AI 202X: decade of major mirroring

While we as humans are living in unsecure times due to global climate change and geopolitical risks, the topic of artificial intelligence has definitively established itself as key revolution in 2024. We have moved beyond the hype of generative AI and are finally able to talk about the real meaning and value of AI. And with that, 2024 marks a turning point in AI evolution that calls for a revolution in the workplace, and forcing organizations to go back to the drawing board: are they still relevant in a couple of years from now? Can they continue running their business as they do today, without becoming irrelevant in their market and for their clients?

AI is like a mirror, reflecting the complex dilemmas and choices we face as a society, and how these decisions shape our future. It's not just pushing organizations to rethink their operations and relevance for the next decade, but also highlighting many interconnected issues. For example, Europe's competitive position is under significant pressure due to strict regulations on AI and data usage. These regulations, while aiming to protect privacy and ensure ethical standards, can also slow down innovation and make it harder for European companies to compete globally. This creates a challenging balance between maintaining ethical standards and staying competitive in a rapidly evolving technological landscape.

We also need to consider the data used for models like GPT-4, Claude, and Gemini. How ethically was this data labeled? Were there any copyright infringements? We don't know exactly what data was used to train these models, which we and our children increasingly rely on as ‘factual’ sources. Are we ready to accept the risks of algorithmic biases and push for faster innovation in Europe? Or should we be even more cautious and take the risk falling further behind competitively? As AI continues to evolve rapidly, we must keep up, both as individuals and organizations, balancing innovation and security.

AI is here, and is disrupting everything we do, see and encounter. On micro and macro level. And to use AI in the best possible form with the highest positive impact, we need to have a different perspective on AI and fully grasp the core essence… ?

So, do you know what AI means?

It’s all about Intelligence

AI is much more than just creating texts or images with ChatGPT, Copilot, or Midjourney... AI represents human intelligence in a new form, but scalable and digital. Algorithms trained on everything we know, think we know, and have gathered as data. This cognitive power allows us to analyze, understand, and transform data into new forms (numbers, text, video, photos, code, and more).

AI is that tireless colleague who works faster, is available 24/7, and can identify patterns that are difficult for humans to perceive. This colleague enables (and even compels) the rest of the company to become more human, personal, and social in their interactions with each other and with machines.

Reflection on today’s state of enterprise AI

After taking time during the holidays for reflecting on my last year and the developments with AI, the way we have approached it as topic within the organization and towards the market, my personal experience and vision on the impact of AI, I decided not to share the reflections on 2024, but most of all my perspective on where we stand today.

1 - Awareness and adoption of AI: increased understanding, but lack of real vision

Although there is still a lot to do in creating sufficient understanding about AI among many of us, we do have seen a large increase in awareness about the strengths of AI among most organizations – especially on leadership level. Where a year ago, mainly younger employees were eager to use AI and got an understanding of what was coming at us, we are now at the stage where most C-levels have a basic understanding of AI and its strengths, and accept that it is going to change their businesses fundamentally. Though, real understanding and vision is still lacking and too often the focus is on individual tooling. So the question of “AI, what is that?” has shifted to “AI: why and how?” This question drives several key steps and activities that need to be taken for organizations to drive the AI development, deployment and adoption towards value driving innovation. Main concerns that are being tackled in the current stage for many enterprise organizations:

• AI Strategy – What is our vision, mission and strategic approach for AI within our organization, specific functions and what business goals can be achieved by using different forms of AI?
• Capabilities – Which capabilities on AI should be in place and are we ready for taking AI really serious? Are we mature enough to cope with the required innovation and optimization?
• AI governance and continuous innovation – As the real value of AI should come from the business itself and not only defined by IT, how should the optimal process be designed in order to enable business driven innovation initiatives? And how does the technical development and IT-foundations interact with this process? Who decides on guardrails, investment and planning?
• AI in Control – And besides the technical and organizational monitoring and management of AI and innovation, how do we make sure that all model usage, development and deployment is secure and ethically responsible? And are we compliant to all different data and AI related regulations that are already in place or coming at us?

2 - Search for business case: defining the value of AI

“What value could AI bring us?” This is perhaps one of the most asked questions during the conversations and workshops on AI I have had over the last year. And as a real consultant would answer: it differs per organization and business goals. We see that 74% of global executives already experience increased productivity of their knowledge workers, improving their organization’s overall performance. But as said, for every organization the value can and will be different and depends on a large set of factors. In general, the following key value indicators are being experienced by organizations: efficiency, productivity, employee happiness, revenue growth, customer satisfaction… So, yes, all of the main business drivers are the core indicators of value of AI for organizations. This differs per organization, but is and should always be connected to the overall business strategy and goals.

