AI as a Service is the biggest new revenue line for MSPs since managed security. But most MSPs are pricing it like they price everything else: a flat per-seat SaaS subscription. That model breaks down fast with AI, and it sets up the wrong conversation with your clients.  

Here's a better way to think about it, price it, and talk about it.  

The problem with per-seat pricing for AI  

A SaaS license costs the same whether an employee logs in once a month or lives in the tool. AI doesn't work that way. Consumption varies wildly by user, by workflow, and by month. If you sell AI at a flat rate, you either overprice it for light users or eat the cost of heavy ones. Either way, you're absorbing risk that should be shared, and your client never sees the connection between usage and value.  

AI is a utility, not a subscription  

Utilities are consumed when they create value. Nobody expects a fixed bill regardless of usage. They expect the bill to reflect the value they're getting.  

That's the mental model shift your SMB clients need, and analogies get them there fast:  

• Electricity. "When you turn on more lights or run more machinery, your electricity bill goes up. Nobody complains because they're getting more value."  

• Water. "Water your garden all summer or fill a pool, and your water bill increases. That's expected. You're using more of the service."  

• Natural gas. "A cold winter means a bigger heating bill. That's not a problem. You're paying for comfort."   

• Mobile data. "Stream videos all month and you'll use more data. You expect to pay for heavier usage because you're getting more from the service."  

• Printing. "Print 50 pages and it costs very little. Print 50,000 and the cost increases. Nobody expects unlimited printing."  

• Shipping. "Ship 1,000 orders instead of 100 and your shipping costs rise. That's simply a reflection of more business."  

• Air conditioning. "On a hot week you run the AC longer. The bill goes up, but you'd rather have the comfort."  

The strongest analogy: AI is electricity for knowledge work  

The more work your people do with AI, the more AI they consume. An employee who automates reports, writes proposals, summarizes meetings, analyzes tickets, and builds workflows all day will naturally consume more AI than someone who sends three emails. That's a sign they're getting value, not wasting it.  

When a client asks "Why did my bill go up?", the utility framing lets you answer: "Because your team is actually using it. Let's look at where."  

How to talk about spend with SMBs  

Here's the positioning that works:  

"Our goal isn't to make AI free. Our goal is to make AI predictable, visible, and governed. Just like any utility, you want to know who's consuming it, what they're using it for, and whether you're getting value from every dollar you spend."  

That shifts the conversation from "Why did my bill go up?" to "I'm glad people are actually using it." And it reframes your role as the MSP: you're not the vendor sending a surprise bill. You're the one who installed the meter, the breakers, and the thermostat.  

The message is, don't eliminate AI consumption. Eliminate unmanaged AI consumption.  

How Hatz makes the utility model work: Extra Usage  

A utility model only works if you can meter and govern it. That's exactly what Hatz's Extra Usage billing gives MSPs.  

Here's how it works, per the Hatz documentation: 

• Every tenant has a monthly credit allowance shared across users and AI features, so each SMB's baseline spend is defined up front.  

• MSP admins set the threshold. With Extra Usage enabled, you set a monthly cap per tenant, in dollars or credits, from the Tenants tab in admin.hatz.ai. Changes take effect immediately, and usage stops automatically when the cap is reached. Extra Usage is billed at the end of the billing period.  

• When a tenant hits 100% of its allowance, work doesn't stop. Hatz can provide limited additional capacity in a light mode: Auto Lite routes requests to lower-credit models suited to everyday tasks, while more credit-intensive models become temporarily unavailable. The client stays operational while spend stays controlled. 

• Admins get alerts at key thresholds: 75% and 100% of the base allowance, then 80%, 95%, and 100% of extra usage. 

• Usage is fully visible. Track credit consumption across tenants, drill into usage by user and model, and spot heavy users before they become billing conversations. 

In utility terms: the allowance is the plan, Extra Usage is the metered overage with a hard cap, light mode is the load-shedding that keeps the lights on, and the dashboards are the meter everyone can read.  

The takeaway  

Price AI as a Service the way it's consumed: a predictable base, governed overage, and full visibility. Your clients get a bill that maps to value. You get margin protection and a growth signal instead of a support ticket. And when usage climbs, that's not a problem to explain. That's the business case proving itself.