Do your homework before signing any AI solution contract with your vendors. We saw this exact playbook with RPA a decade ago — where up to 50% of initial projects failed, yet vendors kept selling and companies kept buying.
AI is genuinely transformative. I use it every single day and it has become as normal to me as opening a browser. The technology is real, the capability is real, and the value is absolutely there for businesses that approach it properly.
The problem is not AI. The problem is vendors exploiting the hype around it.
Right now, vendors are packaging up tools you already have direct access to, adding a layer of complexity and a healthy margin on top, then selling it back to you as something only they can deliver. McKinsey surveys consistently show that while AI adoption is now above 50 percent across organisations, only around a third report meaningful financial impact. That gap exists because most businesses are still experimenting without a clear strategy — and that is exactly the environment vendors thrive in.
Key Insight
Platforms such as OpenAI, Anthropic, Google, Microsoft, and Amazon are already directly accessible to you and your team today. Many businesses are already paying for one or more of them without realising the full extent of what they can do.
The primary barrier to getting real value from AI is rarely the technology itself. It is usually a lack of process clarity, data readiness, and well-defined use cases. That is internal work no vendor can do for you.
Where to Start
Document and genuinely understand your own processes, services, and where the real friction lives.
Identify the problems worth solving, then brainstorm how AI can address them in practical and measurable ways.
In most cases, the AI subscriptions you already have are more than capable of getting you there.
If a vendor is involved, demand clarity on what you are actually getting, what it is built on, and whether it solves a problem that genuinely exists in your business.
Some businesses may benefit from vendor expertise, and there is nothing wrong with that — but go in with your eyes open. If they cannot answer what they are building on top of that you cannot access yourself, you already have your answer.
Before you hand over the budget, ask one simple question:
"What exactly are you building on top of that I cannot access and use myself?"
Have you been pitched an AI solution recently that turned out to be nothing more than a wrapper around a tool you already had access to?
AI Vendor Buyer's Guide — FAQ
Common questions from people researching enterprise AI vendor products before signing a contract.
How do I know if an enterprise AI vendor product is worth the money?
Start by asking the vendor exactly what their product is built on top of. If the core capability comes from OpenAI, Anthropic, Microsoft, Google or AWS — platforms you can access directly — you need a clear answer on what specific value the vendor's layer adds beyond what you could configure yourself. If they cannot articulate that clearly, you have your answer.
Are enterprise AI vendor products often just wrappers around tools I already pay for?
Many are. Vendors have a strong incentive to package generally available AI capabilities behind branded interfaces and consulting margins. That is not automatically bad — there can be real integration work involved — but you should always verify what is proprietary versus what is a thin layer over a commercial API you could call directly.
Do I need a vendor or consultancy to deploy AI in my business?
Usually not to get started. The hardest part of AI adoption is internal work: understanding your own processes, identifying genuine friction, and defining measurable use cases. Those are things no vendor can do for you. The AI subscriptions your team likely already has can deliver meaningful value once that groundwork is done.
What questions should I ask before buying an enterprise AI platform?
Ask what it is built on, which of those components you could access directly, what specific problem it solves in your business, how success will be measured, what data leaves your environment, and what happens at renewal. If the answers lean on buzzwords rather than specifics, treat that as a warning sign.
Why do so many enterprise AI projects fail to deliver ROI?
Most fail for the same reason early RPA projects failed a decade ago — organisations bought technology before defining the problem. McKinsey consistently finds that while AI adoption is widespread, only around a third of organisations report meaningful financial impact. The gap is almost always process clarity and use case discipline, not the tooling.