As organisations rethink processes, introduce new technologies and adapt to increasingly complex business environments, how people experience change deserves to be part of the conversation.
Innovation is changing the way we work.
Artificial intelligence, automation and new digital tools are reshaping processes, roles and expectations across industries. Yet conversations about transformation often focus on what technology can achieve: greater efficiency, better decisions and new capabilities.
There is another side to that transformation. What does it mean for the people experiencing it?
The question matters because mental health at work is not separate from how work is designed, managed and transformed. As organisations look towards new ways of working, there is an opportunity to think about innovation more broadly: not only in terms of the technologies and processes we introduce, but also the working environments we create around them.
Work can have a positive role in our lives. It can provide purpose, structure, relationships and a sense of achievement. But the conditions in which work takes place matter.
The World Health Organization (WHO) identifies excessive workloads, low job control, job insecurity, discrimination and inequality, among other factors, as potential risks to mental health at work. It estimates that depression and anxiety result in around 12 billion working days being lost globally every year, at an estimated cost of US$1 trillion in lost productivity.
These figures illustrate the scale of the issue, but mental health at work should not be reduced to a productivity calculation.
Creating healthier working environments is first and foremost about people. At the same time, organisations cannot ignore the relationship between working conditions and their ability to operate sustainably over time. The WHO notes that safe and healthy working environments are more likely to support staff retention, performance and productivity.
Supporting mental health cannot rest solely on asking individuals to become more resilient or better at managing stress. The design and management of work itself need to be part of the conversation.
The WHO recommends organisational interventions that address working conditions directly, while the International Labour Organization (ILO) highlights measures including worker participation in decision-making, improved organisational communication and greater control over tasks as ways of addressing psychosocial risks.
For Irene Cervera, Senior Global Culture & Internal Communication, supporting mental health in the workplace requires a shared commitment:
“On a personal level, it means understanding ourselves, setting healthy boundaries and respecting those of others. At the same time, organisations need to create a safe environment where people feel able to express their needs without fear. Active listening from leaders is essential, as it helps identify barriers, build trust and, together, create healthier, more human and sustainable ways of working.”
This conversation becomes particularly relevant when we look at how quickly work itself is changing.
Organisations are adopting artificial intelligence, automating tasks, digitalising processes and introducing new ways of collaborating. Much of this transformation is designed to remove inefficiencies, improve decision-making and allow people to focus on higher-value activities.
But technology does not enter an organisation in isolation. It changes processes, responsibilities, interactions and, in some cases, the pace and structure of people’s work.
Research from the European Agency for Safety and Health at Work (EU-OSHA) illustrates this dual effect. Digital technologies can reduce repetitive tasks and create opportunities for people to focus on more creative and human-centred activities. At the same time, depending on how they are implemented, they can contribute to greater work intensity, reduced autonomy, isolation or an “always-on” culture.
The distinction is important.
Technology itself does not determine whether transformation will improve people’s experience of work. How organisations choose to design and implement that transformation matters.
EU-OSHA’s research points towards a human-centred approach that includes worker consultation and considers occupational health and safety from the design stage of digital transformation, rather than addressing its effects only afterwards.
This suggests that responsible innovation is not only a question of what a new technology can achieve. It is also a question of how that technology fits into the organisation around it.
When organisations talk about listening to employees, the conversation can easily become centred on surveys, engagement scores or periodic feedback exercises. But listening becomes more meaningful when it has the potential to influence how work is actually organised.
The people closest to a process often experience its friction first. They know where unnecessary complexity sits, where workloads become difficult to manage, where a new tool saves time and where it creates additional steps. Their experience does not replace business data, technical expertise or strategic priorities. It adds another source of information to them.
When organisations are redesigning a process or introducing a new technology, asking how people actually experience the change can therefore serve two purposes: it can help identify risks to the working environment, and it can reveal opportunities to improve the way work is organised.
This is particularly relevant as artificial intelligence becomes embedded in more areas of working life. The question is no longer simply whether organisations will adopt these technologies. Increasingly, it is how they will integrate them into work in a way that is useful, responsible and centred on people.
There is no single initiative that can create a mentally healthy workplace, just as there is no single technology that can make an organisation innovative.
Both require continuous attention to how an organisation works.
For businesses, this can mean looking beyond individual wellbeing programmes and considering workload, autonomy, communication, leadership, participation and job design as part of a broader approach to mental health. International guidance reflects this perspective: ISO 45003, for example, provides organisations with guidance specifically for managing psychosocial risks within occupational health and safety management systems.
For innovation leaders, it can mean looking beyond implementation and efficiency to understand how transformation changes people’s day-to-day experience of work.
And for organisations as a whole, it means recognising that people should not simply be expected to adapt to every transformation around them. They can also help shape it.
Perhaps that is where the conversations around mental health and innovation meet.
As businesses continue to rethink what work can look like, one of the most valuable questions may not simply be what technology can change, but what kind of working environment we want that change to create.
