As AI agents begin to participate directly in transactions, innovation and decision-making, ecosystems will become faster, leaner and far more complex than any human network could manage manually, Jan Bosch explains.
When I started this series on the AI-driven company, we had developed three maturity models: AI in the company’s business process, the use of AI for the R&D process and AI in the product. Since then, however, we’ve developed a fourth model, which is concerned with AI in the business ecosystem that the company operates in.
The reason for exploring this fourth dimension is that software ecosystems have traditionally focused on human collaboration, ie developers, suppliers, customers and partners co-creating value around shared platforms. From Apple’s iOS to Siemens Mindsphere, ecosystems have allowed companies to innovate faster than any one organization could do alone.
However, the fabric of these ecosystems is changing. With the emergence of generative AI and agentic AI, we’re entering an era where digital agents begin to act and interact on behalf of humans. In the near future, the participants in our ecosystems will no longer be exclusively human. Some will be autonomous software entities making decisions, negotiating and coordinating without human intervention.
This shift isn’t just about efficiency or automation. It represents a fundamental transformation in collaboration, competition and value creation. What used to be a human-centric, trust-based structure is becoming an algorithmic, data-driven web of autonomous interactions. As it says in the title of the paper we wrote on the topic, the ghost is in the machine and it’s starting to talk back. We’re not there yet – for now, ecosystems remain firmly human-centric – but the first signs of a paradigm shift are visible.
The companies we work with see clear opportunities in this space. For example, bringing in AI agents allows us to prototype and experiment much faster and at a much lower cost per experiment. Second, it allows for AI-augmented decision-making where agents provide the relevant input for decisions, even if humans are still making the call. Finally, it allows for smarter collaboration, as AI can mediate interactions between partners, even when formal standards don’t yet exist.
At some point, however, the balance between humans and AI agents will change and we’ll see the emergence of AI-driven ecosystems where AI agents are full participants. These agents negotiate, optimize and even co-innovate across organizational boundaries. Humans remain involved, but primarily as orchestrators, supervisors and strategists rather than operators. Companies will increasingly compete on their ability to orchestrate multi-agent interactions across partners and platforms.
Based on our work with multiple software-intensive companies, several dimensions of this transformation stand out. First, collaboration. Traditional ecosystems rely on personal relationships and trust. With AI, collaboration becomes protocol-oriented and is governed by digital contracts and API-driven negotiation rather than human rapport.
Second, AI dramatically reduces the cost of negotiation and coordination. Consequently, transaction costs will plummet. Smart contracts and agent-mediated interactions allow for micro-transactions, ie small, fast exchanges across partners that used to be too expensive to manage manually.
Third, this will significantly accelerate the tempo of the ecosystem. Agents benchmark offers, compare prices and change partners in real-time. Information asymmetry will be much harder to maintain, causing differentiation to erode faster.
Fourth, the key differentiator will be data because the AI models and algorithms will be accessible to everyone at roughly the same time. Proprietary, high-quality and domain-specific data gives companies a sustainable advantage. In an ecosystem context, success will increasingly depend on strategic choices on the exchange of data, ensuring that certain critical data stays proprietary and isn’t shared.
The fifth dimension is concerned with the power structure of the business ecosystem. Traditionally, the keystone players and the move to online business caused intermediaries to be removed from the ecosystem. With AI, however, we hypothesize that AI intermediaries may become the new power brokers. They provide the integration layers where agents connect and collaborate. These intermediaries control access, coordination and the flow of value across ecosystems.
In this context, strategy becomes increasingly critical, both in terms of definition and execution. In an AI-driven ecosystem, autonomous agents can start to behave adversarially, eg forming cartels. Companies need governance frameworks and human-in-the-loop mechanisms to ensure alignment with defined strategy.
Finally, companies that seek to be leaders in AI-driven ecosystems must focus on owning interfaces (APIs, data contracts), governing agent behavior, forming strong data partnerships and ensuring that humans remain involved and are escalated to for ambiguous or high-risk situations.
To describe how companies will move through this transformation, we developed a five-step maturity model for AI in software ecosystems. They start with a human-centric ecosystem and then bring in AI solutions to assist the humans. In step three, AI agents can autonomously conduct routine transactions, but humans still execute the more complicated tasks. Over time, more and more tasks are moved to agents and there’s continuous data exchange, management of supply and demand and assurance of compliance governed by agents. The final step is the AI-first ecosystem, where AI agents run the ecosystem and humans, for all practical reasons, only provide oversight.
In our interviews, most companies place themselves between steps 1 and 2 and are still human-centric but beginning to explore AI assistance. The later stages, especially 4 and 5, remain largely hypothetical but are rapidly approaching in multiple industries. The implication is clear: organizations that learn to manage both human and machine participants in their ecosystems will move faster, scale more effectively and gain access to entirely new value creation mechanisms. Leading an AI-driven ecosystem means leading through architecture, data and delegation rather than hierarchy and control.
The shift toward AI-driven ecosystems mirrors earlier transformations: from products to platforms and from pipelines to ecosystems. The ecosystem itself is becoming intelligent and adaptive. As AI agents begin to participate directly in transactions, innovation and decision-making, ecosystems will become faster, leaner and far more complex than any human network could manage manually. Ultimately, the future of software ecosystems isn’t humans versus machines but humans designing ecosystems where machines collaborate on our behalf. To end with a quote by Tim O’Reilly: “Every industry and every organization will have to transform itself in the next few years. What’s coming at us is bigger than the original internet.”
