The short version
Every organisation has an AI champion, and they are usually right about the destination. The risk is the timeline. Hardware is a five to seven year capital commitment and most AI strategies driving those decisions are still being written. Before any significant spend, four questions need honest answers:
- What will run on it?
- Who owns it?
- What is already running that shouldn’t be?
- What does ‘AI-ready’ mean in your context?
The right partner runs that conversation before talking about hardware, even when the answer costs them the deal this quarter.
Why is everyone rushing to buy new hardware right now?
Because every organisation has an AI champion and the champion has momentum that has become difficult to question.
Sometimes it is someone in the C-suite. Sometimes it is a head of transformation, who came back from a conference in January, absolutely on fire. Sometimes it is someone in the business who has quietly been using AI tools on their own for six months and has become the most informed person in the building on the subject.
They are not wrong to push. The organisations that treat AI as optional are going to feel that eventually. But here is the pattern I keep seeing in IT teams:
The enthusiast has momentum. They have presented to the board. They have used phrases like AI-ready infrastructure and compute capacity. The board has nodded and IT has been asked to make it happen.
So IT, caught between a strategy they were not fully involved in writing and, a budget conversation they were not fully involved in framing, starts building a hardware case under time pressure, with incomplete requirements. Yet nobody in that process pauses long enough to ask the questions that actually matters.
You are not buying a server. You are making a long-term financial commitment based on a strategy that, in most organisations, is still being written.
What should you know before investing in new infrastructure?
That the gap between ‘we need to be AI-ready’ and a specific, owned, sized workload is enormous, and that gap is exactly where the wrong decision lives.
Most enterprise AI use cases in mid-market and public sector organisations right now are still being identified. The strategy is live but not finished. The ownership question, whether this is an IT project or a business project, has rarely been properly resolved. The use cases have not been validated and the workloads have not been sized.
None of that is a reason to do nothing. It is a reason to answer four questions honestly before committing capital. In most organisations, at least two will not have a clean answer yet.
The key questions to ask before making any hardware decision:
- What specifically will run on this infrastructure, at what scale and by when?
- Who owns AI inside the organisation, IT or the business and has that been agreed?
- What is already running on existing platforms that does not need to be there?
- What does ‘AI-ready’ mean in your specific context, not in vendor language?
If those four have solid answers, you probably have a real investment case and you should move. If they do not, the momentum is ahead of the strategy and a few months of work now will save a platform decision you regret for the rest of the decade.
What does a good pre-investment conversation look like?
It starts with what you are trying to achieve and what you already have, and it does not mention a product until those are clear.
A good conversation maps the workloads the organisation actually wants to run. It surfaces what is already running on existing platforms that quietly does not need to be and settles the ownership question between IT and the business. It defines what success looks like in eighteen months in language the people who have to live with the decision would recognise.
Only then does hardware enter the discussion, because only then can it be sized against something real, rather than against a slide that says AI-ready. This is not anti-AI. We believe in it and we help customers build towards it every week. The most useful thing a partner can do for a lot of organisations right now is not point them at new compute. It is help them slow down long enough to answer those questions first.
That is why we run an assessment conversation before we talk about hardware. The output is sometimes a clear investment case and sometimes it is: not yet, here is what to do first, come back in six months.
When is the right time to invest?
When the workload is specific, the owner is named and the adoption plan exists. Not when the deadline arrives.
The best infrastructure decisions I see are made slowly. The worst ones have a deadline set by someone who presented to the board in January. The difference is rarely the quality of the IT team. It is whether anyone created the space to ask the hard questions before the spend was already politically committed.
Saying ‘not yet’ occasionally costs us a project in the short term. It is also the reason customers come back for the next five years. A partner who only ever tells you to buy is not protecting your budget, they are protecting their quarter. The honest position costs money sometimes. It should.
The second outcome, ‘not yet, here is what to do first’, does not help this quarter’s number. It is exactly why the relationship lasts.
So none of this is a reason to slow your ambition. It is a reason to make sure the ambition and the spending decision are actually the same decision, owned by the same people, pointed at the same thing.
Who is the AI champion in your organisation, and has anyone asked them the hard questions yet?
Frequently Asked Questions
Should we buy AI infrastructure now or wait?
Buy when you can answer what will run on the infrastructure, at what scale and by when. If those answers are still being worked out, a five to seven year financial commitment is premature. The cost of waiting a few months is almost always lower than the cost of buying the wrong platform, for a strategy that was still being written.
Who should own AI in our organisation, IT or the business?
Both, but the split has to be agreed before money is spent. The business owns the use cases, the adoption and the outcomes. IT owns the platform, the security and the integration. Most stalled AI programmes fail because nobody agreed which side owns what, so the infrastructure gets bought without an owner for the result.
What does AI-ready infrastructure actually mean?
AI-ready is not a product specification. It is the point at which you have validated workloads, a sized requirement, an owner and an adoption plan. Vendor language describes capacity and compute. A useful definition describes what you are going to run, who runs it and what good looks like in eighteen months.
How do we know if our AI champion is moving too quickly?
The warning sign is momentum without specifics. If the deadline is set, the board has nodded and the budget conversation has started, but nobody can describe the specific workload, its scale, or its owner – the momentum is ahead of the strategy. That is the moment to slow down, not to sign.
What does a pre-investment assessment involve?
A structured conversation before any hardware is discussed. It maps what the organisation actually wants to run, what is already running that does not need to be, who owns the AI agenda and what adoption looks like in practice. The output is either a clear investment case or a recommendation to address something else first.
Written by…
Andy Smith
Chief Technology Officer
Andy Smith is CTO at Proact IT UK, a specialist IT managed services provider in enterprise storage, hybrid cloud, and data infrastructure, since 1994. Andy leads Proact IT UK’s commercial and technology strategy across Storage, Data & AI, Hybrid Cloud, Operational Resilience, Modern Work and Managed Services, working with IT leaders across enterprise and public sector organisations on AI and infrastructure investment decisions.