Whatever happens in the investment markets over the next few months will not signal the demise of AI. The bursting of the AI investment bubble, which now seems to be a self-prophesying probability, may well pull the plug, for now, on many of the huge, planned AI infrastructure investments that have been announced in the past year. Nervousness in the markets that has triggered intense volatility in tech stocks, also appears to be leading to some nervousness and the delaying of investment decisions in the wider business community, already spooked by what has been deemed as a business unfriendly budget. This could lead to a slowdown in GenAI and Agentic AI activity, which backed up by an MIT study entitled, “The GenAI Divide – STATE OF AI IN BUSINESS 2025” that shows 95% of pilots failing to move into production and future plans being delayed or quietly shelved.
But, for IT teams, this could be a blessing in disguise. To date, siloed thinking — between teams, systems, and data — has become one of the quietest yet most persistent barriers to rapid and sustained progress in moving GenAI and Agentic AI pilots into widespread production. Potentially, while there is a pause in the headlong development and deployment of new AI capabilities, IT leaders across industry will have a unique window of opportunity to make their businesses ready to take advantage of the potential offered by AI, in all its forms.
For a number of years concepts like Minimum Viable Product (MVP), new software technologies such as containerisation, micro-services, and DevOps development practices, have all contributed to a sense that “speed is good”. This has worked well for “cloud native” start-ups who are not constrained by a complex web of legacy applications and sprawling business processes. But speed is still good for established organisations; nobody want to return to the era of multi-year software development projects where no actual application code was written in the first year (Note: back in the 1980s I witnessed that first hand). However, we are now starting to get first-hand accounts of problems being encountered in such areas as using AI coding. This LinkedIn post by Gregory Smith is particularly insightful, based on his hands-on experience of using AI coding. I know Gregory well and I can attest to his technical competency, systems thinking capability and his integrity. He found that, if you don’t think about the architecture up front you can end up with code that is hard to scale or maintain, creating new tech debt. The point he is making is really about the need to ensure that while speed is good, producing something that is sustainable is better.
Over the last year, Salesforce has spent a lot of time and effort into understanding how to develop and scale Agentic AI systems. This summary of their findings by Srinivas Tallapragada, its President and Chief Engineering and Customer Success Officer, is a useful primer for organisations looking to develop Agentic AI. Before you start, take heed of this comment from him. “The shift to agentic AI isn’t just technological — it’s also cultural, operational, and organizational. You can’t simply build an agent, ship it, and expect it to work at scale. Be under no illusion. AI will impact your business and your markets in profound ways, some of which we can’t even envision yet. There are still many questions about AI that need answering: are LLMs the way forward or will SLMs and new models take centre stage; is Artificial General Intelligence (AGI) ever going to be a realistic proposition; will brand new hyperscale AI data centres actually be needed or will processing increasingly move to the Edge. I aim to revisit these questions and more in future blog posts. Much of the change will require a complete reset, of your business models, your systems and your culture. You need to be getting your organisations AI ready now. The world won’t change on 1st January 2026, and you may not spot the signs of change straight away. But that is no reason for you to delay preparing for the reset now.
by Paul Bevan – Tech Industry Forum STAR Group advisory council member
Contact us, if you want to know more, at our contact us page, or email info@cloudindustryforum.org