Because the future of work will not be defined by technology alone. It will also be defined by the choices organisations make about the people working alongside it.
Technology is often presented as a force that makes life easier. It automates processes, removes repetitive tasks, improves access to services and creates possibilities that would have seemed out of reach only a few years ago. Yet one question is still too often absent from the earliest stages of innovation:
An application can work exactly as specified and still be difficult to use for someone with low vision. A speech-recognition system can perform well in testing but struggle with particular accents. An automated recruitment tool can improve efficiency while reproducing historical patterns that disadvantage some candidates. A sophisticated device can still require movements, visual abilities or digital skills that many intended users do not have.
None of these outcomes requires a deliberate intention to exclude. Exclusion can emerge from the assumptions built into research, product requirements, training data and testing. This matters as artificial intelligence, automation and digital services move into increasingly sensitive areas of daily life, including employment, education, healthcare, banking, mobility, commerce and access to public services.
The World Health Organization estimates that 1.3 billion people, or around 16% of the global population, experience significant disability. Demographic change is widening the issue further: by 2030, one in six people worldwide will be aged 60 or over. These are not exceptional needs affecting a small niche. They describe a substantial and growing share of the people who use products and services.
Inclusive innovation responds to that reality. It is not about adding accessibility features at the end of development simply to satisfy a checklist. It means considering different abilities, contexts and ways of interacting with technology while the problem is still being defined and while design choices can still change.
This approach supports social inclusion and regulatory readiness, but its business value is equally important. Products that work for more people can reach wider markets, reveal unmet needs and perform more reliably in the conditions in which they will actually be used.
When a company discovers an accessibility problem after development is complete, correcting it may require extensive change. Teams may need to redesign interfaces, reorganise information, alter the technical architecture, modify physical controls or collect new data. The problem did not suddenly appear at launch; it was simply detected late.
This is why inclusive design begins with different questions:
These questions affect the direction of research and development. Many products exclude people because they were designed around an overly narrow idea of a ‘normal’ user. Once that imagined user becomes the standard, every difference is treated as an exception to be accommodated later. Inclusive innovation challenges that sequence. It treats human variation as a design condition from the beginning.
Product teams often create a representative persona and design around it. Personas can help teams focus, but they become risky when one profile is used as a proxy for everyone. People have different visual, auditory, motor and cognitive abilities. They use different devices, speak different languages and live in different cultural and economic contexts. Some have used digital tools for decades; others may be accessing a particular service for the first time.
Ability is not fixed throughout a person’s life. A limitation may be permanent, temporary or situational. Someone may be unable to hear a video in a noisy place, struggle to read a screen in direct sunlight or need to operate a device with one hand while completing another task. A person recovering from an injury may encounter the same interface barrier as someone with a long-term motor disability.
This helps explain why features developed for accessibility often benefit a much larger population. Captions are useful to Deaf and hard-of-hearing users, but also to people watching a video without sound. High-contrast displays improve access for people with low vision and remain useful in bright environments. Voice interfaces, predictive text and flexible input methods can help users across a wide range of situations.
The World Wide Web Consortium identifies this broader value in its business case for digital accessibility. Designing for varied ways of interacting can encourage innovation, improve user experience and extend market reach. Accessibility, viewed in this way, is not a final compliance test. It is a method for making products more flexible and resilient.
Artificial intelligence adds another layer to the issue. In a conventional product, an inaccessible design decision may create a visible barrier. In a data-driven system, exclusion can be harder to identify because it is embedded in patterns, classifications and probabilities.
AI systems learn from historical data, labelled examples and past decisions. If those inputs do not adequately represent the population on which the system will be used, performance can vary substantially between groups. This does not mean that every difference in model performance is automatically discriminatory. It does mean that representativeness, data quality and evaluation choices have material consequences.
The US National Institute of Standards and Technology has documented demographic differences in the performance of facial recognition algorithms. Its evaluations show that error rates can vary across demographic groups and that the composition and quality of image datasets influence measured performance. The same underlying concern extends well beyond facial recognition.
Consider a system developed primarily with data from one country and then deployed in another market. A language model may have limited exposure to certain languages, dialects or forms of expression. A health tool may be trained on a patient population that does not reflect the people who will eventually use it. A decision-support model may reproduce historical decisions made in a context where unequal treatment already existed.
Technology does not operate outside society, and data does not arrive without context. UNESCO warns that AI systems can embed or amplify bias and contribute to discrimination, inequality and digital exclusion when these risks are not addressed throughout the system lifecycle. The practical question for developers is therefore more specific than whether a model performs well overall: for whom does it perform well, and for whom does it perform worse?
A model can achieve a strong overall accuracy score while producing significantly different outcomes for particular groups. A single average may conceal exactly the information a company needs to understand before deployment.
Evaluation should therefore examine who is represented in the data, whether performance differs between groups, whether smaller populations have enough examples for meaningful testing and how labels were assigned. Teams also need to understand what happens when a model fails and whether users have a clear route to challenge or correct an outcome.
