AI is everywhere right now — and if you’re running an MSP, you’re feeling it from every direction.

Your PSA has AI. Your RMM has AI. Your documentation platform has AI. Your quoting tool probably has AI. At this point, we’re all just waiting for the coffee maker to start summarizing tickets.

But here’s the problem: not all AI is useful. A lot of it is just… noise.

In this Get NIST-y episode, Jared and Mike dig into how to tell the difference between AI that actually helps your business and “AI slop” that introduces risk, cost, and confusion. This article is a roundup of what we discussed. For all the details, use cases, and commentary, stream the episode above!

The Core Question: What Problem Does It Solve?

The easiest way to cut through the hype is also the most overlooked:

What problem is this actually solving?

There are two legitimate categories where AI tends to shine:

• It saves meaningful time on something you already do.

• It enables something you realistically wouldn’t do otherwise.

That second category is where some of the most valuable use cases live. Think AI notetakers that automatically capture and summarize meetings — something most teams know they should do, but rarely execute consistently.

If the feature doesn’t clearly fit into one of those buckets, it’s worth asking whether it belongs in your stack at all.

Because “it has AI” is not a use case.

The Marketing Pressure Problem

A big driver behind AI overload isn’t technical — it’s marketing.

Vendors feel like they can’t show up to the conversation without an AI story. That pressure leads to features that exist more to check a box than to deliver value.

The result? Tools that:

• Add complexity without reducing workload

• Introduce risk without clear controls

• Sound impressive in demos but fall flat in production

This is how AI becomes slop: not because the technology is bad, but because the implementation is shallow.

Before You Let AI Touch Client Data

This is where things get serious.

If your team is using AI tools that interact with client environments, documentation, or tickets, you need clear answers to a few critical questions:

• Where is the data going?

• Is it being stored, and if so, for how long?

• Is it used for training models?

• What controls exist around access and retention?

• Does the contract explicitly define any of this?

If you don’t know where your data is going, you can’t protect it. It’s that simple.

And right now, many vendors are not keeping their contracts or disclosures in sync with their AI features. That gap creates real AI-driven risk for MSPs who assume coverage that isn’t actually there.

The Cost Curve Is Coming

Another factor that’s easy to ignore today: cost.

We’re currently in a heavily subsidized phase of AI adoption. Token costs are low, features are bundled, and pricing often doesn’t reflect the true underlying expense.

That will not last.

As those subsidies disappear, expect:

• AI features to become add-ons

• Usage-based pricing to increase

• Vendors to reevaluate what’s included vs. premium

If an AI feature only makes sense when it’s cheap, it may not make sense for long.

The Real Takeaway

AI isn’t the problem. Thoughtless adoption is.

Use AI where it delivers clear, measurable value. Be skeptical of features that exist only to signal innovation. And most importantly, treat client data like what it is: your responsibility.

Because no matter how useful a tool claims to be, if you don’t understand what it’s doing with your data, you’re the one taking the risk.