Research Finder
Find by Keyword
AllThingsOpen.ai: Applied Intelligence Takes Center Stage
Applied AI, productivity, security, and IBM’s open source contributions were highlighted at the Durham, NC conference.
Analyst Take:
The AllThingsOpen.ai conference on March 17-18, 2025 showed a clear trend: the hype around AI is giving way to applied intelligence. The discussions were not about theoretical possibilities but concrete use cases and showed how AI can be a valuable tool across many types of functions. We saw the typical AI software usage patterns, but also AI for grant writing, drug testing, and even local government, among others.
The presenters were in agreement that AI augments human work to be faster or better, but cannot fully replace it. This year’s conference highlighted AI’s potential to enhance productivity, personalize user experiences, and streamline workflows. AI agents in particular were heavily discussed. However, topics like monitoring, securing, and hacking AI were also featured as the reality of actually deploying and utilizing AI takes hold.
The event benefited from the enthusiastic community atmosphere and the gorgeous setting in the historic Carolina Theater in Durham, North Carolina where Elvis Presley once performed. The strong presence of organizations like IBM, Red Hat, SAS, Acorn, CodeGate, and Auth0 was evidence of the broad interest and engagement in practical AI applications.
IBM’s Open Source Contributions
IBM’s announcement of adding DocLing, BeeAI, and Data Prep Kit to the Linux Foundation reinforces its commitment to open source AI development. These projects aim to deliver tools for multi-agent systems, document understanding, and data preparation. The addition of the projects makes AI technologies more accessible and collaborative. We see IBM’s open source push as promoting innovation and wider adoption of AI capabilities.
AI and Productivity: Augmentation, Not Replacement
A recurring point throughout the conference was that AI enhances productivity. Over and over the presenters talked about AI as a tool to augment human capabilities, not replace them. The discussions frequently demonstrated human-in-the-loop workflows along with the need for human oversight to ensure useful output. The conference also highlighted the importance of prompt engineering, though it’s still up for debate whether shorter, well structured or longer, more detailed prompts yield better results from AI models. However, the use of AI detection for content production is an ongoing issue, with some saying that AI generated content is acceptable while other organizations outright reject any content that is flagged as AI generated, whether the flag is correct or not.
AI Agents and Workflow Automation
AI agents are moving beyond simple tasks to handle more complex workflows. The conference showcased how agents can autonomously research, write, edit, and structure content. Multi-step AI workflows can break down complex tasks into manageable steps and AI assistants can handle scheduling, email processing, and report generation. One of the highlights was a demonstration of Obot, Acorn’s open source AI agent platform that allows users to build, deploy, and share AI agents across their organization. We got to see how straightforward it is to create an Obot AI agent on the platform that automates email responses to specific questions, among other examples.
Building Responsible AI Systems
Security and trust emerged as critical considerations. The need to protect sensitive credentials during AI interactions was emphasized, along with the recommendation for enterprises to self-host AI models when dealing with proprietary data. CodeGate’s solution of a “guardian angel” that is placed between AI coding assistants and LLMs was particularly interesting in this space as many developers and organizations may not realize the potential for AI assistants to send sensitive information to shared LLMs.
The session on Hacking AI was eye-opening. There are many emerging techniques that can wreak havoc on AI systems, such as data poisoning, extraction, and evasion, that often get overlooked within all the AI hype. We watched how adding a few pixels to an image, undetectable to the human eye, can make the image completely unrecognizable to AI. Enterprises and the AI community at large will need to get serious about defending sensitive data from AI hacking.
LLM Selection and the MUX Factor
The discussions around the use of multiplexors (MUX) for selection between different LLMs revealed another trend. This indicates a growing awareness that no single LLM is going to be universally optimal. Enterprises are beginning to explore how to dynamically route tasks to the most appropriate LLM based on specific requirements and use cases. The focus on Model Context Protocol (MCP) also points towards the need to improve how LLMs interact with and interpret data from diverse sources. MCP is an open standard from Anthropic that provides a standard way to connect AI models with different data sources and applications. We see this as a key development as businesses seek to integrate AI into their existing ecosystems.
Looking Ahead
The AllThingsOpen.ai conference provided a clear snapshot of an AI landscape transitioning from theoretical exploration to practical application. The diverse applications showcased, from grant writing to drug testing, indicate a widening acceptance of AI as a tool across sectors. At HyperFRAME Research we see the industry now grappling with the realities of AI deployment, security, trust, and optimizing LLM selection. This indicates both the widespread adoption and maturing of AI technologies.
The emphasis on AI Agents, exemplified by Acorn’s Obot platform and IBM’s BeeAI, suggests a growing interest in further automating complex workflows. However, the concurrent focus on AI hacking and security solutions like CodeGate highlights the inherent risks associated with these powerful tools. The discussion around MUX and MCP are indicative of a move towards a more sophisticated and interconnected AI ecosystem. HyperFRAME Research predicts a need for more “AI middleware” for LLM routing and augmentation in the future.
The growing awareness of AI hacking and security threats will likely drive significant investment in this area, and HyperFRAME Research expects to see the emergence of new standards and best practices for secure AI deployment. The conversation around AI detection and content creation is going to be interesting to watch, as it shows that even with all of the progress in this space, there is still a lot of debate about the best way to utilize AI in a business environment. All in all, the AllThingsOpen.ai conference provided a valuable and engaging experience, marked by impressive presentations, higher than expected attendance, and a highly enthusiastic audience.
Stephanie Walter | Analyst In Residence - AI Tech Stack
Stephanie Walter is a results-driven technology executive and analyst in residence with over 20 years leading innovation in Cloud, SaaS, Middleware, Data, and AI. She has guided product life cycles from concept to go-to-market in both senior roles at IBM and fractional executive capacities, blending engineering expertise with business strategy and market insights. From software engineering and architecture to executive product management, Stephanie has driven large-scale transformations, developed technical talent, and solved complex challenges across startup, growth-stage, and enterprise environments.