Software is about to stop looking like software
The product interface is going through its biggest shift since desktop to web. Most builders are looking the wrong way.
This is Part 2 of the Future of Software series. Read Part 1: Software Is a Proxy. AI Makes It Obsolete.
“AI will replace all software UIs with a chat window.” Open a text box, type what you want, get an answer. No dashboards. No navigation. No buttons.
Sounds elegant. Also wrong.
The opposite take is equally wrong. That AI just means adding a chatbot sidebar to your existing Salesforce instance and calling it transformation. I’ve seen four enterprise clients do exactly this in the last 18 months. Users ignore the chatbot within two weeks. Every time.
The product interface isn’t dying. It’s going through the biggest shift since desktop to web. And most builders are getting it wrong because they think it’s a binary choice. Chat or dashboard. Pick one.
The future is neither.
Why pure chat fails
Before I get into where things are going, it’s worth understanding why the “just a chat box” pitch falls apart the moment someone tries to use it.
Humans think visually, not sequentially. A business owner asks “how is my company doing?” and they don’t want 500 words back. They want to see revenue trending up, one project turning red, cash reserves stable. A well designed visual communicates in 2 seconds what takes 2 minutes to read. We built a project management tool for a healthcare client last year. First version had a conversational interface. Client’s project managers hated it. They needed to see the sprint board AND team capacity AND client deadline all at once. Not one at a time through a chat window.
Chat has no persistent state. Ask “show me overdue tasks.” Get a list. Close the chat. Where’s the list? Gone. Ask again tomorrow, get a new list. Humans need spatial anchoring. The feeling that information lives somewhere stable, not summoned from thin air each time you ask.
Chat can’t handle parallel attention. A project manager needs four data streams in peripheral vision while focusing on one. Chat forces everything into a single sequential thread. That’s a downgrade, not an upgrade.
So pure chat is out. But traditional dashboards are also increasingly inadequate. They show everything to everyone, all the time, regardless of what actually matters right now. They make humans come to the information instead of bringing information to the human.
What’s left?
Four stages of product interface evolution
This transition isn’t a single leap. It’s a progression through four stages. Each one changes the fundamental relationship between the user and the product. I’ve been building B2B software for 18 years and we’re watching this play out in real time across our client base.
Stage 1: Static dashboards (where most products are today)
You navigate to information. The product has a fixed structure. Tabs, sidebars, pages, settings. The same dashboard shows the same layout to everyone, every time.
This is how Salesforce, Jira, HubSpot, Tally, and virtually every B2B tool works today. The UI is the product. Learning the UI is learning the product. The skill is in navigation.
We operate at this stage with most of our enterprise clients still. About 80% of the custom software we build in any given quarter is Stage 1. It works. It’s just not where things are heading.
Stage 2: Conversational hybrid (where the progressive products are heading)
You ask for information, the system responds with both text and visual context. The dashboard still exists but it’s increasingly secondary.
“What should I focus on today?” returns a prioritised list with context about why each item matters. Not because you navigated to a “priorities” page, but because the system synthesised across your tasks, calendar, team status, and deadlines to generate an answer.
We started building Stage 2 interfaces about eight months ago. A healthcare client wanted their clinical coordinators to be able to ask “which patients need follow up today” instead of navigating through four different screens to compile the list manually. The coordinator’s daily workflow went from 45 minutes of screen navigation to a 3 minute conversation. Same data, same decisions, completely different interaction model.
Most users in Stage 2 products spend 80% of their time in conversation mode and 20% in dashboard mode. The inverse of today.Stage 3: Ambient intelligence (where things get interesting)
The system doesn’t wait for you to ask. It proactively delivers the 2 or 3 things that need your attention, through whatever channel you’re already on. WhatsApp, Slack, email, push notification.
8 AM, your phone buzzes: “Two items need attention. One team member is blocked. Reply 1 or 2 to choose an approach and unblock them. One deadline is at risk. Approve a scope adjustment or extend the timeline. Everything else is on track.”
You reply “2” and “approve.” Twelve seconds. Your entire morning interaction with what used to be a complex software product.
The design principle inverts: if the system hasn’t contacted you, everything is fine. Silence is the signal that things are working. This is how a great executive assistant operates. They don’t hand you a 50 page report every morning. They tell you the three things that need your judgment. Everything else is handled.
We’re prototyping this for a logistics client right now. Their operations manager currently spends 90 minutes every morning reviewing dashboards across three different systems. The goal is to bring that down to under 5 minutes of notification responses. We’re about halfway there.
Stage 4: Generated interfaces
No pre-built screens exist at all. When you want to see something, you describe what you want and the system generates a visualisation on the fly. Perfectly tailored to your question, your role, and your context.
“Show me how the team is doing” generates a completely different view for a CEO than for a team lead. The CEO sees cross-team velocity trends and budget utilisation. The team lead sees individual contributor output and blocker resolution times. Same question, different generated interface, because the system understands who’s asking and what they actually need.