The consequences of error matter as much as the error rate. An inaccurate entertainment recommendation may have limited impact. A mistake in a system used to assess credit, screen job candidates, support a diagnosis or verify identity can affect access to essential opportunities and services. In those settings, inclusive development is closely connected to technology risk management, governance and product quality.
Data is not the only source of blind spots. The people who choose the research question, define success and test the product also shape what a technology can see.
No team can represent every possible user experience. Even so, teams that bring together different disciplines, backgrounds and lived experiences are more likely to question assumptions that might otherwise go unnoticed. This is especially important in R&D, where early decisions determine which problem will be studied, which hypotheses will be tested, who will participate in trials, which features will receive priority and which measures will define success.
Diversity in innovation should therefore be understood as more than a human-resources indicator. It can strengthen an organisation’s ability to identify needs. An engineer may see a technical limitation. A designer may notice a barrier in the user journey. A clinician may recognise that a solution does not fit real working conditions. A person who uses assistive technology may identify a problem that other testers never encounter. A commercial team may explain why a technically impressive feature is unlikely to be adopted.
These perspectives do not automatically produce a good product. They improve the quality of the questions, provided the organisation creates a process in which different views can influence decisions rather than simply being present in the room.
Inclusive innovation also changes the role of users in development. Teams cannot reliably infer every need from inside an office. People with disabilities, older adults and other groups that have historically been underrepresented in design processes should not appear only at the end to validate a nearly finished solution.
Their experience can inform initial research, problem definition, prototyping and testing. Co-creation and continuous user research help teams find barriers while the product remains relatively easy to change. The World Health Organization recommends involving people with disabilities and their representative organisations in decisions about the policies and services that affect them. The same principle is directly relevant to technology development.
Participation must also be meaningful. Inviting one person to a final review does not make a process inclusive. Organisations need to recruit a relevant range of users, compensate their contribution appropriately, provide accessible ways to participate and show how their feedback affected the product. When users are involved early, accessibility can become a source of insight. When they are consulted only after launch, it is more likely to become a corrective exercise.
The global demographic shift makes this work increasingly urgent. The World Health Organization projects that the population aged 60 and over will reach 1.4 billion in 2030 and 2.1 billion by 2050. Financial services, digital health, mobility, e-commerce, connected devices and communication tools will serve populations with a wider range of abilities and expectations.
Innovation for ageing populations does not necessarily require separate products labelled for older people. It may require existing products to become easier to understand and operate: reducing unnecessary steps, offering several ways to complete a task, avoiding interfaces that depend on precise gestures, allowing users to enlarge text, improving contrast and designing secure authentication that does not become an access barrier.
Language also matters. Clear instructions support users of every age, while oversimplified or patronising content can undermine trust and autonomy. An inclusive product recognises different needs without treating older users as incapable. This distinction affects adoption. A technically accessible product may still fail if the experience makes its intended users feel that it was designed for someone else.
Regulation is increasing the pressure on companies to act. Since June 2025, the European Accessibility Act has applied common accessibility requirements across the European Union to specified products and services, including computers, smartphones, banking and payment services, e-commerce, electronic communications and certain transport-related services.
Compliance is an immediate reason to review products, but it is a limited starting point for innovation. If an organisation considers accessibility only when a legal obligation appears, the project will tend to aim for the minimum requirement. If accessibility is treated as a product criterion, the business asks a more useful question: how can more people use this product successfully?
That question can reveal underserved markets, inspire functions that improve the mainstream experience and reduce the risk of expensive redesign. The W3C’s business case links accessibility with innovation, market reach, brand value and improved user experience, as well as reduced legal exposure. The strongest business case will vary by organisation, but the underlying logic is consistent. A product that excludes fewer people has more opportunities to create value.
Inclusion is easier to manage when it is part of the original specification. Adjusting contrast or navigation during early design may be straightforward. Reworking an established design system, architecture or physical product after deployment may be costly and slow.
Testing with diverse users requires time and budget. So does discovering after launch that part of the market cannot use the service, that customer-support demand has increased or that a product must be redesigned to meet a regulatory requirement. Early investment should be assessed against those downstream costs, not against a fictional process in which accessibility requires no work at all.
The same applies to AI. Reviewing data coverage, testing subgroup performance and establishing monitoring before deployment require resources. Correcting a model after it has been embedded in business decisions is usually more complex. Organisations may need to gather new data, retrain models, revise workflows, notify affected users and reassess previous outputs.
A final audit remains useful, but it cannot carry the entire responsibility for inclusion. Teams need checkpoints throughout research, design, development, procurement, testing, deployment and monitoring. Each checkpoint should have a clear owner, relevant evidence and authority to change the product when a material risk appears.
Organisations do not need to solve every dimension of inclusion at once before they can improve. They can begin by incorporating five questions into R&D and innovation reviews:
These questions do not replace technical standards, accessibility assessments or formal AI governance. They provide a practical way to change the conversations that shape innovation. They also create a record of assumptions that teams can revisit as the product, market and user population evolve.
