Mark Bateman, Vice President Sales, Global Channels – Spirent says AI adoption isn’t just about acquiring the latest tech—it’s about overhauling infrastructure, energy strategies, and data foundations. In 2025, vendors must move beyond selling products and solutions and start guiding customers through the complex journey of transformation.
New technologies often create a kind of clumsy excitement. Whether it's the cloud, the IoT, or now, Artificial Intelligence (AI), these technologies are acquired eagerly and often without consideration as to what might be needed to make them work. While the excitement around AI has started a flurry of investment and adoption, acquiring a complex new technology is rarely as simple as buying it off the shelf. In fact, as we’ve seen with these kinds of new technologies, post-acquisition problems are common and often serious. While transformative, the early track records of both the IoT and the Cloud were marred by repeated security and technological failures, while AI technologies are proving haphazard, difficult and prohibitively expensive to successfully adopt.
This won’t just be a learning curve for customers, but vendors of AI solutions too. Customers will need to accompany them while they go through the complex journey of AI adoption. This year, we’ll see AI solution vendors being called upon to become consultants too.
AI-sier said than done
Whatever AI can really do for a business, acquiring those capabilities is far more complex than simply bolting them onto an enterprise. In fact, the path to acquiring or developing an AI is a long and winding road, beset with barriers and potholes for even the largest, best-resourced actors. Vendors will need to help their clients anticipate and circumvent these.
Energy
Energy consumption, for example, is a particularly large stumbling block for many AI hopefuls. AI uses an incredible amount of energy. The International Energy Agency (IEA) estimates this technology alone accounts for around 2% of global consumption, and will likely double in the next few years.
For individual enterprises, that will amount to extremely large energy bills at the very least. On a larger scale, however, it may become an even more problematic factor as companies look for data centers that can accommodate their energy demands. Many areas simply can’t draw the power required for AI data centers and a report from the North American Electric Reliability Corporation (NERC) says that those data centers are so power hungry that they may start causing outages across the US and Canada.
The same goes for access to water and cooling. This makes up a significant part of energy expenditure in itself and is a critical component in an AI data center, but one that many are ill-equipped to provide.
“Garbage in, Garbage out.”
On top of that, there’s the other crucial resource that AI depends on - Data. A sufficient amount of good quality, performant training data is perhaps the most important asset an AI draws upon. Without it, or with a sub-standard or insufficient amount of it, an AI model won’t be able to learn or make decisions properly. In fact, an AI model trained on bad data will likely produce bad results. In fact, AI companies are now even having problems with finding enough data to feed their models. One research paper even predicts that if AI models training dataset requirements keep growing, the supply will literally run out between 2026 and 2032.
Building an AI model will require companies to configure their infrastructure for it. As such, there are many areas that stand to gain hugely from AI but are woefully unprepared for it. Take the financial services sector, for example. This sector plans to use AI to deal with the ocean of paperwork and compliance obligations that such institutions are nigh-overwhelmed with. Yet - like many other industries - they won’t be able to merely bolt an AI model onto their existing infrastructure. Instead, they’ll need to redesign their most basic data collection processes, break open their departmental silos and restructure fundamental working practices. Not only will clients need to know that, but they might want to be shown how to do it too.
Without factoring in the myriad factors that could hamstring an AI project, costs will spiral and delivery times will extend uncontrollably. AI projects will suffer delays or ultimate failure. Any attempt at acquiring AI capabilities without careful planning and consideration is sure to stall - leading to spiraling costs and expenditures - or failure.
Vendors need to become partners
For vendors, they can’t simply sell a product and leave the customers to figure out how to use it. In fact, customers will likely demand a more consultative approach from their vendors, as AI becomes the competitive differentiator across a variety of sectors. Customers won’t just want vendors, they’ll want partners too and that will mean being able to understand the unique risks, problems and potential gains that individual customers will deal with in their attempts to acquire AI capabilities. That will mean real understanding of a customer’s current infrastructure and readiness to onboard new technologies; insight into their compliance obligations and risk appetites; understanding what they want to achieve with AI and how they’ll go about achieving it.
In 2025, it won’t be enough to simply be a vendor. Companies that want to achieve will need partners to help them realise their AI ambitions. Rigorous and thorough testing will provide a way for both vendors and customers to understand those facts more clearly, and confidently embark on the road towards AI adoption.
