The strategic battle between tech giants to orchestrate agents
Over 40 years ago, SAP revolutionized the business world with ERP systems and created the market for ERP systems. Other so-called “systems of record” (SoR) systems such as CRM and MIS have since established themselves as the digital backbone for human-driven processes. Today, we are once again on the threshold of profound change: the age of “Systems of Action” (SoA) has dawned. Intelligent AI agents are taking on increasingly complex tasks and acting autonomously – a development that will radically change traditional systems of record solutions such as SAP, Salesforce, and even production planning systems.
Which platforms will prevail, and what should entrepreneurs and investors focus on now? Will these platforms be able to replace human work on a broad scale in organizations through agents?
The big players in the battle for the AI agent market
Four major providers are currently battling to become the new standard for orchestrating AI agents:
1. ServiceNow – Central control and openness
With its “AI Control Tower” and “Agent Fabric,” ServiceNow is positioning itself as the central control platform for autonomous AI agents – across all providers. The strength of this solution lies in its ability to orchestrate agents from all manufacturers in a standardized manner while transparently implementing governance and compliance requirements.
ServiceNow scores with:
- A central governance platform (Control Tower) that provides an overview of AI activities and their ROI.
- Openness through the A2A (Agent2Agent) protocol, which seamlessly integrates agents from different manufacturers.
- Extensive partnerships (including Box, Microsoft, IBM, Google), which avoids vendor lock-in for companies.
2. Microsoft – The integrated AI platform
Microsoft is pursuing an integrated strategy with Azure AI and Copilot tools. It connects internal applications such as Microsoft 365 with complex multi-agent workflows. The advantage: deep integration into the existing Microsoft world.
Microsoft offers:
- Integrated solutions embedded in Office and Azure environments.
- Proprietary frameworks such as Semantic Kernel for creating complex multi-agent systems.
- Support for open standards such as A2A, but with a clear focus on Azure.
3. Google – Openness and developer focus
Google is committed to openness. With Vertex AI Agent Builder, Google offers comprehensive tools for creating agentic systems, while “Agentspace” provides a user-friendly interface for managing and deploying these agents.
The Google approach:
- Maximum openness and developer-friendliness with open source tools such as the Agent Development Kit (ADK).
- A standardized, open ecosystem, strongly driven by the Agent2Agent protocol (A2A).
- Clear focus on speed of innovation and community building.
4. IBM – Automation and control in the enterprise environment
IBM is positioning itself as a leading provider of enterprise automation with Watsonx Orchestrate. Watsonx Orchestrate is tightly integrated into the existing IBM environment and focuses heavily on governance and compliance in the enterprise context. By connecting to IBM Cloud, IBM Watson, and extensive automation capabilities, IBM offers a robust, reliable platform for complex and regulated processes.
IBM scores with:
- Strong integration with IBM’s established enterprise software (e.g., Watson AI, IBM Cloud).
- Extensive compliance and audit capabilities that are essential for highly regulated industries.
- Focus on enterprise process automation with a high degree of control and transparency.
5. Salesforce – Agentforce for intelligent CRM
With Agentforce, Salesforce is focusing on the intelligent expansion of its CRM platform. Agentforce enables companies to embed AI agents directly into their customer-facing processes such as sales, marketing, and customer service. The focus is on personalized customer interactions and improving the customer experience through autonomous AI support.
Salesforce scores with:
- Deeply integrated AI agents within the Salesforce CRM ecosystem, including Sales Cloud and Service Cloud.
- Robust data protection and security mechanisms via the Einstein Trust Layer.
- A strong focus on customer processes and experiences, supported by generative AI models and seamless workflow integration.
In addition to the big players, there are a number of startups and smaller companies positioning themselves in the market. The dilemma for entrepreneurs and investors is that the costs suck enormous value out of companies. If you choose smaller and therefore significantly cheaper solutions, the investment security is much lower.
The science behind the agent revolution
From a scientific perspective, these solutions are based on decades of research into multi-agent systems (MAS). Since the 1980s, research has been focusing on agent models, communication standards (e.g., FIPA, ACL), and the collaboration of autonomous systems. One core problem is that agents must coordinate their actions without losing sight of the big picture.
Modern research (e.g., Stanford Virtual Lab, Generative Agents experiments) impressively demonstrates that teams of specialized AI agents are superior to individual agents. Agents can exhibit emergent behavior, find new solutions, and massively accelerate processes—but they also harbor risks if control is lacking.
An important field of research is therefore currently the governance of autonomous systems: How can we ensure that agents act in accordance with human values and regulatory requirements? Scientists are calling for transparent control mechanisms, human oversight, and clear rule systems – precisely the requirements that ServiceNow, Microsoft, Google, and others are now addressing in their orchestration platforms.
Legal framework: Who is liable when AI makes decisions?
The legal situation regarding the use of autonomous AI in decision-making processes is currently unclear and poses a significant risk for companies. In the case of human decisions, liability is clearly regulated: Companies are generally liable for the mistakes of their employees, albeit to a limited extent and supported by insurance coverage. There are no comparable regulations for AI as yet.