For inclusion to influence product decisions, it needs to be built into the way innovation projects are governed. Organisations can start by defining accessibility and inclusion criteria in the original project brief, rather than treating them as a late-stage review point. Those criteria should be connected to the users, markets and operating environments the solution is expected to serve.
Practical measures include mapping user groups and usage contexts during discovery, involving a relevant range of users in research and testing, and checking prototypes across different devices, input methods, assistive technologies and real-world conditions. For AI and automated systems, teams should review whether the data adequately represent the intended population, evaluate performance across relevant groups and define what happens when the system produces an incorrect or contested outcome.
Inclusive innovation also requires clear ownership. Accessibility, data representativeness and user participation should be reviewed at defined checkpoints across research, design, development, procurement, testing, deployment and monitoring. Teams should document the assumptions they are making, the evidence used to support those assumptions and the product changes made in response to user feedback. This creates a practical record that can be revisited as the technology, market and user base evolve.
Accessibility has often been treated as a specialist review at the end of the process. A product was designed, developed and then checked to determine whether it was accessible. Inclusive innovation reverses that order. It considers human diversity while the problem is still being defined, reviews data before a model begins influencing decisions and involves intended users before the team assumes it understands how the technology will be used.
No product will respond in the same way to every possible need, and no single design can remove every barrier. That complexity is not a reason to postpone inclusion. It is a reason to treat it as a continuing part of research, development and product management.
For organisations, the opportunity extends beyond compliance.
The central question is therefore larger than whether a product meets an accessibility standard. Organisations should also ask how many innovation opportunities they are losing because they continue to design around a limited group of users.
At FI GROUP by EPSA, we understand innovation as the process of turning knowledge and technology into practical impact. That impact will remain limited if some of the people who should benefit from a solution cannot use it. Incorporating inclusion, accessibility and diverse perspectives from the first stages of R&D expands the range of problems a company can solve and the number of people its innovation can serve.
Organisations can begin by reviewing one current innovation project against the five questions above. The answers will show whether inclusion is already influencing decisions or whether it appears only at the point of final validation. That distinction can determine whether accessibility becomes a late correction or a source of better product decisions from the outset.
The European Commission has presented a new tax simplification package which, among other measures, proposes the creation of a common minimum framework for the tax treatment of R&D investments across the European Union.
The main objective is to reduce differences between Member States and encourage greater investment in innovation. For companies developing R&D&I projects, the proposal introduces a new reference point for planning future investment in assets such as machinery, laboratories and pilot plants.
It also highlights the growing connection between tax incentives and European innovation funding policy.
The proposal will now be submitted to the European Parliament for consultation and to the Council of the European Union for adoption before it can enter into force.
Until now, the tax treatment of R&D investment has depended on the rules of each Member State. This has created differences between countries and, in some cases, influenced decisions about where companies choose to develop innovation activities.
The Commission’s proposal introduces a common minimum standard with three main objectives:
Under this approach, taxation becomes an additional policy instrument to encourage R&D investment, alongside existing funding programmes and public support measures.
The European Commission expects the new framework to help strengthen and attract greater investment in R&D activities, as well as support the development of industrial and technological capabilities. Among the expected effects is an estimated annual increase of approximately 0.2% in European GDP.
The Commission also estimates a reduction of almost EUR8 billion in tax compliance costs for businesses.
FI Group by EPSA has actively contributed to this European debate by presenting proposals to reinforce the role of taxation as a lever for R&D&I investment.
“In 2023, we presented a position paper on BEFIT, in which FI Group by EPSA defended the need to include specific tax incentives for innovation within the future common European framework, recovering elements already considered in the CCCTB proposal, such as deductions for R&D investment and specific measures for start-ups,” explains Victor Cruz, Regulatory Public Affairs & Strategy Director.
He adds: “FI Group by EPSA put forward recommendations aimed at improving legal certainty, supporting stable and predictable budgets, promoting eco-innovation and moving towards a common definition of innovation in the European Union.”
In addition to the direct tax treatment of R&D investments, the European proposal includes measures that may influence how companies structure innovation projects at European level.
The proposal includes the removal of withholding taxes on dividends, interest and royalties between companies in different Member States.
This change would make it easier to organise corporate structures with activity in several countries, particularly in areas such as:
The Commission also proposes adjustments to European tax rules that may affect the financing of innovative projects.
These include:
The creation of a common framework may influence how companies plan innovation investment, particularly where the acquisition of equipment, machinery or infrastructure represents a significant part of the R&D budget.
The proposal may be especially relevant for organisations developing new products, processes or technologies that require specific equipment to carry out these activities, including:
Another important point is that the proposal would allow Member States to maintain more favourable national regimes, provided they are aligned with the framework ultimately approved by the European Union. This creates the possibility that future European standards will coexist with existing incentives in each country.
The proposal focuses on tangible investments required to develop R&D activities, including:
The proposal provides for a 100% deduction of the investment, with two possible approaches:
To access the incentive, the assets must also be used continuously for R&D activities for a minimum period of three years. If this requirement is no longer met, the proposal includes a mechanism to regularise the tax benefit obtained.
