Join the event trusted by corporate leaders for almost two decades. VB Transforma connects people building AI Real Enterprise. Learn more
Over the past 100 yearsIN ANDBm Many different technological trends will escalate and fall. What usually wins is the technologies in which there is a choice.
On VB Transform 2025 Today, Armand Ruiz, vice president of the AI platform in IBM, described in detail how Gigantic Blue thinks about generative artificial intelligence and how its company users actually implement technology. The key topic that Ruiz emphasized is that at this point it is not about choosing one supplier or technology of one gigantic language model (LLM). Increasingly, corporate clients systematically reject AI strategies for multi -model approaches, which match certain LLM to targeted operate cases.
IBM has its own AI models with open sources with a granite family, but does not position this technology as the only choice and even the right choice for all loads. This company’s behavior leads IBM to positioning not as a competitor of the foundation model, but as what Ruiz called a control tower for AI loads.
Wielorakie Gate Strategy
IBM’s response to this market reality is a newly issued model gate that provides enterprises with a single API to switch various LLM, while maintaining observation and management in all implementation.
Technical architecture allows clients to launch Open Source models at their own pile of inference for sensitive operate cases, while gaining access to public API interfaces, such as AWS Bedrock or Google Cloud’s Gemini for less critical applications.
“This gateway provides our clients with one layer of a single API to switch from one LLM to another and add observation and management at all times,” said Ruiz.
This approach directly denies the Seller’s joint strategy in the field of locking customers in reserved ecosystems. IBM is not alone by taking a multi -family approach to choose from the model. In recent months, many models routing tools have appeared to direct the load to the appropriate model.
Agent’s orchestration protocols appear as a critical infrastructure
In addition to managing many IBM models, he deals with the appearance of agent-agent communication through open protocols.
The company developed ACP (agent communication protocol) and contributed to the Linux Foundation. ACP is a competitive effort for the Google Agent2agent (A2A) protocol, which this week was brought by Google to the Linux Foundation.
Ruiz noticed that both protocols are aimed at facilitating communication between agents and limiting non -standard development works. He expects that the various approaches will eventually coincide, and currently the differences between A2A and ACP are mostly technical.
Agent’s orchestration protocols provide normalized ways of interaction of AI systems with various platforms and suppliers.
Technical significance becomes clear when considering the company’s scale: some IBM customers already have over 100 agents in pilot programs. Without standard communication protocols, each interaction of an agent-agent requires non-standard development, creating an unbalanced integration burden.
AI concerns the transformation of work flows and how to work
As for how Ruiz sees that AI is affecting enterprises today, he suggests that it really must be more than chatbots.
“If you just do chatbots or try to save costs from artificial intelligence, you don’t do AI,” said Ruiz. “I think that AI really involves the complete transformation of work flow and how to work.”
The distinction between the implementation of AI and the transformation AI focuses on how deeply technology integrates with existing business processes. The internal example of HR IBM illustrates this change: instead of employees asking Chatbota for HR information, specialized agents now support routine queries regarding compensation, employment and promotion, automatically leading to appropriate systems and escalation to people only if necessary.
“I used to spend a lot of time talking to my HR partners for many things. I am now dealing with the majority of HR,” Ruiz explained. “Depending on the question, whether this is something in the field of compensation, whether the point is to simply support separation, renting someone or promotion, all these things will connect to various HR internal systems, and these will be like separate agents.”
This is a fundamental architectural shift from the patterns of human computers to automate work flow via a computer. Instead of employees studying in interaction with AI tools, and learns to perform complete business processes from end to end.
Technical implication: enterprises must go beyond the integration of the API and rapid engineering in the direction of an assignment of a deep process that allows AI agents to autonomous exercise of multi -stage work flows.
Strategic implications for investments AI Enterprise
IMBM implementation data suggest several critical changes for the AI Enterprise strategy:
Give up chatbot thinking: Organizations should identify the full flow of transformation work, and not add conversation interfaces to existing systems. The goal is to eliminate human steps, not to improve the interaction of human computers.
Architect for the flexibility of many models: Instead of committing to individual AI suppliers, enterprises need integration platforms that allow you to switch between models based on the requirements for operate while maintaining management standards.
Invest in communication standards: Organizations should prioritize AI tools that support emerging protocols such as MCP, ACP and A2A, and not the reserved integration approaches that create the supplier’s blockade.
“There is so much to build, and I repeat that everyone must learn artificial intelligence, especially business leaders, they must be the first AI leaders and understand the concepts,” said Ruiz.