I spent around twenty years in the UK IT channel, selling technology into large organisations through partners, resellers and vendors. It's a good industry full of good people. It also has a shape, and most buyers never see it.
Understanding that shape is the single most useful thing you can do before you buy AI.
How the chain works
Very little enterprise technology is bought the way buyers imagine: a need, a search, a careful comparison, a decision. Far more of it is sold. A vendor launches something. Partners are trained and incentivised to take it to market. Targets are set by quarter. Funding appears for pilots, proofs of concept and workshops, often free to the customer, because the vendor is paying to get the product in.
None of that is sinister. It's how new technology finds its first customers. But it means the conversation you're having is usually shaped by someone else's quarter, someone else's target and someone else's product roadmap.
Why AI makes it riskier
With most technology, the product is the product. A firewall is a firewall. With AI, the product is only half the outcome. The other half is your process, your data, your exceptions and who owns the result. No vendor can sell you that half, and the incentive chain isn't built to talk about it.
So the pilot gets funded, the demo goes well, and then it meets the real process. That's where projects stall, and it's why so many end up cancelled.
How to buy it on your terms
You don't need to avoid the channel. You need to know what to ask.
- Ask who is paying for what. If the pilot is free, someone is funding it. That's fine, but know who, and what they expect next.
- Agree the number before the demo. Decide what the process costs today and what success would look like. A demo with no baseline proves nothing.
- Ask for the evidence output, not the feature list. What record does the system leave of what it did, and could your head of risk read it?
- Talk to a customer in production, not in pilot. Ask them what broke and what it costs to run.
- Pay for discovery. A fixed-price look at your process, by someone who'll tell you to stop if the numbers don't stack up, is the best money you'll spend. Free discovery is usually a sales stage.
- Know your way out. Before you sign, know what leaving looks like: your data, your rules, your workflows.
What good sellers will do
The best people I worked with in the channel welcomed those questions. They wanted customers who'd still be customers in three years. If a partner gets uncomfortable when you ask how they're paid or what happens when it goes wrong, you've learned something important cheaply.
That's the point of all this. AI will be sold to you either way. Make sure it's also bought.