The content of the proposal makes it possible to anticipate certain trends that can be incorporated into R&D project planning.
In this context, companies should consider reviewing:
The European framework is moving towards greater integration between taxation, funding and innovation. In this context, planning an R&D project goes beyond identifying a specific incentive. It requires companies to analyse how different instruments can be combined and to adapt their funding strategy as the regulatory environment evolves.
Specialist advice from the early stages of a project can help companies identify funding opportunities, define an appropriate strategy and receive support throughout the project lifecycle. This enables businesses to incorporate regulatory developments into their planning and take advantage of available opportunities at both national and European level.
AI innovation hubs are becoming critical to global R&D strategy. Locations such as King’s Cross, the San Francisco Bay Area, Paris, Toronto, Montreal and Singapore offer more than access to talent. They bring together research institutions, public funding, tax incentives, investors, infrastructure and regulatory environments that can directly influence the cost, speed and commercial success of AI development.
For businesses investing in AI, location is no longer a purely operational decision. It is a strategic R&D choice that can shape access to funding, partnerships, specialist expertise and long-term innovation capacity.
AI innovation is often discussed as though it happens entirely in the cloud. In practice, much of it still depends on place.
The strongest AI ecosystems tend to emerge where talent, research, capital, infrastructure and public policy are concentrated. This is why locations such as King’s Cross, the San Francisco Bay Area, Paris, Toronto, Montreal and Singapore matter to businesses planning AI-led R&D. They are not simply office locations. They are operating environments.
For companies investing in AI, the decision on where to build, test and scale can influence access to people, funding, partners, tax relief and regulatory certainty. The technology matters, but so does the ecosystem surrounding it.
AI R&D is resource-intensive. It requires specialist engineers, data scientists, domain experts, compute infrastructure, research links, commercial partners and patient capital. These inputs are rarely distributed evenly across a country. They concentrate.
Clusters form because proximity reduces friction. Universities train talent and generate research. Start-ups turn ideas into products. Larger companies provide scale, procurement routes and career pathways. Investors bring capital and market discipline. Public bodies provide grants, infrastructure and incentives. When these groups are close enough to interact regularly, collaboration becomes easier.
This is particularly important in AI because many projects sit between disciplines. A healthcare AI company may need machine learning researchers, clinicians, regulatory specialists, data governance support and commercial partners. A robotics company may need AI engineers, hardware expertise, manufacturing partners and testing environments. A financial services AI business may need technical talent, compliance expertise and access to enterprise clients.
Clusters help companies assemble these capabilities faster.
There is also a talent effect. Skilled workers are more likely to move to places where there are multiple employers, strong research institutions and credible career options. Companies benefit from deeper labour markets. Employees benefit from choice. This creates a reinforcing cycle: talent attracts companies, companies attract capital, and capital attracts more talent.
Venture capital adds another layer. According to OECD analysis, AI firms accounted for 61% of global venture capital investment in 2025, worth USD 258.7 billion. That capital is highly concentrated. Businesses outside the main AI hubs can still raise funding, but clusters improve access to investors who understand the technology, the commercial risks and the route to scale.
Knowledge spillovers are harder to measure, but equally important. They occur through conferences, shared labs, alumni networks, founder communities, investor meetings, university partnerships and informal contact. In a fast-moving field, proximity to the right conversations can be strategically valuable.
King’s Cross has become one of Europe’s most visible examples of an urban AI and science cluster.
Its strength comes from the wider Knowledge Quarter, which covers King’s Cross, St Pancras, Euston and Bloomsbury. Within a small area sit the British Library, the Alan Turing Institute, the Francis Crick Institute, UCL, Google DeepMind and a growing group of technology and life sciences companies.
This mix is important. King’s Cross is not only a technology district. It connects AI with biomedical research, data science, public institutions, transport infrastructure, cultural organisations and commercial space. That gives it a broader innovation base than a pure software cluster.
Google’s Platform 37 development adds further weight. The building is expected to bring Google and Google DeepMind teams together in King’s Cross, alongside the AI Exchange, a public-facing space designed to improve understanding of AI. This matters because anchor organisations shape ecosystems. They attract talent, suppliers, investors and adjacent firms.
The UK Government’s 2024 AI Sector Study estimated that the UK had 5,862 AI companies, £23.9 billion in AI-related revenue and more than 86,000 AI-related jobs. London, the South East and the East of England accounted for around three quarters of registered office locations for UK AI firms. King’s Cross sits inside that concentration.
The district also benefits from its life sciences base. UCL analysis has suggested that the Euston and King’s Cross innovation district already generates more than £8 billion in annual GVA and could add up to £3.5 billion more by 2035 with the right planning, transport and investment conditions. The Francis Crick Institute’s partnerships to expand commercial lab space in the area show how physical infrastructure is being built around scientific demand.
