AI use is increasing dramatically
- AI is becoming cheaper and more accessible
This is what it means for you as a decision-maker!
The Stanford Institute for Human-Centered Artificial Intelligence (HAI) publishes the annual "AI Index", a comprehensive report on the state of artificial intelligence. The latest State of AI Index 2025 is now the eighth edition and offers over 400 pages of data and analyses on the rapid development of AI over the past year.
The report shows that AI is a technology-driven transformation of historic proportions. "AI is a civilisation-changing technology - not limited to one sector, but transforming every industry," emphasises Russell Wald from Stanford HAI.
The report provides an empirical basis for corporate decision-makers to assess the significance of these developments and make informed strategic decisions.
The aim is to provide a factual and analytical overview that gives decision-makers the relevant facts as to why action is needed now in order to remain competitive in the face of technological change.
One of the clearest signals from the AI Index 2025 is the rapidly growing spread of AI in more and more areas of life and business. AI is leaving the laboratory experiments and becoming part of everyday life.
Companies are also increasingly relying on AI. The proportion of organisations using AI jumped to 78% in 2024 - up from 55% in the previous year. In other words: Within one year, the use of AI in companies has expanded from just over half to over three quarters. This broad acceptance underlines the fact that AI is no longer just the special project of a few pioneers, but has arrived in the mainstream of the business world. Companies from almost all sectors - from industry to retail and finance - are experimenting with AI or already have productive applications in use. AI-supported tools and processes can be found in areas such as customer communication (chatbots), data analysis, production (predictive maintenance) and the back office (automation of repetitive tasks).
At the same time, the public perception of AI is also taking on interesting characteristics. According to the AI Index 2025, large majorities in some countries such as China (83%), Indonesia (80%) and Thailand (77%) consider AI products to be more useful than harmful. In Western countries, the population is traditionally more sceptical - in the USA and Canada, for example, only around 39-40% are predominantly optimistic. But even here, the mood is shifting: since 2022, optimism towards AI has risen by +10% in Germany, for example. For companies, this means that acceptance of AI applications is growing overall, especially in emerging markets, while in Europe and North America trust needs to be built up through transparent and human-centred AI applications.
To summarise this trend: AI is on the rise everywhere. Artificial intelligence solutions are making the leap from pilot projects to operations and the end user. Business decision-makers should realise that competitors and customers are already coming into close contact with AI. Those who are still hesitant to utilise AI in their own company run the risk of falling behind. The increased spread also shows that there are enough success stories - AI is tried and tested and has delivered real added value, otherwise 78% of companies would not be using it.
Hand in hand with widespread adoption is an unprecedented surge in investment in AI technologies. Companies and investors are investing more money in AI than ever before. According to the AI Index, private AI investment in the USA rose to USD 109.1 billion in 2024 - almost twelve times as much as in China (USD 9.3 billion) and 24 times as much as in the UK (USD 4.5 billion). This enormous sum illustrates the strategic priority that AI has for American companies and investors. By way of comparison, such orders of magnitude were previously only seen in established sectors such as IT or energy. AI is now reaching similarly high levels of investment, signalling that market-defining breakthroughs are expected.
The boom in generative AI (i.e. AI that can generate text, images or other content, for example) is particularly noteworthy. This sector attracted USD 33.9 billion in private investment worldwide in 2024 - an increase of 18.7% compared to 2023. Generative AI models such as large language models (e.g. GPT systems) were very much in the media spotlight in 2023, and in 2024 the headlines were followed by real financing rounds and projects. For companies, this means Generative AI is evolving from a trend to an established business technology. Many companies are investing in the development of their own generative models or in the integration of such AI (e.g. for personalised customer contact, automated reporting, marketing content or product design).
Why is so much capital flowing into AI? One reason is the demonstrable productivity gains. The Stanford report refers to a growing number of studies that prove that AI can increase employee output and reduce skills gaps. gaps. With AI support, for example, less experienced employees can achieve results that previously required expert knowledge - this is referred to as "skill augmentation", i.e. the expansion of the workforce's capabilities through AI tools. AI also automates routine tasks so that skilled workers can concentrate on higher-value activities. These increases in productivity and efficiency ultimately have a positive impact on key business figures, which justifies the high level of investment.
Another aspect is the competitive pressure: as many market players are investing massively in AI, a dynamic is emerging in which no larger company wants to fall behind. AI expertise is becoming a competitive factor, similar to internet expertise 20 years ago. Whoever sets the course for AI today could set the pace for the industry tomorrow.
