AI in the IT Channel: what vendors are getting right (and wrong)
The IT Channel has been full of AI messaging for the past year. Having sat through a lot of it, here's my honest assessment of who's getting it right.
If you've been to any vendor briefing, partner event, or channel conference in the last twelve months, you'll know that every vendor now has an AI story. Some of them are good. Most of them aren't.
I've been on the receiving end of a lot of this as someone who evaluates these tools and has to make credible recommendations to customers. Here's what I've found.
What vendors are getting wrong
Wrapping everything in "AI" without being specific about what it means
The worst category of vendor AI messaging is the vague promise. "Our platform uses AI to enhance security." "Our solution leverages AI for better insights." What AI? Doing what, specifically? Against what training data? How was it evaluated?
When I push vendors on this, about half of them reveal that they've bolted a basic ML model onto existing reporting outputs and called it AI. That's not necessarily wrong, but it's not what the marketing implied.
Ignoring the data quality problem
AI is only as good as the data it's trained on and operates on. I've seen vendors pitch AI-driven insights into customer environments where I know the underlying data is incomplete, inconsistent, or poorly governed. The AI doesn't fix that. It amplifies it.
Overpromising on automation
"Set it and forget it" AI automation is not where enterprise AI is in early 2024. The demos look great. The reality involves constant oversight, exception handling, and human judgment at every step where something non-standard happens. Which is constantly.
What vendors are getting right
Microsoft's Copilot approach (eventually)
Microsoft's M365 Copilot took a long time to go GA (it was November 2023 when it became widely available to enterprise customers) and was expensive from the start at $30/user/month on top of existing M365 licensing. But the product itself is coherent. The integration with Microsoft 365 Graph data is the right bet. The value of Copilot is that it has access to your actual organisational data, not generic internet knowledge.
The early feedback from organisations that adopted it is mixed but substantive. That's actually encouraging; it means people are using it for real work and encountering real limitations, not just doing demos.
Security vendors who are specific about what the AI does
The better security vendors can tell you exactly what their models are trained on, what the false positive rate is, how alerts are tuned, and what happens when the AI gets it wrong. That specificity is a good sign.
Point solutions over platform promises
I'm more impressed by vendors who say "our AI does this specific thing very well" than by vendors who promise AI transformation across their entire product surface. Specificity is a sign of maturity.
The channel opportunity that's being missed
The thing I keep returning to: the real AI opportunity for channel partners isn't reselling vendor AI products. It's helping customers build the governance, data quality, and change management foundations that make AI tools actually work.
Every enterprise AI deployment I've seen fail or underperform has done so because of data governance issues, unclear ownership, or lack of user adoption. Not because the AI technology was bad. Those are solvable problems. They're also problems that channel partners are well-positioned to help with.
That's a more interesting conversation than "have you seen our new AI dashboard?"
