#006: The Chat Bar Isn’t Lazy Design
The chat bar isn’t hiding the product. It’s a new kind of product.
Thought: The Chat Bar Isn’t Lazy Design
This tweet has been making the rounds:
There was a ton of interesting discussion (1, 2, 3, 4, 5, 6, 7, 8). Most of the discussion has been focused on whether this is good or bad design. Instead of debating whether it’s good or bad (the what), I think a better way to understand this shift is to ask two questions (the whys):
Why is everyone building the same thing?
Why does it manifest as a chat bar?
Why is everyone building the same thing?
In my recent article “A Small Figma Update and a Big Signal for SaaS”, I wrote:
The software development process is moving from a discrete model - clear steps with explicit handoff at each step - to a more fluid model with no gates, stages, or handoffs as the process gets increasingly absorbed by the AI itself. Most SaaS tools were either designed to simplify a specific step and/or to improve the handoff between them. So what happens when those steps collapse into a process that needs no handoff?
And:
As this harsh reality is slowly (or rapidly?) crystallizing for every SaaS company out there, they’re now facing a fork in the road - become “the everything tool” to own this new fluid process or get commoditized as another plugin into Claude or ChatGPT.
Figma was a place where design started and lived. Linear was a place where project management started and lived. Notion was a place where documentation started and lived. Now, everything starts and lives in an LLM (Claude Code / Cowork / Codex / etc.) and the tools that used to “own” a step in the process are now plugins into your workflow.
Such a shift radically changes your position in the value chain. Instead of owning the direct relationship with your user, you are now relegated to a supplier role. If you’re a VC-funded B2B SaaS business, your valuation and revenue multiples are grounded in the former world - a world where you owned the user relationship, not rented it from an LLM.
So do you accept your new inferior position and revise your valuation? No, you try to compete. You rebuild your tool for the new agentic process, hoping the work once again starts and lives in your tool.
Why does it manifest as a chat bar?
The core critique of the chat bar UI from the discussion of the tweet went something like this - a blank chat box shifts the burden of navigation onto the user and hides what the product can do. The job of product design is to curate the right tools and actions for the user based on the context.
The problem with this critique is that it’s grounded in the world of tools. And these are no longer just tools. Let me explain.
The best framework that bridges the past and the future here is Jobs-To-Be-Done: customers don’t buy your tools - they hire them to get a job done. Each SaaS app was effectively a toolbox. Our job as designers was to understand the job at hand, surface the right tools in a given context, and make those tools easy to use so you can accomplish the job. With LLM agents, a SaaS app is now more akin to a coworker that goes out and does the job. And how do you talk to a coworker? You chat with them.
Your coworkers don’t “expose their capabilities” or have a sleek dashboard. You just reach out to them on Slack, delegate a task, they may ask you a few clarifying questions, and then they just go out and do it. And that’s what these SaaS apps now are - they’re significantly closer to your coworker than to a toolbox.
Does it mean we no longer need direct access to the tools ourselves with all the knobs, buttons, and forms? No, of course not. And none of the apps discussed above have completely removed their direct manipulation interfaces. They just added a new level of abstraction - an agent that can use the tools for you to accomplish the job (sometimes with your help). But this abstraction will only strengthen over time to the point where direct manipulation and tool use will become an auxiliary need, not a core workflow.
This “coworker” mental model neatly and naturally extends into the future:
The best coworkers are proactive. And so the best agentic SaaS will be proactive - it will anticipate your needs, catch issues before you do, and go out and solve them, only looping you in when needed.
The best tools are custom-made. When your input is vital, the agents will be building the UI for the task at hand, on the fly, exactly as you need it. Your coworker will forge you a perfect tool every time instead of handing you a rusty hammer.
As designers and builders, we need to adapt our mental models in this new world of agentic interfaces. The chat bar isn’t just the latest UI fad - it’s the first visible crack where the toolbox ends and the coworker begins. Our job now isn’t to design better toolboxes. It’s to design the interaction model between humans and their new coworkers.
3 Bits
1. Claude Mythos is Delusional
As you can start to tell, I’m mostly a tech and AI optimist. But the skeptic in me is on alert when the exuberance is running high. This commentary definitely appealed to my skeptic side and it makes some compelling points. It’s good to maintain a balanced perspective when the narrative in tech seems one-dimensional.
2. Hermes Agent
There has been a lot of chatter and excitement online around the new Hermes Agent - a new open-source agent harness which seems to be running circles around OpenClaw in terms of long-term memory, skill-based self-improvement, and better compatibility with local models. I avoided OpenClaw because of cost & setup headaches, but may give this a try with a capable local model to tinker & learn.
3. Synthetic Focus Groups Using Claude
There’s simply asking Claude “tell me which features will be most valuable” and then there’s a robust framework that gets better the better you understand your customers. Craig has been on a roll with thoughtful and practical AI workflows lately (he is the creator of pastmaps.com - a genuinely cool product.)
Carveouts
Sakeeb shared a great post about my new product, Remarc, that I’ve been slowly seeding to early testers. It’s made for builders who care about the details. I’m putting the finishing touches on the public release, but happy to give early access in the meantime.
Sign-off
That’s it for today. Keep tinkering and see you on Thursday.
Appendix: The New Format
As I mentioned in the last issue, I’m moving from the 3/3/3/3 weekly format to something leaner: 1 thought and 3 bits, twice a week (Tuesday & Thursday). The “bits” are interchangeable - could be a post, a tool, a person worth following. I hope you like it. If not, please send me an angry email.
in the spirit of focus in an ADHD land, this new format is working well for me.
have to think that everyone leading with chat on top of claude as their engine is just going to get consumed when anthropic goes direct. latest with Design is a definite threat to most. Default tends to win scale.