For decision-makers, one thing is clear: The record investments signal confidence in the economic importance of AI. There is a broad consensus that AI applications harbour significant business potential - be it through new sales, cost savings or innovation leadership. Companies should therefore examine how they can benefit from these waves of investment, for example through partnerships, investments or internal innovation programmes. At the same time, companies should keep a close eye on the development of generative AI, as this is where much of the momentum lies. In terms of business strategy, it is advisable to have a clear AI roadmap in order to build up talent, data and technologies in good time before the competition does.
Process automation
at a fixed price!
Contact us now.
AIMAX Business Solutions combines excellent solutions with first-class service. Your added value is our goal. Unique AI systems allow us to act independently of the application. With process automation and digital assistance, we unlock new potential in your company.
In addition to distribution and investment, the AI Index also sheds light on the technical performance of AI systems, which once again made enormous leaps in 2024. Last year, new sophisticated benchmarks were introduced to test the limits of advanced AI - for example in areas such as comprehensive world knowledge (MMMU), visual questions (GPQA) or software engineering tasks (SWE-Bench). Just one year later, AI models were able to dramatically improve their results on these tests: Scores increased by 18.8 and 48.9 and 67.3 percentage points on the aforementioned benchmarks , respectively. This shows how quickly research is addressing existing weaknesses and pushing AI systems to performance limits that were previously considered unattainable. It is also noteworthy that AI can now outperform human capabilities in individual settings - the report mentions agent-like AI models that performed better than humans in certain tests. Although such successes relate to specialised tasks, they underpin the trend of steadily increasing competence and autonomy of AI.
The change in who develops AI models is also interesting for companies. In 2024, almost 90% of significant new AI models came from industry, whereas in 2023, 60% came from companies and 40% from academia. The development of powerful AI has therefore shifted strongly to the private sector. Big tech companies and AI start-ups are driving model development with immense resources - both in the training of ever larger models and in the optimisation of smaller ones. This is due to the increased requirements: The training computing power for top models is now doubling every five monthsand the amount of training data used every eight months. Hardly any university can keep up with this speed and budget. For decision-makers, this means that AI innovations reach the market faster today - what is created in the lab is turned into products more quickly thanks to company involvement, whether through open source releases or commercial offerings. However, the ability to build truly leading ("frontier") AI models is concentrated among a few players with large capital and infrastructure.
As a positive side effect of this intense competition, the performance gaps between the top models are shrinking. The AI Index reports that the gap between the best and tenth best models on certain benchmarks has fallen from 11.9% to 5.4% within a year. In some cases, the two leading models were only 0.7% apart. The field of top models is therefore becoming narrower and more competitive - no single model dominates unassailably. For users of AI, this means that there is a greater choice of almost equivalent AI models, whether from different providers or as open source versions. The days of very few AI systems having unique selling points could be coming to an end, which should increase choice and price pressure in favour of users.
Overall, the report shows that AI performance is growing by leaps and bounds and the global race for AI leadership is in full swing. It can be risky for companies to rely solely on one provider or one region when making AI decisions. Instead, the global innovation landscape should be monitored and, where appropriate, multiple sources of AI technology should be considered (such as US and European providers, but also high-quality open source models that are increasingly available). The strong shift of AI development to industry also suggests focussing on practical solutions coming from tech companies and exploring partnerships with such players. Ideally, this should be a solution that offers the greatest possible flexibility in terms of selection and use (including future) AI models.
A key highlight of the AI Index 2025 for practitioners is the realisation that AI systems have become considerably more efficient and cost-effective. Although the cost of developing the very largest models is rising (training the latest cutting-edge systems sometimes costs over USD 100 million, which only tech giants can afford), the opposite is becoming apparent for the use of AI: the application of AI is rapidly becoming cheaper. According to the report, inference costs - i.e. the cost of running a trained model and answering queries - have fallen 280-fold in less than three years.
Specifically: at the end of 2022, it cost around USD 20 to process 1 million tokens (word units) with a model at GPT 3.5 level; at the end of 2024, it was only around USD 0.07. This dramatic drop in price (from $20 to 7¢) fundamentally changes the economic calculation for AI in companies. What was previously only possible with a large budget is now available for a fraction of the cost. Even computationally intensive AI applications - such as a customised chatbot that processes complex requests - are becoming financially attractive.