AI systems often act as a “black box,” making it difficult to clearly assign responsibility.
Currently, there is no clear legal basis in either Germany or the US that regulates liability issues for autonomous AI decisions. We are therefore currently trying to have AI decisions formally “approved” by humans – an inefficient approach that prevents real productivity gains (“man-in-the-loop” approach).
Professional associations such as medical associations and air traffic controllers’ unions are hindering the necessary progress with their obstructionist attitudes and spurious arguments. In doing so, they are also preventing important social productivity gains through AI systems. In the US, on the other hand, large providers are simply creating facts. The legal gray area is much smaller there due to generally stricter liability principles. The US will therefore adapt agent-based autonomous decision-making systems more quickly and pragmatically than Germany and the EU.
Germany could take a leading role here with its new Digital Ministry. It could create clear, legally secure framework conditions, prescribe interfaces and interoperability, and thus enable cross-system governance systems. This would allow companies to deploy agent-based AI systems more quickly and securely.
In concrete terms, this means:
- Creating a legally secure framework for autonomous AI.
- Introducing clear liability concepts, analogous to regulations for human error.
- Establishing binding governance and compliance systems for AI agents.
- Promoting specialized insurance products for AI risks.
Without bold action, Europe risks falling behind again in international competition. We urgently need clear and progressive regulations to unlock the enormous productivity potential of AI, robotics, and autonomous systems.
Which solution will win? Three scenarios
Regardless of the solutions presented by the big tech companies, there are three scenarios for how the agent orchestration market could develop in the future.
Scenario 1: Dominance of a central orchestrator (ServiceNow model)
ServiceNow could prevail if companies prefer a central control authority. Single-pane-of-glass control offers enormous advantages in terms of control and governance, especially in highly regulated industries such as finance and healthcare.
Scenario 2: Federated multi-standard model
Alternatively, a federated solution could emerge in which several specialized orchestrators (e.g., Microsoft for internal processes, Salesforce for CRM, Google for search and knowledge management) are operated in parallel. This would require broad acceptance of open standards (such as A2A) that enable collaboration.
Scenario 3: Decentralized autonomy with open source
A third option would be decentralized, autonomous control. Companies could use open source technologies to build their own customized agent systems and adapt them flexibly. This model would be cost-effective but would require extensive internal expertise.
So whether we see scenario 1 with overwhelming US dominance of another US platform, or a real leap in productivity in our economy through scenarios 2 or 3, also depends on the state and its ability to play a meaningful role in shaping the digital transformation in the future. Added to this is the need to create the necessary legal framework.
Will AI agents replace human labor on a large scale?
The big fear of many employees is: Will AI agents take over the majority of jobs? The short answer: Not everything, but significantly more than we might like.
The fact is, AI agents are already performing routine tasks faster, cheaper, and more accurately than humans.
Anyone wondering whether AI will soon become standard in customer service, IT support, or simple administrative processes has already answered the question. These activities are on the verge of disruption.
Of course, AI still has its limits today, especially when it comes to complex, creative, or highly emotional jobs. But beware: these limits are shifting faster than streaming services took over the cable TV market.
Generative AI and language models are well on their way to massively changing not only simple tasks, but entire job profiles such as lawyers, marketing managers, and analysts.
Companies should therefore start thinking less about short-term efficiency gains and more about long-term strategy as soon as possible. AI is not only replacing tasks, but also fundamentally changing the question of which skills will be in demand in the future. Those who fail to recognize this change early on and act accordingly risk AI overtaking not just individual jobs, but entire companies. Companies that promote the right skills now – especially management, creative problem solving, and strategic control – will be the winners of the next decade. Everyone else should brace themselves.
Conclusion: Regulation will also determine which companies and concepts win
We are now in the same exciting phase as we were between 1998 and 2002 with the Internet age and the emerging web sector. As was the case back then, there are enormous expectations, a great deal of hype, and many of the future winners in what was then a 20-year race are already at the starting line. But even in 2002, it was not yet possible to clearly identify the winners. Interestingly, 10 years later, no new players have managed to work their way into today’s top 10.
But yes, many companies that were still in top positions at the beginning of the Internet boom (such as Netscape or Yahoo) will be weeded out, even though no 20-year-old knows them today.
Internet and web technology took us by surprise in Europe, and we were socially overwhelmed by the Silicon Valley attitude of “just do it.” Neither at the beginning of the wave—in the early 2000s—nor 15 years later were there any serious legal frameworks for consumers. EU regulation, which can rein in the big tech players to some extent, has only emerged in recent years, almost twenty years late.
Unlike back then, the development of autonomous AI agents in Europe is already being slowed down by strict labor regulations relating to liability. In the US, the lack of regulations is much less of an issue due to a different liability regime.
If we in Europe do not very quickly find reasonable political and economic framework conditions for AI liability and for training models, we will not be able to exploit the enormous potential for productivity gains in Germany and Europe. This would leave us lagging behind on the world market with our products!
I would like to see clear impetus from our Ministry of Digital and Economic Affairs for AI governance that will enable us to drive forward this transition to the use of AI agents in all industrial and economic sectors in Germany.
