In January of this year, IBM published a detailed research report explaining why digital transformation only provided an ROI of -5 to 10 percent, rather than the projected 150 percent! That's a huge gap, consistent with our experience with the initial client/server implementation in the 80s, the OS migration in the 90s, the big data implementation in the early 21st century, and the technical implementations analyzed over the past decade.
This is not to say that all technical implementations fail to meet their goals, only that the vast majority fail. The recurring problem is that the technology is not mature enough, with the companies providing it, and the companies deploying it, using impressive sales efforts to convince buyers of its benefits, while lacking follow-up to ensure that the promised value has been realized.
The same thing can happen with artificial intelligence.
When the client/server trend hit, the technology wasn't ready at the time, but it almost bankrupted IBM as the market tried to move to something that wasn't ready.
The reason for this is that sales often go far beyond where the product is located, far beyond the services surrounding the product. When a new trend like artificial intelligence hits, everyone wants a piece of the action. But apart from companies like IBM and Nvidia, which have been working on AI for decades, no one (including Google) has declared it ready for AI.
The reason IBM is so excited this time is that it has watsonx, which is the most mature AI solution on the market today. This time around, IBM is the most mature in terms of enterprise-grade generative AI capabilities, while others survive on little or no foundational sales and marketing commitments.
When sales are ahead of technology, buyers lose. According to IBM data on digital transformation, a lot of people suffer because they didn't do their due diligence.
Solution: Do your homework and follow the process
One of the most successful processes in this case is "from trial to production". Don't deploy in one go. Once you're confident that a vendor has provided you with a mature and complete solution, create one or more pilot projects to validate your conviction. Even a mature product may not be suitable for all situations. You don't want to fail at scale, but failure in a pilot is recoverable and bearable.
But even before the pilot, you need to make sure that the supplier's requirements for revenue and ROI are achievable. Seek references from others who have already deployed and realized the promised benefits, ask vendors to see if they already have the technology deployed on-premises, and ask to speak with the IT staff of the company that uses the technology (they will usually be very honest).
Research and get best practices with others who are trying the same task, and realize that not every solution will work for every company or even every department.
Hybrid multicloud is the practice of providing the best balance between uptime, cost, availability, and reliability. It takes a vendor who understands the concept, has a deep relationship with a cloud provider you trust, and has gained enough experience that it shouldn't be learned in the process of serving you.
Especially when it comes to AI data, quality is crucial, and you need a lot of help to ensure it. Don't want an AI that's biased or hallucinating, just like you don't want an analysis that always provides inaccurate answers.
These new AI capabilities are expected to be multimodal, including natural language, images, audio, video, and even key temporal elements. The use of AI tends to be optimized for one data type and not as well as the others, so you need to understand the differences and let the vendor know that in areas where it doesn't have the capabilities, another vendor might be a better choice.
Finally, you need help with metrics and milestones so that if a vendor is underperforming, you can identify the problem early and either change vendors or teams. If the vendor you're working with can't help you set metrics and goals for your project, you're in the wrong place.
From the client/server of the 80s of the 20th century to today's artificial intelligence, the problem we've often encountered in recent big technologies is that sales far outstrip products and support structures. As a result, deployments don't meet goals and expectations. In many cases, it's wiser to wait until the right partner, the right team, and the right solution are available.