Analyst(s): Nick Patience, Brad Shimmin
Publication Date: May 9, 2025
What is Covered in this Article:
- IBM extended its watsonx.orechstrate offering to enable further the responsible development, orchestration, and adoption of AI agents.
- Enhancing its support for unstructured data in supporting agentic AI development, IBM introduced several new watsonx.data integration and semantic search capabilities.
- Highlighting the value of smaller large language models (LLMs), IBM announced Granite 4.0 Tiny Preview.
- IBM introduced webMethods Hybrid Integration as a means of simplifying the connections between complex and increasingly agentic software applications.
The Event – Major Themes & Vendor Moves: IBM dedicated its annual customer conference in Boston to the same two key themes as last year: hybrid cloud and AI. This time, there was more of a focus on productivity and tangible ROI, or at least the pursuit of it. And like every other major vendor, IBM promoted an aggressive agentic AI strategy built on top of its watsonx platform that elevates the operationalization of complex agentic workflows and underscores the importance of surfacing high-quality unstructured data as fuel for those workflows. To this end, IBM announced several new watsonx.orchestrate and watsonx.data capabilities, intended to gradually fill in the details necessary to demonstrate customer value.
IBM Think 2025 – Watsonx Platform Fuels Agentic AI and Hybrid Cloud Value
Analyst Take: During his keynote, CEO Arvind Krishna declared that “AI is the source of productivity for this era.” That’s certainly the bet that IBM – and all other major technology companies – are placing with agents as the source of that productivity boom. But IBM admitted that its extremely early days in terms of AI-driven productivity improvements being commonplace, but it did cite some of its own findings. Treating itself as ‘client zero’, IBM has taken $3.5 billion out of its costs by applying AI, such as saving 125,000 hours per quarter in case summarization and twice as many hours per quarter in HR functions. In procurement, it’s expected to save hundreds of thousands of hours streamlining supply chain compliance. Deutsche Telekom CTO Peter Leukert cited being able to do four times as many operating system es in 22% of the time, reducing critical vulnerabilities.
Not all of IBM Think 2025 revolved around the best way to code, deploy, and manage AI software. At the outset of nearly all keynote sessions, IBM executives emphasized the power, importance, and responsibility tied up in enterprise data, with Sr. VP, Software, Diniesh Nimral, going so far as to say that data is the most powerful tool any AI agent can have. The trouble with this view, according to IBM, is that AI cannot gain meaningful access to the vast majority of enterprise data, as that data often resides within unstructured or semi-structured sources such as documents, images, audio, etc.
For IBM, the key to unlocking this resource rests within the evolution of the watsonx.data platform, building on several technologies specific to modern, open data lakehouse architectures. IBM showcased a recently announced update to Db2, bringing in-database vector data support to the company’s enduring relational database. Expected to reach customers in June, the new version of Db2 (version 12.1.2) will also feature support for the Apache Iceberg open table format and extend hybrid cloud deployment options now to include not just AWS but also Microsoft Azure.
Rather than just deliver basic unstructured data storage options through Db2 and watsonx.data Lakehouse Architecture, IBM demonstrated several new features and products designed to put unstructured data to work in the service of generative and agentic AI. For example, expected to debut in June as a standalone product, watsonx.data integration will enable data professionals to build complex data pipelines that do more than merely import and vectorize unstructured data as an embedding for use in retrieval augmented generation (RAG) workflows.
This new, visual user experience also extracts metadata and categorizes content, storing that contextual information in a relational format for use alongside semantic search. The idea is to improve upon traditional RAG techniques, enabling companies to query unstructured data more accurately while adhering to privacy, security, and observability governance requirements. These critical capabilities will influence how companies build agentic software going forward.
Like every other major vendor, IBM has an agentic AI strategy and is gradually filling in the details. The most significant announcement in this regard was the continued expansion and deepening of the watsonx platform, particularly watsonx Orchestrate and the introduction of a catalog of 150 pre-built AI agents, alongside partner agents. New features in watsonx Orchestrate include a no-code agent builder for business users and an agent development toolkit for technical teams. The idea being making it easier for organizations to build, deploy and manage agents and thus integrate AI into actual business processes and workflows with ready-to-use components.
