The concept of software components orchestrating across boundaries is not new—it has appeared in mainframe program libraries, Service-Oriented Architectures (SOA) and Enterprise Service Buses (ESBs). APIs became the next wave of this idea, powering cloud, mobile and software-as-a-service (SaaS) ecosystems.
But the next generation is arriving: agentic AI. Where APIs exposed capabilities in rigid ways, AI agents bring interactive integration—flexible workflows that adapt dynamically to user intent and context. Instead of brittle integrations requiring schema alignment and precise calls, agentic AI systems can reason across applications, interpret unstructured inputs and broker actions on behalf of users.
This shift matters because the silos created by sprawling SaaS portfolios remain very real. APIs mitigated them but did not erase them. Agentic AI provides a new layer of connective tissue: a service bus that can flexibly interact with disparate systems, not just move data between them. Our recent Data and AI Study highlights just how critical integration has become to getting value from AI and data, with many of the top 10 issues directly related to the challenges with silos, integration and the ensuing complexity.
Three forces are driving this shift:
- SaaS sprawl has outpaced API-centric integration.
Most enterprises run dozens or hundreds of SaaS tools. APIs exist, but stitching them together via ETL pipelines or IPaaS platforms is brittle, expensive and hard to adapt when business processes evolve. AI agents can orchestrate these applications at the “interaction” level—filling forms, extracting insights, initiating actions—rather than relying only on structured API calls.
- Interactive integration is replacing static pipelines.
Traditional integration meant data mapping and synchronization. Agentic AI allows for conversational, context-driven orchestration. An agent can retrieve data from one SaaS system, reason about its relevance, and trigger actions in another—even when APIs are incomplete, inconsistent or absent. This enables business users to stitch together workflows that were previously locked behind IT-heavy integration projects.
- Service billing is finally granular.
API-based billing models remain uneven, but AI has made service-level billing real. Token-based pricing from OpenAI, Anthropic and others makes the cost of individual AI-powered calls transparent. This allows organizations to evaluate the ROI of integrations at a far finer level than per-seat or per-dataset SaaS billing ever allowed. For the first time, organizations can measure the economics of machine-driven workflows in unit terms.
The emergence of agentic AI as the new service bus carries major implications:
- Providers that expose capabilities for agents will win: Just as APIs became a competitive necessity, providers that optimize their applications for AI-driven orchestration will see adoption accelerate. Those that remain closed will be left behind as agents automate around them.
- ETL and iPaaS are not enough: Standard ETL tools cannot easily address the messy, interactive nature of SaaS workflows. Enterprises will demand AI-native integration layers that combine reasoning, orchestration and execution across applications.
- Cost clarity enables governance: Token-based billing makes it feasible to monitor and meter integration usage at a granular level. CIOs can now see not only the existence of integrations, but their economic footprint. This visibility was missing in the API economy and will drive new practices in governance, budgeting and optimization.
- New agility for enterprises: Agentic AI reduces the latency between business need and system execution. Instead of waiting for IT to design, test and deploy integrations, business users can delegate integration tasks to agents, while IT provides governance and oversight. This accelerates reaction times to market changes and reduces integration backlogs.
The API economy provided the first wave of democratized integration, but it never fully delivered on its promise of eliminating silos or enabling fluid workflows. Agentic AI represents the next generation: a service bus that thinks, intermediating between software silos and delivering interactive integration.
Our research suggests that enterprises should consider providers that expose their systems in ways that agents can easily interact with; that build governance models around tokenized usage and cost control; and that embrace flexible, AI-driven orchestration as a complement (and, in many cases, a replacement) to traditional ETL pipelines.
Where the ESB and API economy offered rigid connectors, agentic AI offers adaptive interoperability. This is not just another service bus—it is the foundation of a new enterprise integration paradigm.
Regards,
Alex Bakker