Every conversation about AI in the enterprise circles the same $1. The software dollar. The license, the API seat, the platform fee. People argue about which model is smartest, which vendor will win, which SaaS will get disrupted. That argument is about one dollar out of seven.
The other six sit mostly unexamined.
Eric Siu put the ratio in plain terms earlier this year: for every dollar a company spends on software, it spends six on services. Implementation, integration, customization, training, ongoing support, managed operations. The ratio has held for decades because software alone has never been sufficient. You buy the tool; you still need people to make it work.
AI does not collapse that ratio. It changes who captures it.
The Wrong Way to Play This
Most agencies and operators are using AI to compress the cost of delivering that $6. Faster reports. Cheaper decks. Smaller headcount to produce the same output. That is a margin play on a commodity deliverable. You are not changing what you sell. You are just making it cheaper to produce.
The problem with that game: every competitor plays it simultaneously. If AI makes it 40% cheaper to produce a strategy deck, it does that for everyone. Your competitive position is exactly where it was. The clients who were buying on price will push harder. The clients who were buying on quality will notice the output starting to feel similar across vendors.
Bolt AI onto the same model, and you get faster decks and cheaper reports. The same fragile business underneath.
What you have built is a blender for the same ingredients. The output flows faster. The economics are marginally better. And the moment a client or a competitor figures out the same configuration, the advantage disappears.
This is the wrong game. Not because efficiency is bad, but because efficiency on the wrong unit is just organized decline.
The Right Play: Sell the Outcome, Not the Labor
Traditional services sell pieces. A deck. A report. A workshop. An analyst on retainer. The client buys units of human effort and hopes the accumulation of those units produces a result.
The AI-native operator sells something structurally different: a managed outcome. Not the labor that used to produce it. The result itself, running continuously, without the client having to manage the people who make it happen.
Sierra, the AI customer service company, prices exactly this way. They charge only when their agent resolves the issue. Nothing when it escalates to a human. That is not a labor pricing model. That is an outcome pricing model. The value delivered is a resolved customer problem. Sierra owns the definition of "resolved." That definition is the asset.
Sarah Guo described the underlying work well: the value in applied AI is doing "unglamorous work, arranging a company's private reality so a model can act on it." Integration. Maintenance. Context management. That work runs as long as the relationship does. It is not a project. It is infrastructure. And infrastructure commands a different price than a project deliverable.
When you own the outcome definition and the infrastructure that delivers it, you are no longer in the labor arbitrage game. You are in the managed outcomes game. That is a different business entirely.
What Changes with Fable-Class Models
The argument above has existed in theory for years. What is different now is that it is executable.
Claude Fable 5, released in June 2026, is Anthropic's most capable general-use model. It can run for hours on a single task. It tests its own work against defined criteria. It produces output at expert level across analytical, legal, and technical domains without a specialist human in the loop for each step.
Before Fable-class models, the $6 of services required proportional human labor at the delivery end. You could automate pieces, but the judgment layer needed people. A Fable-class model changes that constraint. The analytical work that used to require a senior analyst, a consultant, or a specialist team can now run autonomously at a fraction of the cost.
- Senior analyst required for judgment calls
- Labor scales with scope; cost scales with headcount
- Delivery speed constrained by human availability
- Output quality varies by who is on the engagement
- Client buys hours, hopes for outcomes
- Model handles analytical judgment layer autonomously
- Scope decoupled from headcount at the margin
- Delivery speed set by compute, not calendar
- Output quality consistent across runs
- Client buys the outcome; operator owns the loop
The operator who arranges the private context (the proprietary data, the workflow integrations, the outcome definitions) and deploys Fable 5 against it captures the $6 with a fraction of the headcount that used to be required. The delivery economics flip. The value shifts entirely to the person who owns the context and owns the definition of done.
The $6 Is Not a Homogeneous Block
It is worth being precise about what the $6 actually contains, because not all of it moves at the same speed.
Reporting, summarization, data joins, market monitoring, contract review. This moves first. Fable 5 handles it better than junior staff. Operators who have not automated this layer are already behind.
Connecting models to live company data, building the feedback loops, keeping the context current. This is the unglamorous work Guo describes. It runs continuously and is the primary source of defensible margin for an AI-native operator.
Deciding what "resolved" means. Setting the success criteria. Owning the relationship when the system fails. This layer is not automatable. It is the operator's core product. The previous two layers exist to make this one valuable.
Layer 1 is a commodity race. If you are competing only on Layer 1, you are competing on who can configure the same models cheapest. Layer 2 is where the moat starts. Layer 3 is the business.
What Scaling Companies Need to Understand
If your company is currently buying $1 of software and $6 of services from separate vendors, two things are likely true. First, the services vendor is not using Fable-class models to their full capability. They are using AI to make the old delivery model cheaper, not to change what they deliver. Second, the $6 is being fragmented across relationships that do not compound on each other.
The question worth asking is not which AI software to buy. It is who captures the $6 on your behalf, and whether they are pricing outcomes or labor.
An operator who embeds in your company, arranges your private context, defines what resolved looks like function by function, and deploys Fable 5 against that definition is not selling you software or services in the traditional sense. They are selling you both sides of the ratio, integrated. The $1 and the $6, priced as an outcome.
That is a structurally different engagement from buying a SaaS license and a consulting retainer separately. The economics are different. The accountability is different. The compounding is different.
The $1 Is a Race to Zero
Model capabilities are converging. The gap between frontier models will continue to narrow. The commodity dynamic in software is well established and accelerating under AI. Every dollar of software pricing is under pressure from open-source, from vertical competitors, from foundation model providers moving up the stack.
The $6 is not a commodity race. It is a context race. Whoever holds the deepest, most current map of your company's private reality and owns the outcome definitions running against it is not easily replaced. That is the durable position.
The $1 is moving fast and getting cheaper. The $6 is not.
We capture the $6 with you.
One free conversation. 30–45 minutes. We'll map which of your services budget is ready to be delivered differently and what outcome-priced AI actually looks like in your company.
Book the Diagnostic →