Over the last few months, one topic has come up consistently across MSPs and their customers: 

“Why is AI usage — and credit consumption — going up so much?” 

On the surface, it can feel like something is wrong. 

But the reality is the opposite. 

What we’re seeing right now is one of the clearest signals yet that AI is working. 

The Misconception: “Costs Should Be Going Down” 

There’s a natural assumption in the market: 

• Models are getting cheaper 

• Intelligence is becoming more efficient 

• Therefore, total spend should decrease 

That logic makes sense. 

But it’s not what actually happens. 

What’s Actually Happening: Usage Is Exploding 

As AI becomes: 

• More capable 

• More accessible 

• More embedded into daily workflows 

People use it more. 

A lot more. 

This is a classic example of Jevons Paradox — when efficiency increases, total consumption increases. 

We’re seeing this play out in real time: 

• A company starts with a handful of users 

• Those users find value quickly 

• Usage becomes daily (emails, proposals, research, support) 

• More employees want access 

• Workflows get more advanced 

And suddenly, AI is no longer a tool — it’s part of how the business operates. 

The Big Shift: Agentic AI Is Driving Usage Through the Roof 

There’s another major shift happening right now: 

The move to true agentic AI. 

AI is no longer just responding to prompts. 

It’s: 

• Running multi-step workflows 

• Calling tools 

• Searching data 

• Executing tasks autonomously 

And that changes everything. 

Token consumption is going through the absolute roof. 

The average usage per: 

• Individual user

• Entire company 

…has increased dramatically. 

To put that into perspective: 

I personally used around 1 billion tokens last month — roughly the equivalent of 7,500 books worth of words. 

Now, that’s on the high end. 

But even when you scale that down across:

• Teams 

• Customers 

• Entire MSP portfolios 

…it becomes very clear why systems are hitting limits and why credit usage is spiking. 

It’s Not Just More Usage — It’s More Users 

At the same time: 

Agentic AI is bringing entirely new users online. 

As tools become: 

• Easier to use

• More powerful 

• More capable of doing real work 

People who would never have used AI before are now:

• Engaging daily 

• Running workflows 

• Depending on it 

So you’re seeing:

• More usage per user 

• More users overall 

• More complex workflows 

All happening simultaneously. 

The Shift in the Models Themselves 

Alongside this, the model landscape is evolving fast. 

While many models are getting cheaper, a new class of frontier models has emerged that are: 

• More powerful 

• More capable 

• Designed to execute real work 

These models can consume: 

10x to 100x more tokens than lighter alternatives. 

And that’s expected. 

Because they’re doing far more. 

What You’ll See Next: A Shift Toward Cost-Efficient Intelligence 

One of the most important trends to watch from here: 

Companies will start experimenting much more aggressively with lower-cost models. 

This has already started. 

Leaders across industries have highlighted they use open-source models over ChatGPT in certain scenarios because they were:

• Faster 

• More cost-effective 

This is a signal of where the market is going. 

In the agentic era, it’s not just about raw intelligence anymore. 

It’s about: 

Intelligence per unit of cost. 

Why This Matters Right Now 

If you’re an MSP or a business leader, you’re probably asking: 

“How do we get ahead of rapidly rising AI costs?” 

That’s the right question. 

Because while usage is increasing — and will continue to — the real opportunity lies in: 

• Optimizing which models are used 

• Matching task complexity to model cost 

• Managing usage across teams 

• Creating visibility into consumption 

The Short-Term Reality for MSPs

There’s an important truth: 

You will feel this in the short term. 

As adoption increases:

• Usage spikes 

• Costs rise 

• Customers start asking questions 

This is happening across the entire industry. 

Not just in one platform. 

Where the Market Is Heading 

We are in a clear transition: 

Phase 1: Access & Activation 

Get AI into users’ hands quickly.

Phase 2: Usage & Adoption 

Drive real, daily usage. 

Phase 3: ROI & Optimization (next phase) 

• Optimize cost vs. performance 

• Understand usage deeply 

• Tie spend to outcomes 

The conversation is shifting fast: 

From “access” → to “efficiency” → to “ROI.” 

Spend Management Becomes the Competitive Advantage 

This is where MSPs win. 

Anyone can provide AI access. 

Very few can: 

• Optimize usage 

• Manage cost intelligently 

• Deliver ROI clarity 

• Guide customers through this transition 

Spend and credit optimization becomes a competitive advantage. 

Not as a restriction… 

…but as a value creation layer. 

Where Hatz AI Comes In 

This is exactly the problem Hatz AI is built to solve. 

As the market moves toward: 

• Higher usage 

• More models 

• More complexity 

MSPs need: 

• Cost optimization 

• Credit optimization 

• Visibility

• Control 

And that’s where Hatz delivers:

• Intelligent model routing 

• Multi-model access 

• Usage visibility 

• Optimization over time 

Because the future isn’t about choosing one model. 

It’s about managing many — efficiently. 

Final Thought: Don’t Fight the Trend 

If there’s one takeaway: 

Rising AI usage is not a problem — it’s a signal. 

The winners won’t be the ones who: 

• Restrict usage 

• Slow adoption 

They’ll be the ones who: 

• Enable it early 

• Guide it properly 

• Optimize it continuously 

Because AI doesn’t plateau. 

It compounds. 

And with agentic AI, that compounding is accelerating faster than most people realize.