Learn how to optimize AI costs with smart model selection, cost-efficient workflows, and spending controls that keep your budget predictable and your team productive. 

The best way to manage AI spending isn't to use AI less. It's to stay in control of how and when you use it. Hatz makes that simple. 

• 80% of business tasks work fine on our lighter, faster models 

• Teams defaulting to high-cost models spend 3-5x more than necessary 

• Smart automation saves both time and credits 

Quick Wins for AI Cost Control 

• Default to Auto Lite. Most everyday work doesn't need premium models. Escalate to Performance or Turbo only when the task genuinely requires it, then drop back down. 

• Check the multiplier before you choose. Hover over any model in the selector to see its cost relative to the baseline (1x). A 3-4x model across long sessions adds up fast. 

• Enable subagents. They route specialized tasks to lighter models so your primary model isn't doing everything. 

• Start fresh chats for new topics. Long conversations carry full history as input, which costs more with every message. New topics should start in new chats. 

• Turn recurring prompts into Workflows or Agents. If your team types the same thing weekly, automate it instead. 

• Keep context lean. Upload only the pages you need. Paste excerpts, not entire documents. Input size is part of your cost. 

• Think in ROI, not just spend. A 50-credit conversation that saves two hours is an excellent trade. The question isn't how to use fewer credits, it's whether you're getting real value. 

If you run out, work doesn't stop. Hatz has built-in safeguards and optional overage billing so you stay in control. 

Why Your AI Costs Vary More Than You'd Expect 

Credits are Hatz's unit of AI consumption. Every message you send draws credits based on two factors: the amount of text processed (input and output) and which model you use. Since every model carries a different multiplier (cost efficiency rating), your monthly budget stretches further or shorter depending almost entirely on which models you use and how much context you carry into each conversation. 

The core variables are simple but compound quickly: 

• Which model is handling the request 

• How much conversation history has accumulated 

• Whether tools or files are involved 

• How you structure your chats and workflows 

A long conversation on a high-cost model with large attachments costs significantly more per message than a fresh, focused chat on Auto Lite. 

The key insight: these aren't problems to avoid. They're dials you can turn. 

The Single Biggest Lever: Smart Model Selection 

Model selection is where most teams find their biggest opportunity to reduce costs. The issue: leaving a high-cost model selected out of habit, when 80% of business tasks work fine on our lighter, faster tier. 

How Auto Lite works: Our Auto Model Selection feature routes each message to the most capable model the task actually requires. On Lite tier, that means fast, efficient responses for everyday work like drafting emails, summarizing documents, brainstorming, and routine analysis. 

When a task genuinely needs more compute power (complex reasoning, nuanced writing, multi-step analysis), escalate to Performance or Turbo for that message only, then drop back down. You're not switching chats or managing model selection manually. Auto routes per message, so you get the right tool for the job, automatically. 

The cost difference: Teams that default to Turbo for everything often spend 3-5x more per message than Performance tier, and up to 12x more than Lite, even though most of those messages didn't need premium performance. 

Understanding Model Multipliers 

Every model in Hatz has a multiplier that shows its cost efficiency compared to an industry flagship at 1x baseline:

• 0.5x - roughly half the cost per message 

• 1x - matches the baseline 

• 2x or higher - roughly double or more 

How to find multipliers: Hover over any model in the model picker to see its multiplier in the tooltip. Or go to Model Selector > Multipliers tab for a sortable, filterable comparison. Watch for promo badges—we occasionally run promotional pricing on specific models, and the displayed multiplier always reflects the current rate. 

Before escalating from Auto Lite: Always check the multiplier. A 3-4x model across a long session compounds quickly. 

Use Subagents for Task Routing Efficiency 

Subagents are purpose-built specialists that work alongside your primary model. Instead of your main model handling everything, subagents automatically delegate specialized tasks to lighter, more efficient models. The result: better outputs and smarter credit usage. 

If your workspace doesn't have subagents enabled, turn them on. The efficiency gains are automatic and passive. You don't manage the routing yourself. 

Start Fresh Chats for New Topics 

Every message in a conversation carries the full thread history as input. The longer the chat, the more tokens get processed with every new message, even simple follow-ups. A conversation that started as strategy brainstorm and drifted into scheduling is burning credits on irrelevant context with every response. 