Measuring the real impact of AI is therefore one of the most difficult tasks to execute, because of the many different aspects to measure. Not only direct results, like efficiency or revenue increases, should be involved, but also more indirectly and/or longer term related values like customer satisfaction and business model disruption. In early 2025, I expect to have a clear formula ready that enables us and organizations to predict and measure the expected value of every AI project. Do you want to stay in the loop or get involved in this AI-formula development, please reach out!

3 - Development of AI solutions - 2023: Trial, 2024: experimentation, 2025: scaling?

It appears that every enterprise organization is already trying to do something with AI, in a bigger or smaller manner. Research shows that 93% of enterprises have move beyond their early stage of AI journey and have at least started testing AI for their core business processes, albeit in limited forms. Where, as stated before, almost 74% of executives experience value of AI for their employees, only 31% is able to scale this throughout the organization. And this is where the biggest hurdle for organizations comes to the surface. Many factors are key for success to scale AI, like infrastructure, platform and business architecture, cross-functional collaboration (breaking the silos), governance and most important: AI adoption and training. And because of these many strategic components, enterprise organizations seems to be lacking behind. But step by step, enterprises are at the moment of preparing for scaling the potential of AI throughout the organization.

4 - Too(l) focused: finding problems for solutions – let’s turn this around

AI has been such a big topic over the last years, that many of us decided that we had to do something with AI. No matter what, we had to start using AI in our business. In addition, everything we see and read, is all about tools that generate texts, videos or images, or specific tasks. These individual tools have been pushing the awareness and development around AI, but still don’t bring the value on its own. And while most organizations have been searching for the right problems to add these solutions to, a change in thinking is required and, fortunately, happening. Because AI is able to solve a lot of challenges, the solution itself should never the be starting point.

5 - Rise of shadow IT – balance between secure and speed

When it comes to adopting AI, most enterprise organizations are understandably cautious. They show interest in implementing AI solutions, but these must be thoroughly tested and evaluated first. And with good reasons of course, due to several key factors, like potential risks as data privacy issues, algorithmic biases and security threats associated with AI. Furthermore, integrating AI into existing workflows and systems poses significant challenges. Enterprises need to ensure their infrastructure can support AI applications, which may require substantial investments in hardware, software, and employee training.

In the meantime, the workforce is eager to start using these AI tools that they have seen in the market and are already able to use in private life. Because most publicly available tools are currently more advanced than the secured, scalable and internal AI solutions (like your own CustomGPT platform), the risk of employees using their private tools for work is extremely high. This so-called shadow IT has become a large issue for CIO’s in many organizations, as they loose control on the systems being used and therewith the data that might be used, misused and potentially leaked. Enterprise organizations, really need to start offering the minimum required solutions to their employees in order to start controlling the use of AI and also to expand the potential benefits of it. Only when every employee is transparent in the use of AI, the impact of AI can be measured. Because you can’t measure what you don’t know… Start putting the tools in the hands of your people and you’re able to win back control on your IT-landscape.

6 - Data accessibility, quality and platforms – so much data, but still not sufficient

While exploring the possibilities and value of AI for enterprise organizations, eyes have been opened about the foundations required. This has led to conclusions that the current situation of many organizations are not orchestrated correctly and that there are many data-related topics not organized sufficiently. Over the last decade, organizations have collected immense amounts of data, but it seems not to be sufficient: data quality is too low (missing values or just outdated), accessibility is limited and availability of data is fragmented. The need for modernization of data is growing and urgency is rising to update platforms, to make different decisions regarding data management and governance, and foundational systems should be replaced or migrated to the Cloud. Furthermore, which data is relevant to use for what purposes? How to structure this and what about identity and access management? Besides the focus on structured data in combination with systems like Databricks, there is an increasing focus on the use of unstructured data for (generative) AI purposes. This brings important challenges for many organizations to leverage the knowledge and experience that they have in-house.