This is the point businesses should recognise. King’s Cross has grown because several conditions overlap: research density, corporate commitment, transport links, public institutions, commercial space and access to finance. It is not a single-factor success story.
King’s Cross is part of a wider global pattern. AI hubs are developing in different ways, shaped by their talent base, policy environment and funding model.
The Bay Area remains the commercial benchmark for AI. It combines world-class universities, large technology companies, venture capital, experienced founders and a deep technical workforce. Its advantage is scale. Companies can access capital, talent, infrastructure providers and customers in one market.
The cost base is high, and competition for talent is intense, but for frontier AI companies the Bay Area remains difficult to ignore.
Toronto and the wider Ontario ecosystem offer a different model. The Vector Institute, the University of Toronto and the Toronto-Waterloo corridor have helped build one of North America’s strongest AI talent bases. Ontario also benefits from Canada’s SR&ED tax incentive programme and provincial support such as the Ontario Innovation Tax Credit.
For companies looking for North American AI capability outside the Bay Area, Ontario offers strong research depth and a structured incentive environment.
Montreal’s AI position is closely linked to Mila, the Quebec AI Institute, and its academic network across Université de Montréal, McGill, Polytechnique Montréal and HEC Montréal. The city is especially strong in deep learning research and talent development.
Quebec has also strengthened its innovation support through the CRIC tax credit for R&D, innovation and pre-commercialisation activity. That gives Montreal a clear proposition: research strength, public support and a lower-cost environment than many US hubs.
Paris has become a stronger European AI contender through a mix of state strategy, academic depth and start-up infrastructure. Station F has increased its AI activity, while France 2030 has backed nine AI clusters with €360 million to strengthen research, training and innovation.
France also has one of Europe’s most established R&D tax incentive systems through the Crédit d’Impôt Recherche. For companies assessing European AI locations, Paris combines market access, talent and public support.
Singapore’s strength lies in coordination. Its government has treated AI as part of a broader national innovation strategy, linking talent, regulation, digital infrastructure and regional market access. The National AI Strategy and Enterprise Innovation Scheme give businesses a clear policy framework.
Singapore is especially relevant for companies looking at Asia-Pacific expansion. It offers connectivity, a pro-business environment and a growing base of AI partnerships across professional services, finance, healthcare and advanced industries.
AI hubs do not grow through market forces alone. Public policy often shapes whether a cluster becomes durable.
Governments use incentives for a practical reason: R&D creates benefits that extend beyond the company funding it. New knowledge, skilled workers and technical capability can spread through supply chains and wider industries. Without support, businesses may underinvest.
R&D tax incentives help reduce the cost of eligible activity. In the UK, the merged RDEC scheme applies from 1 April 2024 at a 20% credit rate, while Enhanced R&D Intensive Support provides additional support for qualifying loss-making R&D-intensive SMEs. In France, the Research Tax Credit supports qualifying R&D expenditure. Canada’s SR&ED programme remains a major federal incentive, often layered with provincial schemes. Singapore’s Enterprise Innovation Scheme offers enhanced deductions or allowances for qualifying innovation activity.
These incentives affect the real cost of experimentation. For AI businesses, this can include work on model development, data processing methods, software architecture, automation, testing, technical uncertainty and integration challenges, provided the activity meets the relevant scheme rules.
Grants play a different role. They can help companies start projects that would otherwise be delayed or reduced in scope. UK programmes such as BridgeAI support AI adoption and development in priority sectors. EU AI Factories are designed to improve access to compute, data and expertise. France 2030 is using public investment to strengthen national AI capability. Quebec’s support for Mila shows how direct funding can reinforce research institutions and talent pipelines.
Public-private partnerships are also central. Innovation hubs need convenors: institutions that bring companies, researchers, funders, standards bodies and public agencies together. The Alan Turing Institute, Digital Catapult, Innovate UK, the Francis Crick Institute, Station F and Singapore’s EDB all play different versions of this role.
For businesses, the lesson is clear. Incentives should not be treated as an afterthought once a location has been chosen. They are part of the location decision.
AI location strategy should start with the work being undertaken. A company developing frontier models has different needs from a manufacturer applying AI to production processes, or a healthcare company building regulated decision-support tools.
Five factors deserve close attention.
AI hubs are becoming strategic engines of R&D because they bring together the conditions that help companies move from technical possibility to commercial use. Talent, capital, infrastructure, research institutions and incentives all matter. Their value increases when they work together.
King’s Cross shows how this can happen in a dense urban district. Its advantage is not only the presence of major AI companies. It is the combination of AI, life sciences, public research, transport, universities and commercial investment in a small geography.
Other hubs offer different strengths. The Bay Area has unmatched capital depth. Ontario and Montreal offer research strength and structured public support. Paris benefits from national AI investment and a strong R&D tax regime. Singapore offers regional access and coordinated policy.
For businesses investing in AI, the location decision should be treated as part of the R&D strategy. The right hub can improve access to talent, shorten collaboration cycles, reduce project costs and open routes to funding. The wrong location can make the same project slower, more expensive and harder to scale.
