Technology Refresh
3 conversational turns exploring this topic
Q: What's emerging vs declining in AI technologies this quarter?
Subject: The AI Hype Cycle is Eating Itself (and What to Do About It) - Ins7ghts
Forget "AI is changing everything." The speed at which AI trends are being born, hyped, and discarded is the real story. This quarter's data paints a stark picture: While AI in Healthcare gains momentum, Agentic AI fades from the conversation. Yes, you read that right. Something can be both emerging and declining simultaneously. Welcome to the AI hype cycle on hyperdrive.
Here's what's actually important, and what's just noise:
The Real Deals:
AI in Healthcare (Mentions: 62): This isn't just about chatbots diagnosing illnesses. We're seeing real progress in drug discovery, personalized medicine, and robotic surgery. Companies like PathAI (cancer diagnostics) and Recursion Pharmaceuticals (drug discovery) are leading the charge, and the regulatory landscape is (slowly) starting to catch up. This is a long-term trend with massive potential, if* we can solve the data privacy and ethical challenges.
Data Governance (Mentions: 22): Surprise! The boring stuff is becoming sexy again. All this AI needs clean, reliable, governed* data. The rise of AI is forcing companies to finally take data governance seriously. Expect to see more investment in tools like Collibra and Alation, and a surge in demand for data governance professionals. This is the unglamorous but essential foundation upon which all AI success is built.
- AI-Ready PCs (Mentions: 20): The quiet revolution. Forget cloud-only AI. The ability to run AI models locally on your laptop is a game-changer for privacy, latency, and accessibility. Intel, AMD, and NVIDIA are battling it out for dominance in this space. The real winner will be the developer ecosystem that builds compelling applications that leverage these new capabilities.
The Overhyped (or Already Fizzling):
- Agentic AI (Mentions: 39 Emerging, 39 Declining): This is the poster child for the AI hype cycle. The idea of AI agents autonomously performing complex tasks is seductive, but the reality is far more complex. While companies like Adept AI are making progress, the technology is still in its infancy. The current hype is unsustainable, and we're already seeing a backlash as people realize the limitations.
- Data Mesh (Mentions: 9 Declining): The promise of decentralized data ownership was appealing, but the execution has been messy. Many companies are finding that Data Mesh is too complex and requires too much organizational change. Expect to see a move towards more pragmatic, centralized approaches to data management.
My Hot Take:
The AI market is starting to resemble the crypto market: a constant stream of new coins (technologies) being launched, pumped, and dumped. The signal-to-noise ratio is abysmal. The real value lies in the application of AI to solve real-world problems, not in chasing the latest shiny object.
Advice for Data Professionals:
- Focus on the fundamentals: Data quality, data governance, and data literacy are more important than ever.
- Be skeptical of the hype: Don't get caught up in the latest buzzwords. Evaluate new technologies based on their potential to solve real business problems.
- Invest in skills that are transferrable: Learn how to work with data, regardless of the specific AI technology. Python, SQL, and cloud computing are good places to start.
- Become a translator: The biggest need in the AI space is for people who can bridge the gap between technical experts and business stakeholders.
The AI revolution is real, but it's not going to happen overnight. It's time to tune out the noise and focus on building a solid foundation for long-term success.
– Yves Mulkers
Q: What should we sunset?
Subject: Re: The AI Hype Cycle is Eating Itself (and What to Do About It) - Ins7ghts
You asked what to sunset. The answer isn't about killing projects (yet). It's about killing mindsets.
The biggest drag on data teams isn't technical debt; it's conceptual debt. We're clinging to ideas that were shiny last year but are now actively hindering progress.
Here's what needs to go:
1. The "GenAI Solves Everything" Delusion:
| Mindset | Why It's Toxic | What to Replace It With |
|---|---|---|
| "Just throw GenAI at the problem!" | Ignores fundamental data quality issues, misaligned incentives, and the fact that most problems aren't GenAI-shaped. | "What's the smallest possible intervention that solves the biggest pain point?" |
| "LLMs are the future of everything." | Overlooks the compute cost, hallucination risk, and the fact that most tasks don't require a 175B parameter model. | "Can we solve this with a well-tuned, smaller model, or even a clever rules-based system?" |
Hot Take: The decline in "Generative AI" and "LLMs" mentions isn't because they're useless. It's because reality is setting in. The real value is in applied AI, specifically in areas like healthcare (62 mentions this week). Stop chasing the shiny object and start building something useful.
