You’re Using Claude Wrong #171b – Why AI Costs Are About to Become a Leadership Problem

On July 13, 2026, Anthropic changed how organizations pay for its most advanced AI model, Fable 5. Instead of being included in a subscription, it moved to metered billing, making every prompt and generated response a direct cost.

Joerg Storm’s central message is simple:

The biggest AI mistake companies will make is using the most powerful model by default.

Choosing an AI model is no longer just a technical decision—it’s now a financial and governance decision.

The Big Change

Until now, many teams could freely use Fable 5 as part of their subscription.

Now:

• Every token processed has a price.

• Fable 5 costs approximately:

  • 2× more than Claude Opus

  • Around 5× more than Claude Sonnet

That means organizations that continue using Fable for every task could see their AI costs increase dramatically without realizing it.

Storm argues that this isn’t bad news—it simply requires better decision-making.

The New Mindset: Don’t Default to the Best Model

Many users assume:

Best model = Best choice

Storm argues this thinking is outdated.

Instead, organizations should ask:

• Does this task actually require frontier-level intelligence?

• Would a cheaper model produce nearly identical results?

• Is the additional quality worth paying several times more?

Only a small percentage of work truly needs the most expensive model.

Model Routing Is the New Superpower

One of the biggest ideas in the newsletter is model routing.

Instead of everyone always using Fable 5:

• Simple tasks go to inexpensive models.

• Medium-complexity tasks go to Sonnet.

• Only the hardest reasoning problems reach Fable.

Think of it like choosing transportation:

• Walking for short distances.

• Driving for longer trips.

• Flying only when necessary.

You wouldn’t book an international flight to travel across the street.

The same principle now applies to AI.

Leaders Need AI Spending Policies

Storm believes organizations need formal governance around AI usage.

That includes:

• Defining which teams can use premium models

• Setting approval rules

• Monitoring AI spending

• Tracking return on investment (ROI)

• Creating routing guidelines

• Preventing unnecessary costs

Without governance, AI expenses can quietly grow into a significant operational cost.

The Hidden Cost Most Companies Ignore

The newsletter explains that model pricing isn’t just about the advertised cost.

Actual spending depends on factors such as:

• Tokenization efficiency

• Prompt size

• Response length

• Context window usage

• Cached prompts

• Batch processing discounts

The “headline price” is rarely the amount organizations ultimately pay.

Understanding these factors can significantly reduce costs.

Use Expensive AI Only Where It Creates Real Value

Storm introduces the idea of scoring tasks before assigning them to a model.

Premium AI should be reserved for work such as:

• Long-term strategic planning

• Complex research

• Multi-step reasoning

• Resolving conflicting information

• Critical business decisions

• High-value customer interactions

Routine work—like drafting emails, summarizing meetings, formatting content, or basic coding—rarely justifies the premium price.

Separate Thinking From Doing

One of the most practical recommendations is splitting AI into two roles:

Advisor

Uses the most capable (and expensive) model to:

• Develop strategy

• Solve difficult problems

• Create plans

• Make complex decisions

Executor

Uses lower-cost models to:

• Write content

• Perform repetitive tasks

• Execute workflows

• Handle routine operations

This architecture keeps most AI usage inexpensive while still benefiting from frontier intelligence when it matters.

Always Have Fallback Models

Storm recommends designing systems that automatically switch models when:

• Costs exceed budget

• A premium model isn’t required

• The preferred model is unavailable

• Response quality is sufficient from a cheaper alternative

This prevents unexpected billing spikes while maintaining productivity.

AI Governance Is Becoming a Business Discipline

The newsletter includes guidance for organizations on:

• Cost governance

• Spending controls

• Executive approval policies

• Team-wide routing strategies

• Risk management

• Rollout plans over 7, 30, and 90 days

The goal is to make AI spending predictable, measurable, and aligned with business value.

Key Takeaways

• Fable 5 is now a premium-priced AI model, making model selection a financial decision as much as a technical one.

• Most work does not require the most expensive AI model.

• Implement model routing so tasks are matched to the appropriate level of intelligence and cost.

• Separate strategic reasoning (“advisor”) from routine execution (“executor”) to maximize efficiency.

• Establish governance, spending controls, and fallback mechanisms before scaling AI across teams.

• Measure the return on investment for premium AI usage instead of assuming higher capability always delivers greater value.

Final Thoughts

Joerg Storm’s message goes beyond comparing AI models. He argues that the next competitive advantage won’t come from simply having access to the most powerful AI—it will come from using the right AI model for the right task at the right cost.

As frontier models become more expensive, successful organizations will be those that combine intelligent routing, clear governance, and disciplined spending. In the emerging AI economy, **efficiency—not just capability—will determine long-term success.

https://drstorm.substack.com/p/youre-using-claude-wrong-171b?utm_campaign=email-half-post&r=5xcpdm&utm_source=substack&utm_medium=email