Several factors contribute to this cost reduction. Firstly, hardware is continuously improving: according to the report, the cost of computing power, especially in the AI sector, is falling by around 30% per year, while the energy efficiency of AI hardware is increasing by 40% per year. Modern AI chips (GPUs, TPUs, etc.) can therefore perform more and more calculations per watt, and the price per computing unit is falling. On the other hand, more efficient AI models and algorithms are emerging. A new generation of smaller, optimised models now achieves the same performance as previous giants, but requires significantly fewer computing resources. As an example, the AI Index states that open, freely available AI models were almost on a par with the large, proprietary models on certain benchmarks in 2024 - the performance gap fell from 8% to 1.7% within a year. These "small" models (sometimes only a few billion parameters instead of hundreds of billions) can often be run on off-the-shelf hardware, which saves on expensive infrastructure.
All of these trends significantly lower the barriers to entry for AI. Companies that previously shied away from the high costs of AI projects are now finding much more favourable framework conditions. Cloud providers are offering AI services at ever lower usage prices, open source communities are providing pre-trained models and AI can also be operated on-premises, i.e. on a company's own infrastructure, with less effort. The latter in particular is becoming attractive again: as efficient models require fewer resources, they can also be run locally within the company. This suits companies that are reluctant to use cloud AI for data protection or latency reasons.
For decision-makers, the message is clear: "AI has become affordable." The financial hurdles are falling, which means that the question "Can we afford AI?" can increasingly be answered with "Yes". Instead, the focus is shifting to how AI can be used most effectively. Because when AI technology becomes a cheap commodity, the competitive advantage lies less in the mere possession of this technology and more in its clever application and integration into the business model.
Those who integrate AI efficiently into their processes at an early stage can realise a cost advantage - for example, by automating repetitive tasks and thus saving on human resources or offering new AI-supported services that scale with minimal marginal costs. At the same time, companies should examine where open source AI represents an alternative to expensive proprietary solutions. The trend of open models catching up makes it possible to act more independently of large AI providers in many cases and further reduce costs.
Responsible AI
With the increasing spread of AI, the attention of regulatory authorities and the requirements for the responsible use of AI are also growing. The AI Index 2025 shows that a real regulatory activism has begun worldwide. In 75 countries surveyed, references to AI in draft legislation increased by 21.3%, continuing the trend of a nine-fold increase in legislative activity on AI since 2016. Governments are clearly recognising the need for action to shape and control AI. At the same time, immense public funds are being mobilised to keep pace technologically: Canada has announced $2.4bn, China a $47.5bn semiconductor fund, France €109bn, India $1.25bn and Saudi Arabia as much as $100bn ("Transcendence" project) to promote AI and related technologies. In short, countries are not only investing in rules, but also in resources to ensure that their economies remain competitive in the age of AI.
For companies, this two-pronged development - more regulation, more promotion - means two things: on the one hand, they need to prepare for compliance and standards. Frameworks for trustworthy AI are already emerging: organisations such as the OECD, EU, UN and the African Union have issued guidelines on the transparency, fairness and safety of AI systems in 2024. In the EU, for example, the AI Act is just around the corner, which imposes strict requirements for AI applications depending on the risk class. Companies should check their AI applications at an early stage to see whether they fulfil ethical and legal requirements - for example with regard to data protection, non-discrimination and the traceability of decisions. On the other hand, government funding initiatives also offer opportunities, e.g. through grants for AI research, infrastructure or pilot projects. It can be worthwhile using available programmes to advance your own AI strategy.
A look at the report also shows that the economy itself still has some catching up to do when it comes to "Responsible AI". Although AI incidents (such as wrong decisions or discriminatory results) are on the rise according to the incident database, major AI developers rarely use standardised assessment procedures for responsible AI. In other words, while many companies are aware of the risks posed by AI, they are not yet acting consistently enough to minimise them. There is a gap between recognising problems (bias, lack of explainability, security gaps) and taking actual action. There is growing pressure from regulators, but also from customers and partners who expect trustworthy AI. For example, more and more clients are demanding proof that the AI systems used are tested and monitored.
One aspect of responsible AI use concerns the protection of sensitive data. Many companies are reluctant to unleash cloud-based AI solutions on their internal data for fear of information leakage or misuse. This is where solutions that allow AI applications in the local environment come into play, so that data remains in-house - a point that we will discuss further in the context of EMMA. Overall, it is clear that AI ethics and governance are moving from a theoretical concept to a practical requirement. Those who establish their own guidelines early on (e.g. internal AI principles, test steps before deploying new AI models, training on bias avoidance) will give themselves a head start as regulation and customer expectations continue to increase.
To summarise, decision-makers should not only look at AI from the perspective of "What's in it for me?", but also "How do I use it responsibly?". In future, the licence to operate for AI-driven processes will depend on being able to guarantee security, fairness and transparency. It pays to actively follow the development of AI regulation and perhaps even help shape it (for example through industry initiatives) in order to avoid unnecessary risks and build trust among stakeholders.