The rule: When you're starting a new topic, start a new chat. Cost per message drops, and the model gets a cleaner signal to work from. 

Long threads make sense when continuity matters (iterating on a document, working through a multi-step problem). For everything else, fresh is cheaper and sharper. 

Automate Repetitive Work with Workflows and Agents 

If someone on your team is opening a chat and typing a similar prompt regularly, that's a Workflow, App, or Agent waiting to exist. 

Rebuilding context, re-explaining formats, and re-typing instructions every time wastes both time and credits. Agents and Workflows handle repeatable work in a structured, focused way, faster and more consistent because the context is baked in, not rebuilt each time. 

The Workshop Assistant can build this for you. Describe what you want in plain language: 

• "Take our existing chat and turn it into an Agent or Workflow" 

• "Create a workflow that takes a weekly report and writes an executive summary" 

• "Build an agent that drafts client follow-up emails based on meeting notes" 

It can create, modify, run, and explain workflow items all from one conversation. Find it in the Tool Selector. No setup required. 

Keep Your Context Lean and Focused 

Everything you put into a conversation is part of the cost. Uploading a 40-page document when you only need two pages adds input tokens the model processes regardless of whether they're useful. 

Habits that make a real difference: 

• Trim file uploads to only the relevant sections 

• Paste targeted excerpts instead of full documents 

• Avoid re-uploading the same file across conversations (if it's referenced repeatedly, it belongs in an Workflow or Agent) 

• Keep questions focused. One clear question per message typically costs less and gets better answers 

Precision saves money. Every irrelevant token is money you didn't need to spend. 

Think in ROI, Not Just Cost 

Credits are a resource, not a tax. A 50-credit conversation that saves two hours of research or prevents an error is an excellent investment, regardless of absolute cost. 

The right question: "Am I getting real value from the credits I'm using?" Not "How do I use fewer credits?" 

High usage from AI doing meaningful work is the system working as intended. High usage from runaway context, defaulting to Turbo, or repetitive manual prompting is worth addressing. Not because spending is bad, but because those patterns produce less value per credit than they could. 

What Happens When You Run Out of Credits 

Running out of credits doesn't mean your team hits a wall. Starting July 1, 2026, Hatz handles overages with built-in safeguards and transparent controls: 

1. Step 1: Auto Lite Buffer: Once monthly credits are exhausted, Hatz automatically switches all tasks to Auto Lite tier. Every plan includes a dedicated buffer. Work keeps moving. 

2. Step 2: Graceful Pause: If the buffer is also exhausted, access pauses until your billing cycle resets. Nothing is lost. Everything is safely paused. 

3. Step 3: Optional Overage Billing: Extra Usage Billing is opt-in and on your terms. You set a threshold in credits or dollars for how much additional usage you're comfortable authorizing. Any overage bills at the start of your next cycle. No surprise charges. 

Why this matters: The rest of the AI industry is moving toward unpredictable, metered billing. Anthropic separated agentic workloads from base plans. Microsoft moved Copilot to AI Credits. Claude's recent tokenizer changes quietly raised per-call costs up to 27%. Hatz runs the opposite direction: transparent, forecastable spending that puts control in your hands. 

For Admins: Scaling Cost Control Across Your Team 

If you're managing AI adoption at scale, Hatz gives you the controls to stay in charge: 

• Set role-based credit limits to establish per-user or per-role spending thresholds proactively 

• Restrict high-cost models for roles that don't need them. A support team on Lite by policy costs a fraction of an unrestricted deployment 

• Use Auto Mode policies to keep teams defaulting to Lite or Performance automatically 

• Monitor the Adoption Tab to see where credits are going, who your heaviest users are, and whether usage patterns are producing real outcomes 

The Bottom Line 

AI cost management isn't about using less. It's about using smarter. The tools are in your hands: choose the right models, structure your conversations intentionally, automate what repeats, and keep your context focused. 

Default to Auto Lite. Escalate only when necessary. Start fresh chats for new topics. Turn repetition into automation. Keep spending predictable. 

With Hatz, your budget goes further, your team stays productive, and you stay in control. 

Ready to Get Started? 

Explore our Credits & Usage guide to dive deeper into cost optimization. 

Or talk to our team about how Hatz helps your organization scale AI with confidence and control.