsheet.

If the founder at the top of this piece sounds familiar, our case file on a founder with four banks and no single view shows what the fix looks like, and what a family office does once it is in place.

Sources

All pages opened and checked on 4 October 2026.

Family businesses and family offices

• Deloitte Private, “Deloitte Private global report reveals widespread AI adoption among family businesses but many still lag in digital readiness”, press release, 31 March 2026. Supports: 1,587 family businesses with revenue of USD 100M or more, 35 countries, fieldwork March–June 2025; 86% using AI (44% in many areas, 42% in select functions, 2% not using or exploring); uses: process efficiency 40%, risk management 39%, CRM 39%, customer experience 38%; productivity 68%, efficiency 67%, decision-making 65%; 48% investment falls short; 51% see inadequate adoption as moderate or high risk. deloitte.com
• PwC, “PwC 2025 Global Family Business Survey”, press release, 13 October 2025. Supports: 1,325 family businesses in 62 countries and territories, fieldwork 1 April–17 June 2025; 61% cite experimentation with AI as a growth opportunity; 3% looking to reinvent their business; 46% of public family businesses report GenAI boosted revenue and profitability (family-business data from PwC’s 28th Annual Global CEO Survey) against 29% and 32% of non-family CEOs. pwc.com
• PwC, “Global NextGen Survey 2024: Success and succession in an AI world”, 2024. Supports: 917 interviews in 63 territories (13 November 2023–23 January 2024); 73% of NextGen see generative AI as a powerful force for transformation; 49% of family businesses have prohibited or not started to explore AI; 7% have implemented it; 14% have a person or team responsible for generative AI. pwc.com
• Citi, “AI in the Family Office”, 11 May 2026, summarising the Citi Wealth 2025 Global Family Office Report. Supports: 22% automating operational tasks or using AI for investment analysis (13% in 2024); 16% using AI for investment performance reporting; 57% cite lack of internal expertise as the biggest barrier; data privacy as a condition of adoption. citigroup.com
• UBS, “UBS Global Family Office Report 2026”, press release, 28 May 2026. Supports: 307 family offices in more than 30 markets, average family net worth USD 2.7 billion; 65% already invested in AI across the value chain. ubs.com

Exposure and economy

• International Monetary Fund, Kristalina Georgieva, “AI Will Transform the Global Economy. Let’s Make Sure It Benefits Humanity”, IMF Blog, 14 January 2024. Supports: almost 40% of global employment exposed; about 60% in advanced economies, 40% in emerging markets, 26% in low-income countries; roughly half of exposed jobs may benefit. imf.org
• McKinsey & Company, “The economic potential of generative AI: The next productivity frontier”, 14 June 2023. Supports: USD 2.6–4.4 trillion a year across 63 use cases; about 75% of value in customer operations, marketing and sales, software engineering and R&D; banking, high tech and life sciences among the most affected industries; banking USD 200–340 billion; retail about USD 310 billion; up to 50% fewer human-serviced contacts; UK GDP of USD 3.1 trillion in 2021. mckinsey.com
• McKinsey Global Institute, “A new future of work: The race to deploy AI and raise skills in Europe and beyond”, 21 May 2024. Supports: about 27% of hours worked in Europe and 30% in the US could be automated by 2030; about 20% without generative AI; demand trends by occupation. mckinsey.com
• Goldman Sachs, “Generative AI could raise global GDP by 7%”, 5 April 2023, summarising Goldman Sachs Research, “The Potentially Large Effects of Artificial Intelligence on Economic Growth”. Supports: 7% (almost USD 7 trillion) GDP increase and 1.5 percentage points of productivity growth over ten years; 300 million full-time jobs exposed; two-thirds of US occupations exposed; a quarter to a half of workload in exposed occupations. goldmansachs.com
• OECD, “General assessment of the macroeconomic situation”, OECD Economic Outlook, Volume 2026 Issue 1, June 2026. Supports: no signs of widespread labour displacement from AI at the industry level; vacancies in the most AI-exposed industries rose more than in other sectors over the year to April 2026 in most economies, the US excepted; list of AI-exposed sectors. oecd.org

Adoption in Europe

• Eurostat, “20% of EU enterprises use AI technologies”, Eurostat news, 11 December 2025. Supports: 20.0% in 2025, 13.5% in 2024; Denmark 42.0%, Romania 5.2%. ec.europa.eu
• Eurostat, “Use of artificial intelligence in enterprises”, Statistics Explained, data extracted December 2025. Supports: large enterprises 55.0%, medium 30.4%, small 17.0%; information and communication 62.5%, professional, scientific and technical 40.4%, real estate 24.8%, construction 10.8%. ec.europa.eu
• Eurostat, database table isoc_eb_ai, “Artificial intelligence by size class of enterprise”, updated 15 June 2026. Supports: enterprises with 10+ employees using any AI technology in 2025: Spain 20.3%, Portugal 11.5%; Portugal large enterprises 49.2%, small 9.4%. ec.europa.eu

Regulation

• European Union, Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). eur-lex.europa.eu

The sector exposure table is PWA’s editorial reading of the sources above, not a forecast. This article is educational. It is not investment, tax or legal advice, and it is not a recommendation to buy any technology or to invest in AI.

---

Pedro Souto is the founder of PWA — Private Wealth Advisory. He works with UHNW founders, families, and single-family offices to build an independent operating layer around complex wealth — combining markets, analytics, technology, and governance to turn fragmented advice into one coherent picture, and to make someone, finally, responsible for the whole of it.