Look Away
Our customers asked for dashboards. What they actually wanted was the freedom to stop watching one.
By Kelly Breakstone Roth, Co-Founder & CEO of Prysmic · June 2026 · 6 min read
Last week I was listening to Lenny Rachitsky's podcast. His guest, the legendary Tony Fadell, told a story about the first iPhone. When Apple was building it, the team spent months arguing about a keyboard. For years a phone meant keys you could feel, and the BlackBerry was the thing to beat. So the team built both, a physical keyboard and a slab of glass, and tested them against each other again and again. The glass kept losing on speed and on accuracy. It stayed worse, until one day it was good enough, and they made the leap that defined the next twenty years. Touch became the way we trust a phone, so completely that we forgot it had ever been a question.
Operations software is in the middle of its own keyboard moment, and we have been building right at the center of it. When we started Prysmic, we built dashboards. Beautiful ones. If you run operations for a consumer brand you know the shape of them, a single board that pulls every shipment and every purchase order and every exception into one place so you can sit in front of it and know where things stand. That is what operations software has always been. It is what our customers asked us for, and it is what we were proud to put in front of them. We made those screens good.
And then we noticed something we had not expected. The people we built them for barely looked at them. One logistics lead we had built a board for told us, politely, that he could already see a good deal of it elsewhere, and that honestly it sat a little too far in the weeds for his job. What he cared about were the three or four shipments, out of hundreds, that could actually go wrong that week. The rest, he was glad to never open again.
The dashboard was our keyboard
None of this was a layout problem we could fix with a cleaner screen. It was the same message arriving from every direction, and it took us a while to hear it, because it sat in direct contradiction to what those same people had asked us to build. They had asked for dashboards. Every one of them. Of course they had. The dashboard was the only language operations software had ever taught them to speak, the only shape the thing had ever come in. When the tool you know is a screen full of status, a screen full of status is what you ask for.
The dashboard was our physical keyboard. The best answer the last era could give, asked for so many times it stopped looking like a choice.
The real work was learning to translate. What a customer asks for and what a customer needs are rarely the same sentence, and our job was to dig underneath the request until we found it. Underneath every ask for a better dashboard sat the same unspoken need: to stop being the monitor, to hand the operation over and trust that it was handled, to be pulled back in only when a person was truly required. The deeper want was to be free of the looking altogether.
When the software becomes the engine
Here is what sits under all of it. In the software all of us grew up with, the human is the engine of the work. The product stores the information and lays out the controls, and then it waits. You are the one who reads the state of things, decides what matters, moves between tools, and takes the next action. A dashboard is the perfect interface for that arrangement, because when you are the engine, you need to see everything at once.
Agents break that arrangement. A system that can take in a goal, work through the steps, judge its own progress, and act no longer needs you to be the engine. And the moment the software does the interpreting and the doing, the screen stops being the place where the work happens. The work happens underneath it. What the human should see becomes a completely different question than the one a dashboard was built to answer.
In the software we all grew up with, the human is the engine. Agents are the first thing that can take the engine's seat.
The obvious next move is to drop a chat box where the dashboard used to be and let people type their questions instead of clicking for them. Chat earns a place. When you want a fast answer pulled from across the whole operation, asking in plain language and getting it back is a real step up. But an answered question is only half of what these teams are reaching for. The heavier half is the work itself, and the responsibility that rides on it. They do not only want to ask what is happening. They want it handled, the lower-level decisions lifted off their plate for good.
Chat is a good feature. It was never going to be the product.
What replaces the dashboard
Mark Weiser, who led research at Xerox PARC, wrote in 1991 that the most profound technologies are the ones that disappear, the ones that weave into the fabric of daily life until you stop noticing them. Operations software spent the next thirty years doing the opposite, demanding to be watched. Agents are the first real chance it has had to recede. So what replaces the dashboard is not a single thing, and that is the point. The operation runs itself, mostly out of sight, and surfaces to you in whatever shape the moment needs. It might be one line in the channel your team already lives in: this shipment is running two days late, I have moved it to the next sailing and told the warehouse, and your launch date still holds. The status arrives with the action already taken and the consequence already worked out. When a real decision needs you, the choice shows up right there, with nothing to navigate.
The screen does not vanish in any of this, and it should not. It changes jobs. One of our customers put it better than I could. The agents had gotten so good at handling things on their own that he could no longer see the work, and he wanted it back in view. Not to do it himself. To trust it. Think about how you bring on a sharp, motivated new hire. Before you hand them the hardest calls, you want to watch how they think, how they reach a decision, where their judgment is already sound and where it is still forming. Autonomy is something they earn as you come to trust them. An agent is no different. So the screen becomes the place you watch an agent reason, check what it did, sign off on the moves that carry real weight, and reconstruct what happened when something unusual occurs. It is an audit and approval surface, and it has little in common with a wall of live tiles.
The lines are the product
Which means the real design work, the part that used to be a dashboard layout, is now a harder set of questions. What should the system do entirely on its own? What should it only ever recommend? What needs a human to sign off every single time? Which actions are cheap and reversible, and which are the forty-thousand-dollar kind that should never happen without a person in the loop? Gartner, which a year ago was urging everyone to adopt agents as fast as possible, now warns that many enterprises will end up pulling their autonomous agents back, almost always because they deployed them without drawing those lines. The lines are the product. Graded autonomy, visible permissions, and clear approval boundaries are the new interface, even when none of them look like a screen.
So the human is not edged out of the operation. The work moves up. Less time spent as the engine, refreshing portals and chasing statuses, and more spent on the things only a person should hold: the intent, the risk thresholds, the decisions that carry consequence, and the accountability for all of it. That is the part of the job most operators wanted more of in the first place.
The interface is assurance
Strip all of this down, and what an operator actually wants from an interface is assurance. The peace of mind that the operation is being handled, correctly and reliably, without them standing over it. The more useful question is the one our customers kept leading us back to: how do you actually want to work with your operation, now that it can carry itself? Because the interface that matters in this new era is no longer just the thing a human looks at. It is the whole arrangement, the intent and the context and the tools and the memory and the permissions and the approvals, organized so that a person and a system can run an operation together without either one becoming the bottleneck. And it does not have to look the same for everyone. A planner, a VP of operations, and a finance lead can each get their own view and their own moment to step in, because the interface is now composed around the person instead of built once for the average user.
That is the question we built Prysmic to answer. We started where everyone in this category starts, with the screen, and our customers kept pointing us past it to the thing they were really after: an operation that runs, holds, and earns their trust, and asks for them only when it counts. The dashboard had a thirty-year run as the answer to a question operators were never quite asking. We are building the answer to the one they were. And the most valuable thing you can put in front of an operator, it turns out, is the confidence that they can look away.