AI may be digital, but innovation still has an address.
Artificial intelligence has moved beyond experimentation. It is now a core driver of competitiveness, efficiency and strategic differentiation across industries. As adoption accelerates, however, a parallel force is reshaping its development: regulation.
Governments worldwide are no longer taking a passive approach to AI innovation. They are actively defining the conditions under which it must evolve. The European Union’s AI Act, alongside emerging frameworks in the United States, Asia and Latin America, signals a structural shift. AI is no longer simply a technological opportunity. It is now a regulated domain.
This creates a fundamental question for organisations: does regulation slow innovation, or does it redefine how innovation happens?
The answer is increasingly clear. Regulation is not a constraint on innovation. It is a filter. It determines which innovations are viable, fundable and scalable.
For businesses, the question is no longer whether to comply. It is how to use compliance to unlock investment, accelerate adoption and build long-term competitive advantage.
The first wave of AI adoption was defined by experimentation. Companies tested models, explored use cases and prioritised speed over structure. Governance, explainability and risk management were often secondary considerations.
That phase is over.
We are now entering the era of trusted AI, where systems must demonstrate not only performance, but also accountability, transparency and control. This shift is being driven by regulators, but also by market expectations. Clients, investors and partners increasingly require assurance around how AI systems operate, use data and make decisions.
In practice, organisations must redesign their AI roadmaps around three core pillars:
As internal governance frameworks mature, companies are formalising processes that include risk classification, human oversight and traceability. This aligns with emerging standards such as ISO/IEC 42001, where AI systems are managed through structured, auditable processes across their lifecycle.
The implication is direct. AI is no longer a standalone capability. It is an integrated, regulated system that must be designed for trust from day one.
One of the most underestimated effects of AI regulation is its direct impact on funding.
Public funding programmes, particularly in Europe, are increasingly aligned with regulatory compliance. Funding is no longer awarded solely on the basis of technological ambition or innovation potential. It is increasingly linked to a project’s ability to demonstrate ethical use, security and responsible governance.
This is especially visible in the European ecosystem, where the AI Act sets a clear framework for acceptable practices. Projects that incorporate risk management, data governance and human oversight are better positioned to access funding instruments, including R&D incentives and strategic innovation programmes.
At the same time, governments are prioritising investment in sovereign and strategic technologies. AI sits at the centre of this agenda, but only when it aligns with broader objectives: resilience, security and control over critical data and infrastructure.
For companies, this creates a structural shift:
Organisations that treat compliance as a checkbox will fall behind. Those that integrate it into their innovation strategy will be better positioned to access capital that others cannot.
A second major shift is taking place in how AI systems are designed.
Historically, compliance was reactive. Systems were developed first, and regulatory requirements were addressed later, often at significant cost. Today, this approach is no longer viable.
AI systems are increasingly being built to be compliant by design. This means that requirements relating to data protection, cybersecurity, traceability and governance are integrated from the earliest stages of development.
This shift has several consequences:
For industrial AI projects, the impact is direct. Digital transformation programmes must now include governance layers, documentation standards and risk controls as core components, rather than add-ons.
Companies that adopt this approach achieve two key advantages. First, they reduce the risk of costly redesigns or compliance failures. Second, they accelerate adoption by building trust with stakeholders from the outset.
A common pattern is emerging in successful projects: when governance and compliance are embedded early, organisations move faster in later stages. They face fewer barriers in deployment, scaling and market acceptance.
While regulation is becoming universal, its implementation varies significantly across regions. These differences have direct implications for companies operating internationally.
Europe leads with a structured and comprehensive approach. The AI Act introduces a risk-based framework that defines clear obligations, particularly for high-risk systems. The European model prioritises trust, safety and fundamental rights. While this creates higher compliance requirements, it also establishes a strong foundation for scalable and trusted AI.
The United States adopts a more flexible, market-driven approach. Regulation exists, but it is less centralised. Innovation tends to lead, with governance frameworks evolving in response. This creates a more dynamic ecosystem, but also greater uncertainty over the long term.
Asia, particularly China, combines strong state involvement with rapid technological deployment. Regulation is present, but often aligned with national strategic objectives. The focus is on scaling AI capabilities quickly while maintaining control over key sectors.
Latin America is at an earlier stage. Adoption is accelerating, but regulatory frameworks are still developing. This creates opportunities for rapid implementation, but also risks linked to future alignment with global standards.
For businesses, the implication is clear: AI strategies must be designed for regulatory diversity. Solutions that are compliant in one region may not scale globally without adaptation.
Despite the visibility of regulation, many organisations are still underestimating its business impact.
The first misconception relates to cost. Compliance is often seen as an expense that reduces margins. In reality, its absence can be far more expensive. Retrofitting compliance into existing systems, facing regulatory penalties or rebuilding non-compliant solutions can significantly increase costs.
The second blind spot is scalability. AI systems that are not designed with regulatory requirements in mind face major barriers when expanding into new markets. What works locally may not be transferable internationally.