2. The "Data Mesh is the Answer" Cult:
Data Mesh (declining mentions) promised decentralized data ownership. In reality, it often delivers decentralized chaos. Unless you're a massive, hyper-complex organization like Netflix (who, let's be honest, can afford to throw bodies at problems), the overhead of implementing a full-blown Data Mesh is rarely worth it.
Instead: Focus on federated governance. Let individual teams own their data, but establish clear, centralized standards for data quality, security, and discoverability. Think of it as "Data Mesh Lite."
3. The "Agentic AI" Obsession (with a caveat):
Agentic AI is both emerging and declining. Confusing, right? Here's why: The idea of AI agents autonomously solving complex problems is still compelling. But the reality is that we're nowhere near that level of sophistication.
Sunset: The expectation that Agentic AI will magically automate everything tomorrow.
Embrace: The components of Agentic AI. Focus on building smaller, more specialized agents that can handle specific tasks, like automated data cleaning or anomaly detection. Think of it as "micro-agents" rather than a single, all-powerful AI overlord.
Specific Advice:
Audit your "innovation" backlog: Kill any project that starts with "Let's use GenAI to..." and replace it with "Let's solve this specific problem* with the most appropriate tool."
Re-evaluate your data governance strategy: If you're still trying to force-fit Data Mesh, take a step back and ask if it's actually* solving your problems.
Invest in AI-Ready PCs (yes, really): The rise of "AI-Ready PCs" (20 mentions) signals a shift towards edge computing and on-device AI. This is where the real* efficiency gains will be found. Think about how you can leverage local processing power to reduce latency and improve data privacy.
Stop chasing the hype. Start building solutions. That's where the real value lies.
– Yves
Q: What should we adopt?
Subject: Re: The AI Hype Cycle is Eating Itself (and What to Do About It) - Ins7ghts
You asked what to adopt. Honestly? Less "shiny new object" and more "boring but essential."
The hype cycle is a vortex, not a wheel. What goes up, comes down...hard. This quarter's data proves it: While AI in Healthcare gains momentum, Agentic AI fades from the conversation. Notice anything odd? Agentic AI appears in both the emerging AND declining lists. That's not a typo. It's a symptom.
Here's the brutal truth: We're so busy chasing the next unicorn, we're tripping over the basics.
| Trend | Status | Why You Should (Actually) Care |
|---|---|---|
| AI in Healthcare | Emerging | This isn't just about chatbots diagnosing sniffles. Think personalized medicine, drug discovery, and robotic surgery. Real-world impact, real ROI. Companies like PathAI and Paige are leading the charge. Watch them closely. |
| Data Governance | Emerging | Sexy? No. Critical? Absolutely. All this AI wizardry is useless without clean, reliable data. Invest in your data lineage, metadata management, and access control. Collibra and Alation are your friends here. |
| AI-Ready PCs | Emerging | The shift to on-device AI processing is real. Think lower latency, improved privacy. This will change how we build and deploy AI applications. Intel, AMD, and Qualcomm are battling for dominance here. |
| Physical AI | Emerging | AI isn't just software anymore. It's robots, drones, and smart sensors interacting with the physical world. This is where the real disruption will happen. Boston Dynamics is the obvious example, but look closer at smaller players like RightHand Robotics. |
My Hot Take: Generative AI is plateauing. The low-hanging fruit has been picked. The real value lies in applying AI to specific domains and ensuring the underlying data is trustworthy. Stop chasing the next LLM and start building robust data pipelines.
What to Adopt (and How):
- Double Down on Data Governance: I'm serious. This is the unsexy foundation that everything else is built on. Invest in tooling, training, and, most importantly, a data-centric culture.
- Focus on Vertical AI: Forget generic AI solutions. Find the specific problems in your industry that AI can solve. Healthcare is a great example, but every sector has its own unique opportunities.
- Prepare for On-Device AI: The future is distributed. Start thinking about how you can leverage AI-Ready PCs and edge devices to improve performance and reduce costs.
Specific Advice:
- Audit your existing AI projects: Are they solving real problems, or just chasing hype? Be ruthless.
- Talk to your data team: What are their biggest pain points? Chances are, they're related to data quality and governance.
- Read "Data Quality: The Accuracy Dimension" by Jack Olson: Yes, it's old. But the fundamentals haven't changed.
Stop being distracted by the noise. Focus on the signal. The future of AI isn't about bigger models, it's about smarter applications and rock-solid data.
Yves