The Two Faces of AI Value
Every successful AI product creates value in one of two ways. Never both. And the companies that fail do so because they confuse the two.
Production efficiency — AI as a compressor. It compresses time, mental effort, and trial-and-error cost. The user is a producer: a writer, a coder, a researcher. They want either the same output in less time, or more iterations in the same time. ROI is calculable: hours saved × hourly rate.
Consumption friction-removal — AI as a translator. It translates "something I need to figure out" into "it is already done for me." The user is a consumer: they open an app, receive a result, and are satisfied. ROI is felt, not calculated. "This just works" vs "why do I have to deal with this."
These two modes demand completely different product strategies, business models, and user interactions. Mix them up, and you build something nobody wants.
A 15-Year Case Study: Siri
Siri launched in 2011 with a promise that felt like magic: talk to your phone like a person.
What followed was 15 years of disappointment.
Why? Because Apple asked consumers to act like producers.
Every Siri interaction was a command: "Set a reminder for 3 PM." "Send a message to John saying I will be late." "What is the weather like today?"
This is production language. You are telling a system what to do, specifying parameters, and expecting it to execute. It is the same cognitive mode as writing code, drafting an email, or operating a machine. It is work.
For 15 years, Apple told consumers: "Learn our voice command syntax. Memorize the patterns. Structure your thoughts into commands." And for 15 years, most people simply... did not.
Siri was not technically broken. It was structurally misaligned with how consumers want to interact with technology.
WWDC 2026: The Half Step
This week, Apple announced Siri AI — rebuilt with Google Gemini as its foundation. On-screen awareness. A dedicated app. Visual intelligence. Standalone Mac integration.
Is this different?
Partially. On-screen awareness is real progress: when Siri can see what you are looking at, you do not need to specify. "Where is that restaurant from the Instagram post" works without you copying and pasting. That is friction removal. That is consumer logic.
But the dedicated Siri app? That is production logic wearing consumer clothes. A standalone chatbot app that you open, type into, and review output from? That is the same cognitive model as ChatGPT, Claude, and Gemini. It is producer software marketed as consumer software.
The Dynamic Island integration is smart — it is where you already are. The conversational mode is better than commands. But Siri AI still fundamentally asks you to produce: formulate a request, wait for a response, evaluate the result.
Products That Got It Right
TikTok recommendation algorithm. You open the app. You scroll. Content appears. You never prompt, never specify, never iterate. AI is the engine, invisible, producing a perfectly tuned feed. The user role: pure consumption.
Spotify Discover Weekly. Every Monday, a playlist appears. You hit play. That is it. AI analyzed your listening history, compared it to millions of others, and delivered a result. You did not lift a finger.
Google Maps automatic rerouting. You are driving. Traffic is bad up ahead. Google Maps silently changes your route. You do not notice. You just arrive faster.
iPhone Smart HDR. You press the shutter button. The photo looks good. Behind the scenes, AI is compositing multiple exposures, optimizing dynamic range, and balancing colors. You never see it happening.
These products share a single pattern: AI does the work. The user just... consumes.
Not a single one of them asks the user to "write a prompt," "review and edit," or "iterate until satisfied." They absorb all complexity and deliver a finished result.
The Producer-Consumer Contract
This framework is a direct application of a deeper principle I have been developing: the producer-consumer contract.
The producer side: Take complexity upon yourself. Deliver simplicity to others. If your product has exposed knobs — settings to tweak, parameters to adjust, decisions to make — you have not finished absorbing the complexity.
The consumer side: Do not overthink. Do not second-guess. Delegate to professionals. If you find yourself struggling with a tool, that is not your failure — it is the tools failure to absorb its own complexity.
Now apply this to AI:
Production AI (co-pilot mode):
- AI sits alongside the producer, offering suggestions and accelerating their work
- The user stays in control, makes the final call
- Value is measured in efficiency gains
- Examples: GitHub Copilot, Midjourney, AI-QC verification pipelines
Consumption AI (engine mode):
- AI is embedded invisibly, doing the work before the user even notices
- The user does not make decisions — they receive results
- Value is measured in friction removed
- Examples: TikTok feed, Google Maps navigation, Smart HDR
The tragedy of the current AI industry is that most companies build producer tools and sell them as consumer products.
The Diagnostic
Here is a simple test for any "AI for Everyone" product:
- Does it ask the user to formulate a request (prompt)?
- Does it ask the user to evaluate and iterate on the output?
- Does it require the user to learn a new interaction pattern?
If you answered yes to any of these, you are building a producer tool. That is fine — producer tools are valuable. But market it honestly: to producers, as an efficiency multiplier. Do not tell consumers they need to "learn how to use AI."
And here is the darker implication: true consumer AI is invisible. You cannot build a brand around it. You cannot put "AI-powered" on the box. It just makes things work better, and nobody thanks you for it because they did not notice.
What This Means for Builders
| If you are building for... | Do this | Do not do this |
|---|---|---|
| Producers | Lead with measurable efficiency. "Save 3 hours/day." Calculate ROI. Let them stay in control. | Tell them it is fully autonomous. Ask them to trust decisions they cannot verify. |
| Consumers | Embed AI into existing behavior. Do not mention AI. Deliver finished results, not drafts. | Create a new "AI app" they need to learn. Ask them to prompt, review, and edit. |
The hardest lesson: these are different product categories with different logics. A consumer AI product that asks users to write prompts has an identity crisis. A producer AI tool that hides its controls from power users is equally confused.
The Siri AI Verdict
After WWDC 2026, my assessment is:
What got better: On-screen awareness, conversational interaction, system-wide integration. These reduce friction. These respect the consumer role.
What stayed the same: The fundamental interaction is still producer-oriented. You prompt, you review, you decide. A standalone Siri app is a chatbot, and chatbots are producer tools.
What is still missing: True invisible AI. Siri should observe, predict, and deliver — not wait to be asked.
Siri AI is better than Siri 2011-2025. But it is not yet consumer AI done right. It is a half step toward a framework that someone — maybe not Apple — will eventually fully execute.
The Bottom Line
AI creates value in exactly two ways: by compressing production time or by removing consumption friction. Products that serve both roles serve neither well.
Build your product for one. Market it to one. Do not make your consumers do producer work, and do not take control away from power users.
The producer-consumer contract is a reminder that elegance is not in the product — it is in what the user does not have to deal with.
This framework builds on ideas from my ongoing series on the Five-Layer Operating System and the producer-consumer contract. English posts on dev.to, Chinese translations on WeChat.
Follow me on Bluesky: @keeperlant.bsky.social
Top comments (0)