As a seasoned advocate and expert in cloud computing and generative AI, I’ve observed the immense transformative potential these technologies offer. Yet, we’re doing things just as stupidly as we did in the early days of cloud computing.
If you have not noticed lately, enterprises are running around in circles to fix mistakes they made 10 years ago in migrating and building new cloud-based systems. Repatriation is shorthand for “whoops!” The lack of planning and understanding has led to huge bills that nobody expected, and CIOs are attempting to mitigate. This means instead of focusing on innovation, we’re looping back to fix things after the fact.
Functional failure
Generative AI is not just a shiny new tool; it demands a radical rethinking of architectural principles. Current trends show a heavy reliance on reusing and repurposing old strategies without substantial innovation. This approach may offer short-term convenience but leads to long-term stagnation and, more often, “functional failure.”
I use the term “functional failure” since all of these poorly planned architectures work, but they cost between 2 and 10 times more than they should. Moreover, they return no value (or return negative value) to the business. This has been a common scenario in the world of cloud computing. It is generally a huge resource drain and hugely annoying if you work in these IT groups, but it likely won’t kill the business.
AI systems are a lot different. For most businesses they have the potential to be a huge innovative differentiator, meaning the business becomes that AI system. One example is intelligent supply chain systems that are being built around AI for some industry players; they are able to leverage this technology to build things faster and cheaper, and, at the same time, provide a better customer experience. Although most businesses will tell you that they do that already, most don’t, and a business that figures that out will disrupt its market like Uber and Netflix did.
It’s a bespoke architecture, dummy!
Most generative AI systems crafted today resemble each other too closely, which is concerning since the businesses they aim to serve differ significantly and have specific needs. As I stated above, generative AI should be an innovative differentiator with unique solutions designed and built for your specific business use cases. But that’s not what’s being done.
When systems mirror each other, as they do with common frameworks, tools, and approaches, they fail to leverage their full potential and instead become costly liabilities. I dread hearing, “This is the way that this guy showed me at a conference.” Good architecture is designed for a very specific use case. The chances you can replicate what you saw at a conference to a value-delivering ending are nil.
The architect working for a specific cloud provider in Hall C at the 2:15 talk can only provide you with general patterns, most of which are not useful to your specific use cases and architecture. Sorry, there is no easy button for this. Even during my presentations, if I see people taking photos of my slides, I’m quick to remind them that the task is not to move through some static and reusable approach but to come up with unique and innovative solutions for your specific situation.
Success in generative AI depends on how well the technology aligns with a business’s unique needs. It is crucial to move away from one-size-fits-all solutions and to craft bespoke architectures that are robust and agile enough to evolve with changing business demands. Instead of pursuing rapid deployment for immediate gains, we need a shift towards thoughtful, strategic integration of AI technologies. Only then can we ensure systems provide long-lasting benefits and adapt to evolving needs.
Getting smart quickly
The role of a generative AI architect should go beyond merely applying existing technologies; it should involve pioneering new methodologies and pushing the boundaries of what’s possible. As leaders, we must foster a culture that not only encourages innovation but actively rewards it.
Are we questioning established norms and continuously seeking opportunities to improve and innovate? Are we blindly following other people’s approaches to completely different business problems? It’s time to stop imitating architectural processes from hyperscaler conferences or reusing frameworks, spreadsheets, and slides developed for another project by whatever consulting firm. You need to get smart quickly and stop copying off other people’s papers.
The journey toward exceptional generative AI architecture for use in or out of the cloud is challenging yet crucial. It requires a break from tradition, a commitment to deep customization, and a resolve to innovate. I wish I could tell you this is easy, but we’re about to embark on building core IT systems that will define the business’s value. Get it wrong, and the business is likely to be displaced. No pressure.