Skilled labour, acceptance and technology limits
Despite all the successes and positive trends, the AI Index 2025 does not hide the fact that there are still hurdles and unanswered questions that companies should consider. One of these is the shortage of skilled labour and lack of training in the AI sector. It is true that the availability of AI education has increased significantly in recent years - two thirds of all countries have now introduced AI or computer science lessons in schools or are planning to do so (twice as many as in 2019). Nevertheless, in the USA, for example, less than half of computer science teachers feel adequately prepared to actually teach AI topics. The situation is similar in companies: The tools may be getting simpler and simpler, but the workforce needs further training in order to use AI sensibly. Companies should therefore invest in training and talent development alongside technical implementation. This includes training for existing employees, interdisciplinary teams (bringing domain experts together with data scientists) and attractive career paths for AI specialists in order to attract and retain talent.
From a technical perspective, current AI still has its limits when it comes to certain skills. One buzzword here is the "reasoning" problem: while AI is excellent at recognising patterns and can even deliver creative output, it still struggles with complex logical thinking. The report points out that large language models still lag behind human performance in rigorous reasoning tasks (e.g. multi-step mathematical proofs or strategic planning) and are often unreliable. For example, they fail on tasks that require precise reasoning or formal logic, which limits their use in highly critical areas (e.g. legal reasoning, safety-critical control systems). Companies must therefore be aware of the limitations of AI systems: Not every task should be blindly left to AI, and human expertise remains essential in many areas. A pragmatic approach is to use AI specifically where it is clearly better or more efficient and at the same time to define processes in which humans intervene to check or supplement (keyword: human-in-the-loop).
Another issue is sustainability. Despite more efficient hardware, overall energy consumption has increased enormously as a result of the AI boom. The largest AI models today cause thousands of tonnes of CO₂ emissions during training and operation. This issue is coming into focus as many companies pursue climate targets. It will be important here to focus on energy-efficient AI solutions - for example, favouring smaller models, operating data centres with clean energy or optimising the use of AI where it brings the greatest added value. Providers such as large cloud platforms are already responding (e.g. by investing in climate-friendly data centres), and sustainability can become a selection criterion for AI services for users.
Finally, acceptance within the company itself should be mentioned: the introduction of AI can bring internal resistance - be it fear of job losses or scepticism towards "black box" systems. This is where change management and transparent communication are crucial. Employees should understand that AI is a tool that relieves them of routine work and opens up new opportunities. Experience shows that the best way to successfully introduce AI is for managers to involve all stakeholders at an early stage, make successes visible and communicate a collaborative vision of human-machine cooperation.
"The report impressively demonstrates the transformative effect of AI and thus confirms our own personal feelings.
The change is not limited to one sector, but is cross-sectoral.
Those who fail to act now will soon be forced to do so by the competition. Do you want to act or react?
I would be happy to talk to you personally."
In view of the developments outlined above, the question arises as to how companies can actually make the leap onto the AI bandwagon. This is where the EMMA and AIMAX® solutions come into play, which address precisely the points identified. The trends described in the AI Index - further spread of AI, lower costs, generative AI capabilities, need for accountability - are reflected in the focus of these tools.
Cognitive AI EMMA
EMMA is a no-code RPA solution with cognitive AI. RPA (Robotic Process Automation) makes it possible to automate repetitive business processes by mimicking human interactions with software. No-code means that no programming knowledge is required - specialist users in the departments can define processes using a graphical user interface. In this way, EMMA addresses the shortage of skilled labour: AI-based automation can be anchored in the company even without a team of data scientists or developers. This fits in with the trend of democratised AI (AI is becoming more widely accessible), as shown by the AI Index.
EMMA also integrates cognitive AI components - in other words, the software can go beyond pure if-then rules to understand text input or read documents, for example. These capabilities can also be used to automate unstructured tasks, which increases productivity and relieves employees of monotonous work steps.
What is remarkable about EMMA is that no interfaces to the applications to be automated are required. The solution interacts like a human being with various systems via the existing user interface. This drastically reduces integration costs and effort - an important factor as, according to the AI Index, workflow integration will be a key differentiator in the future as AI technology itself becomes more available. EMMA can be operated locally at the customer's premises, which directly addresses the issue of data security and compliance: all data remains within the company's sphere of influence and even strict data protection guidelines (e.g. in regulated industries such as finance or healthcare) can be adhered to. At a time when regulators and customers are placing the highest demands on the protection of sensitive information, EMMA thus offers a responsible way to use AI - in line with the trend towards more responsible AI observed in the AI Index.