IBM also announced IBM Granite 4.0 Tiny Preview – an early version of the smallest model in the forthcoming Granite 4.0 model family. Despite only being trained on 2.5T of its planned 15T+ tokens, this model’s performance already rivals IBM Granite 3.3 2B Instruct. This is achieved with fewer active parameters and approximately 72% less memory, according to IBM. Granite 4.0 Tiny will be launched in the summer, along with Granite 4.0 Small and Granite 4.0 Medium, but IBM has put this preview version on Hugging Face so developers can try it out. IBM says it is usable on consumer-grade GPUs.
In terms of hybrid integration and infrastructure modernization, IBM announced Webmethods Hybrid Integration, which aims to simplify the complex task of connecting disparate systems (APIs, file transfers, B2B workflows) and promised significant downtime reduction.
Having announced its Z17 mainframe in April and dubbed it the first mainframe engineered for the AI age, IBM announced the LinuxONE 5 mainframe server at Think the difference being unlike Z17, which runs on z/OS, LinuxONE runs the anonymous operating system. This fifth generation of LinuxONE is powered by the IBM Telum II processor and is focused on the most secure Linux-based applications, with financial service customers to the fore. IBM cited its 20% YoY growth for LinuxOne and adoption by new clients, including those without prior mainframe experience.
What to Watch:
- Can IBM effectively maximize pending acquisitions such as DataStax, legacy product updates (e.g., Db2 12.1.2) and new capabilities like watsonx.data intelligence software, blending those into a single, compelling platform offering that can be effectively integrated into each customer’s unique data estate with little or no professional services support? IBM has struggled to present a unified platform, and while the company has now settled upon a unified yet extensible brand under the watsonx moniker, the company’s recent slate of acquisitions and the rapid evolution of watsonx.data, watsonx.governance, watsonx.orchestrate, et al., will test IBM’s ability to bring new functionality to market without fragmenting its overall go-to-market messaging.
- IBM has been banging the hybrid cloud drum for a while now, and it has a strong story, being able to deploy on any hypersaler, more than a thousand other cloud service providers around the world and of course, on-premises. That was a significant differentiator while the hyperscale cloud companies were insisting that everything will eventually move to the cloud. Now those same hyperscalers are revisiting their own hybrid strategies and pushing them harder (though public cloud is still their main focus by a long way) as customers push back on moving everything to the cloud and are even repatriating some cloud workloads back on-premises.
- How will IBM leverage its new unstructured data and agentic AI capabilities to capture the minds of enterprise developers? IBM has a strong history in supporting enterprises in building Java applications. And with a strong infrastructure play built on Red Hat OpenShift, OpenShift AI, and Ansible, IBM is well-positioned to capture executive buy-in by de-risking agentic AI software development. However, IBM must extend its focus to also incorporate critical languages, such as Python, TypeScript, and Rust. This effort may receive a boost once the company concludes its acquisition of DataStax, maker of the popular AI app orchestration framework, LangFlow.
You can read more about IBM’s agentic AI and data-specific announcements on the company’s website.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Other insights from Futurum:
IBM Posts Strong Q1 FY 2025 Results, Backed by Automation and Hybrid Cloud
Microsoft Advances Enterprise AI With New Copilot Agents and Collab Features
Oracle Database Analyst Summit 2025: Oracle Raises its Database Game
IBM Places Bet on Efficiency with New Granite 3.0 Foundation Model Family
Image Credit: IBM
Author Information
Nick is VP and Practice Lead for AI at The Futurum Group. Nick is a thought leader on the development, deployment and adoption of AI - an area he has been researching for 25 years. Prior to Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, with responsibility for 451 Research’s coverage of Data, AI, Analytics, Information Security and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.
Brad Shimmin is Vice President and Practice Lead, Data and Analytics at Futurum. He provides strategic direction and market analysis to help organizations maximize their investments in data and analytics. Currently, Brad is focused on helping companies establish an AI-first data strategy.
With over 30 years of experience in enterprise IT and emerging technologies, Brad is a distinguished thought leader specializing in data, analytics, artificial intelligence, and enterprise software development. Consulting with Fortune 100 vendors, Brad specializes in industry thought leadership, worldwide market analysis, client development, and strategic advisory services.
Brad earned his Bachelor of Arts from Utah State University, where he graduated Magna Cum Laude. Brad lives in Longmeadow, MA, with his beautiful wife and far too many LEGO sets.