7 - Sectoral transitions – AI driven by urgency or position of advantage

Several sectors have made significant progress in their AI transformations, driven either by an urgent need to innovate and adapt or by a head start in digitization and data modernization. Healthcare emerges as a frontrunner, leveraging AI for advancements in diagnostics, personalized medicine, and operational efficiency. The main focus within healthcare is the accelerated modernization of data platforms and cloud migration, enabling a scaled strategic approach for AI development and adoption. Examples include AI predicting no-shows in hospitals and AI agents (Copilot) in ICUs for speech-to-text purposes, reducing administrative burdens and allowing more time for personal patient care.

The public sector is also showing substantial progress, with governments implementing AI to enhance public services, streamline administrative processes, and improve decision-making. This accelerated focus on AI is driven by labor shortages, increasing regulatory pressure, and administrative burdens within public organizations. For a broader perspective on the state of AI in public sector, please check out our recent publication – based on qualitative research conducted by Carlijn Hattink , Deborah Hofland and myself.

Thirdly, the financial sector is making significant progress in scaling AI solutions within organizations. The advantage these organizations already had in modernizing their data landscape, as well as in applying AI models for credit risk, due diligence, fraud detection, and more, enables them to adopt generative AI themes more quickly and scale new solutions on a stronger foundation. This primarily involves secondary, supportive processes such as finance, HR, and procurement.

Let's conclude on the current state of AI in enterprise

In conclusion, the journey of AI adoption in enterprises is marked by significant milestones and ongoing challenges. While awareness and understanding of AI have grown, there is still a need for a cohesive vision and strategy. Organizations must focus on defining clear AI strategies, building necessary capabilities, and ensuring robust governance to drive continuous innovation. The search for measurable business value from AI remains a priority, with enterprises exploring various metrics to measure its impact. As we move from experimentation to scaling AI solutions, the importance of infrastructure, cross-functional collaboration, and employee training and adoption cannot be overstated. Shifting from tool-centric to problem-solving approaches is essential for maximizing AI's potential. Addressing shadow IT and ensuring data quality and accessibility are foundational steps. Sectoral transitions show diverse AI applications, driven by urgency and strategic advantage.

As we reflect on the current state of enterprise AI, it is clear that while significant progress has been made, the journey is still just beginning. As Ethan Mollick states in his book Co-intelligence - Living and Working with AI: “Assume this is the worst AI you will ever use. We're early, early days still. I mean, there's a lot of stuff still being built”. This year, 2025, will bring us significant new areas of AI to focus on and while AI is growing exponentially, we are expected to keep up with the pace. So let’s dive into some of the key topics we should expect to change our business even more.

AI in 2025: Scalability and functional AI transformations

AI is no longer a nice-to-have, but becoming a must-have for organizations.

My perspective on what 2025 is about to bring us in terms of AI and key topics for enterprise organizations? Like all the ‘trend watchers’ dive into to have another hype, Agentic AI is one of the major topics for 2025. For me personally, this is nothing new and something that should be integrated in all the value-driven AI solutions that are developed and deployed in any function. Agentic AI workflows (mentioned by Andrew Ng , early 2024), are systems of multiple agents collaboration to perform a set of tasks within a specific function and should not to be confused with multi-model agents – which able to process and generate different forms of data like text, visual, video, speech, etc.. Let’s take Agentic AI serious, but not as separate hype and treat it as core requirement for automating business processes.

In AI Agent approaches, using suitable data sources for knowledge management and context is crucial. Instead of thorough training and finetuning specific language models, Retrieval Augmented Generation (RAG) has become a key focus over the last months. With the rise of Agentic AI systems, RAG will now evolve into Agentic RAG, linking various agents and retrieval methods in one system, enabling more accurate outcomes and decrease of hallucinations.

Furthermore, an increased focus on Small Language Models (SLM) is to be expected. The known chats like ChatGPT, Perplexity, and Gemini mainly use Large Language Models (LLMs) like GPT-4, o1, Claude 3.5 Sonnet, and others. But as these models know a lot about a lot, in many functional areas we require models that know a lot about certain functional areas and understanding of the context, jargon, and for example financial metrics. In these cases, SLMs are much more accurate as these are trained on specific data sets related to a topic or function. Well-known examples are Mixtral, Llama 3, Qwen and the by Microsoft recently announced SLM Phi-4, which excels at complex reasoning in math.

As these models are smaller, they can run faster and require lower costs (both in tokens and energy use). Because they are trained on domain-specific datasets and have a more accurate understanding of what you’re looking for, they experience lower hallucinations and can execute your specific tasks more accurately. These models don’t require large servers and could even run on our smartphones and smartwatches.