The third and most critical risk is obsolescence. AI solutions developed today may become non-compliant tomorrow if they do not anticipate regulatory evolution. This can create stranded investments and limit long-term value creation.
Organisations that fail to address these challenges early face a compounding disadvantage. They risk losing access to funding, facing operational constraints and struggling to scale.
By contrast, companies that integrate governance frameworks, including structured AI management systems, mitigate these risks. They create a foundation where innovation can evolve within clear, sustainable boundaries.
AI regulation marks a turning point in how innovation is conceived, developed and scaled.
It is not simply a set of constraints imposed on technology. It is a framework that defines what sustainable, scalable and investable innovation looks like.
For businesses, the implication is strategic. Compliance is no longer a legal requirement operating in isolation. It is a core lever for funding, differentiation and growth.
Organisations that embed governance, transparency and accountability into their AI systems will not only reduce risk. They will position themselves to capture new opportunities, access strategic funding and build trust in increasingly competitive markets.
The shift is already underway.
The question is not whether regulation will shape AI innovation. It is whether companies are ready to use it to their advantage.
On 8-9 June 2026, FI Group by EPSA participated in the European Association of Innovation Consultants (EAIC) Annual General Meeting in Zagreb, Croatia.
EAIC is a European association representing innovation consultancy firms and supporting dialogue between consultants, institutions and industry on research, development and innovation funding policy.
The event brought together key stakeholders from EU institutions, industry and consultancy to discuss the future of innovation funding in Europe, with a particular focus on the next European long-term budget for research and innovation, the Multiannual Financial Framework (MFF) 2028-2034.
These discussions are highly relevant for organisations seeking to anticipate how European innovation funding will evolve in the coming years.
FI Group by EPSA’s participation in the EAIC AGM reinforces its position as an active contributor to the development of European innovation policy and funding frameworks.
By being present where key decisions are discussed and shaped, FI Group by EPSA gains early visibility of policy and programme developments while contributing to the design of future funding mechanisms. This enables the company to create tangible value for clients by:
This involvement supports FI Group by EPSA’s ability to provide clients with forward-looking insight, stronger strategic guidance and a clearer understanding of future EU funding programme opportunities.
During the EAIC AGM, Luca Pira, Public Affairs and International Grant Manager and Coordinator of the National Funding Schemes Working Group, played a central role in both implementation and strategic discussions as part of the Working Group Coordinators panel.
In this context, he presented FI Group by EPSA’s contributions to key topics, including the future design of EU innovation funding at national and regional level under the MFF 2028-2034, as well as regulatory and State aid frameworks.
This included input into the recent public consultation on the General Block Exemption Regulation, where FI Group by EPSA contributed as part of a coordinated effort at both national and European level alongside key industry associations: the European Association of Innovation Consultants (EAIC), the Asociación Española para el Fomento de la Financiación de la I+D+i (AFIDI), the Associazione Italiana per la Ricerca Industriale (AIRI), and BusinessEurope.
Luca also highlighted upcoming work on a European comparative report on R&D and technological innovation incentives, coordinated by Víctor Cruz, Regulatory, Public Affairs & Strategy Director. This further demonstrates FI Group by EPSA’s expertise in tax incentives and regulatory intelligence.
FI Group by EPSA’s active participation in the EAIC AGM also strengthens its international partner network, particularly in countries and regions where the company does not yet have a direct presence, such as Croatia and the broader Balkan region.
These markets are becoming increasingly strategic within the European innovation landscape. Through EAIC, FI Group by EPSA is also able to leverage collective advocacy capacity at European level, enhancing its influence on policy and funding frameworks beyond what a single organisation could achieve independently.
This approach creates stronger opportunities for FI Group by EPSA and its clients in international and cross-border innovation initiatives.
Active engagement across EAIC Working Groups
FI Group by EPSA is actively engaged in several EAIC Working Groups, covering:
This involvement strengthens the company’s expertise, expands its network and enhances its ability to anticipate and shape market trends.
As Europe defines its next innovation funding cycle, FI Group by EPSA will continue to contribute actively within EAIC, remaining at the forefront of policy and programme developments. This includes participation in the EAIC Summit on the Future of Funding in Europe, taking place on 3 November 2026 in Brussels. The event is expected to feature speakers such as MEP Christian Ehler, a key figure in EU research and innovation policy, who plays a leading role in shaping the MFF 2028–2034 and the upcoming FP10.
This engagement is aligned with a broader shift in the sector: from a primarily operational role focused on identifying opportunities and preparing applications, towards a more strategic advisory role. This requires the ability to combine technical, financial and regulatory expertise to position projects, structure partnerships and support clients across the full innovation lifecycle.
By leveraging EAIC’s collective advocacy capacity at European level, FI Group by EPSA strengthens its ability to influence frameworks, anticipate change and create tangible value for clients, while reinforcing its competitive position in the European innovation funding landscape. FI Group by EPSA can support organisations in identifying upcoming opportunities, interpreting regulatory changes and positioning projects for future funding success.