AI agent AIMAX®
AIMAX® in turn, is a generic AI agent that can interact with EMMA. AIMAX® can be thought of as a kind of artificial colleague with generative AI capabilities. It can be used across industries and departments, i.e. its expertise is not limited to a specific area - an echo of the broad applicability of AI across many domains mentioned in the AI Index. AIMAX® complements deterministic automation with generative AI expertise: this means that where EMMA executes rigidly defined processes, AIMAX® can act flexibly, draw conclusions or generate content. Together, they form a team in which EMMA takes care of the precise, rule-based processing of tasks, while AIMAX® provides the creative, knowledge-based component. This interaction reflects the latest developments in the AI world, in which hybrid approaches are becoming increasingly important - such as the combination of classic algorithms and modern AI models. For companies, this opens up the possibility of automating end-to-end processes: From the interpretation of an unstructured input (e.g. an email enquiry, which AIMAX® helps to understand and answer) to the execution of specific transaction steps in the IT systems (which EMMA takes care of).
The strength of AIMAX® lies in the fact that it brings Generative AI directly into business processes. After the hype surrounding ChatGPT & Co, many companies are asking themselves how they can utilise such capabilities in practice. AIMAX® offers a solution here by acting as a universally applicable AI module that works with EMMA in customer service, sales, HR or any other area, for example. It can thus support departments without the need to develop a separate AI model for each application. What is important here is that AIMAX® - just like EMMA - can be integrated into the existing system landscape without disrupting it. It "talks" to EMMA, which in turn interacts with the existing applications. This allows companies to gradually add AI functionality instead of having to set up processes from scratch. This also significantly lowers the barrier to AI adoption. The generative expertise of AIMAX® ensures that even unclear or changing requirements can be managed - a major advantage over purely deterministic systems, which reach their limits in every exceptional case.
EMMA and AIMAX®
- a platform to utilise the opportunities
To summarise, EMMA and AIMAX® provide companies with a platform to exploit the opportunities identified in the AI Index: They make it possible to roll out AI broadly (no-code approach, generic agent), save costs (through automation and by eliminating large integration projects), act responsibly (local operation, data protection) and implement new capabilities such as generative AI securely. Especially in light of the fact that AI technology is becoming increasingly affordable and the major challenge now is to cleverly interweave AI with business processes, such solutions offer a practical approach. Instead of having to train AI models themselves, decision-makers can use EMMA and AIMAX® to immediately access tried-and-tested AI functionalities and make them effective in their company.
The State of AI Index 2025 paints a picture of a technology area that is growing in importance at breathtaking speed. For companies, the key message is clear: AI has gone from the future to the present. With over 78% company adoption, drastically falling utilisation costs and more and more use cases across all industries, there is little doubt that AI will have a lasting impact on competitiveness. Anyone looking into the executive suite today will recognise that competitors, partners and customers have long since begun to use AI to their advantage.
Decision-makers should see this development as a call to action. Now is the time to formulate and implement a clear AI strategy. This does not mean mindlessly chasing after every trend, but rather making informed decisions: Where can AI add the most value to your own business model? Where is there unused data and processes that can be automated or optimised? Which decisions can AI support? Equally important: what risks do you need to be aware of and how can you ensure that AI is used ethically and in compliance with the law?
The good news is that the hurdles have never been lower. AI technology is more accessible and affordable than ever, there are proven solutions (from open source models to enterprise platforms such as EMMA and AIMAX®) and a growing pool of expertise. Companies can quickly achieve success in pilot projects and then scale them up. You should think big, but start small: In other words, set ambitious goals, but start with manageable projects from which you learn. Experience shows that an iterative approach - experiment, adapt, roll out - makes sense when introducing AI in order to create acceptance and minimise risks.
However, one thing must not happen: Standstill. In view of the speed and investment targets from the AI index, doing nothing would probably be the most dangerous option. Companies that wait and see now run the risk of being left behind in a few years' time when competitors are faster, cheaper and more innovative thanks to AI. This does not mean that every company has to invest millions in AI immediately. But every company should now have a plan for developing AI expertise - be it through its own projects, cooperation with service providers or the use of platforms that provide AI capabilities. The important thing is to get started.
In conclusion, AI is a transformative factor, comparable to electrification or the internet. The Stanford AI Index 2025 provides the data that proves this transformation. For decision-makers in companies, it is now important to translate these findings into their own strategy. With considered, proactive action - supported by suitable solutions such as EMMA and AIMAX®, which make it easier to get started - it is possible to utilise the opportunities offered by AI while avoiding stumbling blocks. In this way, the company remains competitive in the digital transformation and helps to shape the future instead of being driven by it.