In line with these more technological AI-topics, I foresee the following AI focus areas to make the biggest difference for enterprise organizations in 2025:

a)????? Functional AI – from individual task augmentation to end-to-end AI workflows.

Where up till so far, the main focus of enterprise organizations has been on tools and augmentation of individual employees, the extended power of AI is set to transform entire functional areas that consume and require vast amounts of data. Main functional areas that are going to make the first transformational steps are, among others, finance and risk management. This transformation necessitates a shift in focus from simple tooling to comprehensive transformations, and includes the hyper-automation and even complete redesign of business processes. This functional transformation will be the key reason for organization to put additional efforts into the development and integration of Agentic AI systems and workflows, ensuring seamless and efficient operations. This so-called enterprise augmentation, is bound to make organizations more lean and efficient in work load.

b)????? Customer AI - focus from customer service to customer experience.

In recent years, for many organizations the first focus has predominantly been on customer service use cases, as it has been low-hanging fruit to optimize. However, this evolution will encompass a broader scope, integrating AI to provide personalized and seamless customer interactions across various touchpoints. And therewith, I expect AI to evolve from purely enhancing customer service to revolutionizing the entire customer experience. This requires integrated and streamlined end-to-end supply chain, procurement, delivery, sales and service departmental processes. More ecosystem thinking and collaboration between departments, breaking silos and enabling data enrichment and data sharing, without the need for copying data. Another key driver for improved customer experience is the strengthening of employee experience in offering AI solutions and optimizing their way of working.

c)????? AI Enterprise Search – setting the foundation for future AI solutions

Knowledge bases are becoming the cornerstone of AI enterprise search, forming the foundation required for effective information and knowledge management. These so-called knowledge search and assist platforms make information readily available, streamlining access and improving organizational efficiency. By developing this so-called Enterprise AI Knowledge Search & Assist platforms in a scalable, modular manner, organizations are able to use this foundation for future AI solutions and development, supporting individuals tasks, roles and business processes. The role of unstructured data is becoming more and more important in this practice, increasing focus on data management and governance, and shifting to information management and governance.

d)????? Data & AI – foundational challenges to overcome

As data has been an important topic over the last decades, we are at turning point to have a different perspective due to AI. Where our Big Data wave asked for collecting more and more data, the challenge now is that it is often not usable for AI for multiple reasons. Data is outdated, too low quality, fragmented, incomplete or just not accessible in legacy systems. This is putting organizations into a dilemma: should we redo our complete infrastructure or can we start with piloting smaller projects? I would say both, have two separate streams. One stream in which you reconsider your foundations and make important decisions (Cloud, on-premise, one lake, multiple ponds, etc.) that have the best fit for future and has the most secure and innovative solution in the meantime. In parallel, in the second stream you can start developing individual solutions, like the AI Knowledge Search & Assist, to support employees in specific functional areas. Just make sure that you align both streams and that the modular setup in stream two is able to migrate to your future fitting foundational solutions.

In addition, there is two data-related topics I want highlight that have a continued focus during 2025 and are enabling future acceleration of AI use case development:

• AI for data management and quality monitoring – where most people are thinking of data management for AI, we also see a shift towards usage of AI for data management, classification and especially quality monitoring. ?
• Unstructured data – where data management mostly focuses on optimizing structured data, AI has now given the value to unstructured data. This raises questions about management of this data, governance, priority and vectorizing.

e)????? Platforms and AI – acceleration of enterprise AI through embedded AI.

Related to the data topic, the state of data platforms and modernization of key operational systems (e.g. ERP, CRM) is key for growth with AI in 2025 and beyond. Besides Microsoft and its collaboration with OpenAI, all platform organizations (like SAP, ServiceNow, Salesforce, Bullhorn, etc.) that are using, processing and offering massive amounts of data are now accelerating their focus on embedded AI solutions. By tapping into strategic partnerships with the larger AI-service providers (like Amazon and Anthropic, Google DeepMind, Azure OpenAI), the platforms are embedding multiple AI solutions into the core of their current service offerings, making analysis faster, more personalized, accurate and even more predictive.

Over the last several years, many startups have gained ground in getting a piece of the AI market globally. But now the larger platform companies are putting all of their effort into bringing the best AI into their platforms as a single-experience, the market for the companies is becoming more and more difficult to grasp. Like the CEO of Microsoft, Satya Nadella recently stated: the SaaS business is dead. It is all about integrated platforms, and where AI agents replace standalone apps and SaaS platforms.

