The difference between managing startup finances with a traditional bank vs Brex is like using a Google Maps instead of a paper map. Both technically work, but one is smarter.
Brex is my AI-powered bank. It automatically categorizes expenses, enforces budgets, and integrates with my accounting tools. I sleep better knowing I have actual control over company spend, earn 4.37% on cash, and get 20x higher credit limits than traditional banks.
They hooked up Greg’s Letter readers with serious perks:
- Up to $500 in travel points or $300 cash back after $10K spend
- 50,000 bonus points
- $5K in AWS credits
- $2.5K in OpenAI credits
Only available through this link. Thanks Brex! Now onto today's post....
AX vs UX
This is an important post for founders who want to build lasting, money-making, impactful next generation software (mobile apps, saas etc).
There’s something happening to software that most people haven’t noticed yet, but once you see it, you can’t unsee it. We’re reaching the end of interfaces as we know them.
I don’t mean interfaces are disappearing. I mean the fundamental relationship between humans and software is changing from transactional to conversational, from stateless to stateful, from tools to teammates.
Most software today treats every interaction like meeting a stranger. You open an app, tell it what you want, it gives you what you asked for, and then forgets you exist. Every session starts from zero. This made sense when software was simple and computers were dumb, but it’s becoming obviously wrong as AI gets smarter.
The Relationship Problem
The current model of software interaction is based on a fundamental misunderstanding of how people actually work. Humans don’t think in terms of isolated tasks. We think in terms of ongoing projects, evolving goals, and accumulated context.
When you open your email app, you’re not just “checking email.” You’re continuing dozens of conversations, managing multiple projects, and trying to stay on top of commitments that span weeks or months. But the software treats each email as an isolated event with no connection to anything else.
This context amnesia forces users to become their own memory system. You have to remember which emails are important, which threads need follow-up, which attachments relate to which projects. The software could remember all of this, but it doesn’t, because it’s designed around the metaphor of tools rather than relationships.
Why Now
Three things converged to make a new model possible. First, AI became good enough to actually understand context rather than just executing commands. Language models can now infer intent, maintain conversation history, and make reasonable decisions based on incomplete information.
Second, the cost of maintaining state dropped dramatically. When memory and storage were expensive, stateless interactions made economic sense. Now that compute is cheap, there’s no technical reason why software can’t remember everything about how you work.
Third, users became sophisticated enough to trust AI with more autonomy. People are comfortable with systems that make suggestions, automate routine tasks, and learn from their behavior. The trust threshold shifted from “do exactly what I say” to “help me accomplish what I want.”
What Changes
This graphic was created by my team at LCA. If your company is looking for AX, reach out to them here
The shift from tools to teammates changes everything about how software gets designed and used. Instead of optimizing for task completion, you optimize for relationship building. Instead of measuring clicks and conversion rates, you measure trust and delegation.
Traditional UX design assumes the user knows what they want and the software’s job is to help them do it efficiently. Agentic experience assumes the user has goals that evolve over time and the software’s job is to help them achieve those goals even when they change.
This has profound implications for how software companies think about their products. Netflix isn’t really in the business of helping you find something to watch; it’s in the business of learning your taste well enough to surprise you with things you didn’t know you wanted. Spotify doesn’t just play music; it becomes your personal DJ who gets better at reading your mood over time.
The Trust Gradient
The most interesting aspect of agentic experience is how trust develops differently than with traditional software. With tools, trust is binary - either the software works or it doesn’t. With agents, trust is a gradient that builds over time through successful collaboration.
Early in the relationship, the agent shows its work extensively. It explains why it’s making certain recommendations, asks for confirmation before taking actions, and makes its decision-making process transparent. Users can see exactly what the system is thinking and correct it when it’s wrong.
As the relationship develops and the agent proves its competence, users gradually hand over more autonomy. The agent starts making more decisions independently, requiring less explanation for routine actions, and taking initiative on tasks the user hasn’t explicitly requested.
This mirrors how trust develops in human relationships. You start by giving someone small responsibilities and gradually increase their authority as they prove themselves reliable. The difference is that AI agents can maintain this trust relationship with thousands of users simultaneously.
The Network Effect of Context
Agentic software creates a new kind of network effect based on context accumulation rather than user growth. The more you use the software, the more valuable it becomes because it understands your patterns, preferences, and goals better.
This creates much stronger switching costs than traditional software. When you switch email apps, you lose your folders and filters. When you switch from an agentic email system, you lose an AI assistant that understands your communication patterns, knows which emails are urgent based on your behavior, and can draft responses in your voice.
The value is in the accumulated understanding that the system has developed about how you work and not just features anymore. This understanding becomes a moat that’s extremely difficult for competitors to replicate.
What Gets Automated
The boundary between what users do manually and what gets automated shifts dramatically in agentic systems. Instead of automating predefined workflows, the system learns to automate whatever the user finds tedious or repetitive.
Your calendar app learns your scheduling preferences and starts automatically finding optimal times based on your energy levels, travel patterns, and work priorities. Your project management tool learns to predict which projects are at risk and suggests interventions before problems become critical.
The automation is learned. This makes it much more flexible and personalized than traditional workflow automation, but also more unpredictable and harder to understand.
The Design Challenge
Using Perplexity's Browser "Comet" to figure out what this complicated software actually costs (source)
Designing agentic experiences requires thinking about user interface design completely differently. Instead of designing screens and flows, you’re designing relationships and trust-building mechanisms.
The challenge is creating systems that are powerful enough to be truly helpful but transparent enough that users understand what’s happening. Users need to feel in control even when they’re delegating authority to the system.
This requires new design patterns that don’t exist in traditional UX. How do you design an interface that adapts to each user’s mental model? How do you visualize an agent’s confidence level in its recommendations? How do you let users correct the system’s understanding without breaking the overall experience? How do you use copy (tone, length etc) as one of the fundamental building blocks of AX?
Why This Matters
The companies that figure out agentic experience first will have an enormous advantage because users who experience true AI partnership can’t go back to dumb tools. Once your email app understands your communication style and can draft responses for you, using regular email feels broken.
This creates a winner-take-all dynamic in many software categories. The first company to build truly agentic experience in a category get customers who become extremely difficult to switch away from. We're helping companies become agentic at LCA and it's been pretty eye-opening to see what categories there are opportunities in.
We’re still in the early stages of this transition. Most “AI-powered” software today is just traditional tools with chatbots bolted on. True agentic experience requires rethinking the fundamental architecture of how software works, not just adding AI features to existing interfaces.
The shift is inevitable because the technology now exists to build software that actually understands users rather than just executing their commands. The only question is which companies will recognize this opportunity and build accordingly.
Users will stop tolerating software that makes them repeat themselves once they experience systems that actually remember and learn. The future belongs to software that becomes more valuable the more you use it, not because of network effects with other users, but because of the relationship it builds with you.
Note to self: will try this. Probably could automate this with an agent that delivers you the news twice a day based on preference. There's a startup idea somewhere in there ;).
