Operations Calling is Tulip’s flagship event for the global manufacturing community — a space where leaders, engineers, and innovators come together to exchange ideas and see the future of operations unfold. The 2025 edition gathered more than 750 attendees from across industries for two days of keynotes, workshops, and live demos exploring how AI, composability, and human ingenuity are defining a new era of connected, intelligent operations.
One of the event’s most thought-provoking sessions, Breaking the Loop: AI Beyond the Hype, brought together three distinct perspectives on the state of artificial intelligence in manufacturing and beyond. The panel featured Dr. Pattie Maes, Professor at the MIT Media Lab and pioneer in human-computer interaction; David Rogers, Senior Solutions Architect at Databricks; and Ashtad Engineer, Worldwide Head of Manufacturing, Supply Chain, and Sustainability Solutions at AWS.
Together, they explored the promises and perils of AI’s rapid evolution — from its potential to drive industrial productivity to the responsibilities that come with deploying it at scale. What emerged was a candid, cross-disciplinary discussion that looked beyond short-term excitement to the deeper, structural questions shaping the field: What does meaningful progress look like? How do we balance autonomy with oversight? And how can the next generation of AI amplify human capability rather than obscure it?
The conversation began by reexamining a question that still defies easy definition: what exactly is “industrial AI”? For Ashtad Engineer, it’s not a buzzword but a category of applied intelligence shaped by the constraints of operations. In manufacturing, data comes from physical systems, not just digital platforms — and the challenge lies in transforming that operational data into context-rich insight without compromising safety or control.
Dr. Maes offered a broader lens, framing industrial AI as part of a continuum that extends from perception to reasoning to action. Where traditional automation focused on rule-based execution, modern AI enables systems that learn, anticipate, and adapt. “We’ve been talking about agents for decades,” she noted, “but now we have the computational and contextual capabilities to make them truly collaborative.”
Despite their differing backgrounds, both speakers converged on a shared idea: industrial AI is defined not by model size or processing power but by purpose. Its value lies in improving decisions, not replacing them — and in doing so, bringing human and machine intelligence closer together.
If AI represents the mind of modern industry, agentic systems are its emerging nervous system. These software entities can perceive data, reason over context, and execute tasks semi-independently. But as the panelists pointed out, the reality of today’s agents remains more pragmatic than the hype suggests.
Ashtad described current industrial agents as advisory rather than autonomous, supporting teams in defined workflows such as quality inspection, data validation, and system orchestration. Their power lies in speed and consistency — surfacing the right information to the right person at the right time — not in replacing human operators.
Dr. Maes traced the concept’s roots back to early AI research in the 1990s, when “software agents” were envisioned as digital assistants capable of learning user preferences. Today’s generative and agentic tools finally deliver on part of that promise, but their autonomy remains bounded by trust and governance. Real progress, she argued, won’t come from building fully independent systems but from creating ones that understand context and collaborate responsibly.
As AI systems become more complex, the conversation inevitably turns to trust. For industries where safety and validation are essential, blind confidence in algorithmic output isn’t an option.
Dr. Maes illustrated the point with an anecdote from a recent experiment by Anthropic. Researchers had tasked an AI-powered vending machine with deciding which snacks to dispense based on user input. The system performed well at first — until it began interpreting requests creatively, producing unpredictable outcomes. The takeaway, she explained, wasn’t about malfunction but about misunderstanding. Even well-trained models can misinterpret context, and that unpredictability is unacceptable in high-stakes environments.
Ashtad expanded on this theme from an industrial perspective. In manufacturing, every decision must be traceable and reproducible. Validation frameworks, audit trails, and “explainability by design” are what enable AI to scale safely. “Trust isn’t built through marketing,” he emphasized, “it’s earned through transparency.”
The panel agreed that progress in AI must be measured as much by reliability as by capability. Governance and human oversight aren’t barriers to innovation — they’re what make innovation sustainable.
The conversation concluded with a look toward the next phase of AI’s evolution. Ashtad predicted a future where software architecture becomes modular, built around interoperable agents that sit between systems of record and user interfaces. These agents won’t just automate workflows; they’ll coordinate them — acting as connective tissue between humans, machines, and data systems.
Dr. Maes added that for this future to take shape responsibly, openness and collaboration are critical. Proprietary silos slow innovation and concentrate risk. The next wave of progress, she argued, will depend on shared standards and transparent design. “AI doesn’t live in isolation,” she noted. “Its behavior is a reflection of the systems and the people who build it.”
Rogers closed the discussion by returning to the session’s central theme: moving beyond hype. Each speaker agreed that the future of AI in manufacturing isn’t about limitless autonomy or abstract intelligence — it’s about meaningful, measurable outcomes. The factories of tomorrow won’t just be more connected; they’ll be more comprehensible, blending data-driven insight with human intent.
Breaking the Loop captured the tension and optimism defining AI in 2025. It was a reminder that while the technology continues to evolve rapidly, the questions surrounding it remain deeply human: trust, accountability, and purpose.
The panel’s message was clear. AI’s potential to transform operations is immense, but realizing it requires balance — between innovation and restraint, autonomy and oversight, ambition and humility. True progress won’t come from escaping the human loop, but from strengthening it.
🎥 Watch this session and all other Operations Calling 2025 presentations on demand.