In 2025 I foresee a strong acceleration of usage of these embedded AI solutions, which brings multiple other challenges like business disruption, change of work and capabilities, and different cost structures. It is more than just activating AI-solutions. Let’s see how enterprise organizations are going to cope with these challenges.

a)????? Cyber AI – protecting against AI and leverage AI for protection.

At the moment, there is an acceleration in attention for cybersecurity in relation to AI. Both by the technological advancements, but also because of geopolitical situations globally. And this focus on cybersecurity and AI has a double-edged sword. The landscape will be dominated by an AI-driven arms race between attackers and defenders. Threat actors will leverage advanced AI to create more sophisticated and personalized attacks, including hyper-realistic phishing campaigns, AI-generated malware, and autonomous cyber weapons capable of adapting in real-time. To counter these threats, organizations will increasingly adopt AI-powered defense systems that can analyze vast amounts of data, detect anomalies, and respond to incidents at machine speed. These AI-driven security solutions will include real-time threat detection, automated incident response, and predictive analytics to anticipate and mitigate potential attacks before they materialize. However, the rapid deployment of AI tools also introduces new vulnerabilities, making AI systems themselves potential targets for breaches and manipulation. As a result, cybersecurity strategies in 2025 will need to focus on developing robust AI governance frameworks, implementing cross-border AI defense networks, and establishing clear accountability measures for AI-driven security decisions.

Key dilemmas mirroring our unbalanced state

Above key topics will have a large impact on enterprises, and the complete AI market in the Netherlands, during this year. We are going to see more and more breakthroughs and concrete, value-driven use cases are popping up more frequently. 2025 is going to be the year that foundations are ready and acceleration is started.

But there is several important dilemmas that remain and which we need to keep an eye on at enterprise organizations.

• AI Risk: Finding the right balance between risk and opportunity will remain crucial. As AI technologies advance, the potential for data breaches, ethical concerns, and job displacement will grow. Organizations will need to implement robust risk management frameworks, including regular audits, ethical guidelines, and comprehensive training programs, to mitigate these risks while maximizing AI's potential.
• Business Case value: The challenge of evaluating short-term gains versus long-term benefits will persist. While short-term projects may offer quick wins, they might not be sustainable. Long-term projects, on the other hand, require significant investment and patience. To address this, organizations will need to develop a balanced portfolio of AI initiatives, combining quick-win projects with strategic, long-term investments. This approach will ensure immediate value while building a foundation for future growth.
• Transparency in data and costs: Many organizations will continue to struggle with clearly articulating the data and cost implications related to their business cases. Transparency will be essential for gaining stakeholder trust and securing investment. Solutions will include adopting standardized reporting frameworks, leveraging advanced analytics to provide clear insights into data usage and costs, and maintaining open communication channels with stakeholders to ensure they understand the value and impact of AI initiatives.
• AI and ESG: Ensuring that innovation does not come at the expense of our planet will be a growing concern. AI can contribute to sustainability by optimizing resource use and reducing waste, but it can also have negative environmental impacts, such as increased energy consumption. Organizations will need to integrate ESG considerations into their AI strategies by adopting green AI practices, such as using energy-efficient algorithms, investing in renewable energy sources, and ensuring that AI applications align with broader sustainability goals.

So, AI is transitioning from a nice-to-have to a must-have for organizations. This year is set to bring significant advancements in AI. These advancements will drive concrete, value-driven use cases, making 2025 a pivotal year for AI integration and acceleration in enterprise organizations. However, challenges such as AI risk management, balancing short-term and long-term business case value, transparency in data and costs, and ensuring sustainability will remain crucial considerations.

Prepare yourself to stay relevant

There is no point anymore in looking away and ignoring the rise of AI. We are at the verge of the next revolution, changing our lives and organizations. Increase productivity and ensuring we earn more with working less. And getting the chance to become more social and personal again. Let’s take this opportunity to redefine our existence. The way we live and operate our organizations. It’s time to go back to the drawing board… back to the core of our why.

And for this: make sure you are ready, expecting the unexpected. Ready to stay relevant in a period where organizations have access to the same intelligence – where your people, creativity and excellence are your differentiator. It is about setting the right foundations, environment and controls to augment people and businesses to become the best version of themselves. It’s time for AI.

Note: This article is written from a personal perspective and is based on experiences in my role at KPMG, external sources and a vision of the market. I have used AI (Bing Copilot, KPMG AdvisoryGPT) for checking and optimizing wording and sentences.