“Compare this quarter with last quarter” doesn’t load a pre-built comparison page. It generates a side by side analysis highlighting specifically what changed, what caused the change, and what actions might address it. The visualisation is ephemeral. It exists for this moment, for this question, and then dissolves.
There is no “analytics page” with 15 pre-configured charts that someone built once and nobody updates. Every view is generated on demand, contextual to the user, and disposable after use.
We haven’t built a full Stage 4 system for a client yet. But we’ve built pieces. A government reporting module where the interface generates different compliance views depending on which regulatory body is requesting the data. Same underlying dataset, completely different presentation. That’s a taste of where this goes when it matures.
Five properties of the future interface
Regardless of stage, the direction is clear. Future interfaces share five properties that are fundamentally different from today’s software.
1. Exception based, not comprehensive
Today’s products show you everything. All tasks, all deals, all transactions. The user’s job is to scan through comprehensive views and find what matters. That’s an enormous cognitive tax.
Future products show you only what needs attention. Three items are red. Everything else is fine. The 200 transactions that reconciled correctly are invisible. The 3 that didn’t are surfaced.
The metric for a well designed future product is not “time spent in app.” It’s “decisions made per minute of attention.” If a user can process their entire daily workload in 90 seconds of interaction, that’s a triumph. Not a retention problem.
I keep having this argument with product managers who are worried about engagement metrics dropping. If your users accomplished everything they needed and left in 45 seconds, you won. If they spent 20 minutes browsing dashboards and left without taking an action, you lost. The engagement model of the future is the opposite of consumer social media.
2. Multi-modal, not screen bound
Today’s products live on a screen. You open the app, use it, close it.
Future products exist across multiple channels simultaneously. Morning summary arrives as a WhatsApp message. Time sensitive approval pops up as a push notification. Deep analysis request generates a visual in the web app. Quick status check answered by voice assistant.
The product meets you where you are, in the format appropriate for the interaction’s complexity. Channel selection isn’t a user preference to configure. It’s a decision the system makes based on urgency, complexity, and context.
We’re seeing this with our own internal tools. Our marketing system runs across 10 tracks with automated alerting. When ad spend drifts more than 15% overnight, the alert goes to Slack. When a review drops below 3 stars, the notification goes to the responsible person’s phone. When someone wants to dig into campaign performance, they open the dashboard. Different channels for different urgency levels. Nobody configured that. The system decides.
3. Conversation first, visualisation second
The primary interaction model is natural language. But the response isn’t always text. That’s the nuance the “just a chat box” crowd misses.
When you ask “how are we doing on revenue?” the best answer is a chart. When you ask “who’s falling behind?” the best answer is a ranked list. When you ask “what happened last Tuesday?” the best answer might be a timeline.
The AI should know when to respond with words, when to respond with a visualisation, and when to respond with an action. The interface is generated in response to the question, not pre-built awaiting it. Pre-built dashboards assume the designer knows what the user will want to see. Generated visualisations assume the system can figure it out in real time. The latter is almost always more useful, because what matters changes every day.4. Progressively deep. Glanceable to explorable
The initial response should be absurdly simple. A traffic light. A single number. A one sentence summary. “Everything’s on track” or “Two things need your attention.”
But depth should be instantly available. Tap on “two things need attention” and you get the specifics. Tap on a specific issue and you get the full context. Conversation history, timeline, root cause analysis, recommended actions. Tap further and you get the raw data.
The design principle: the top layer should be understandable in 3 seconds. Every deeper layer is opt in. Most days, most users never go past the first layer. That’s success, not failure.
Think of it like a newspaper versus a research library. Today’s software is a research library. Everything is there, find what you need. Tomorrow’s software is a newspaper with a library behind it. The headlines tell you what matters, and you can always dig deeper.
5. Action oriented, not information oriented
This might be the most underappreciated shift. Today’s products end at “here’s the information.” The user then has to decide what to do and go somewhere else to do it. See a problem in the dashboard. Think about the solution. Switch to email to communicate it. Switch to the task tool to update it. Switch back to verify.
Future products end at “here’s the recommended action. Approve?”
“A team member is blocked on a decision. Here are both options with trade offs. Reply 1 or 2 to unblock them.” That’s not information delivery. That’s a decision point with pre-packaged execution. The user makes a judgment call, the system handles everything else. Communicating the decision, updating records, adjusting timelines, notifying stakeholders.
Every interaction should conclude with the work being done, not with the user having more information to act on manually.
What this means if you’re building
If you’re building a B2B product right now, or redesigning one, a few things follow from this.
Your home screen should not be a dashboard. It should be “What needs your attention right now.” Two or three cards, maximum. Everything on track is invisible.
Your primary interaction model should be conversational. A text input at the centre, not a navigation menu. “What’s behind schedule?” gets an intelligent answer. “Reassign the homepage task to Sarah” is executed, not routed to a form.
Your notification layer is your most important surface. More users will interact through notifications than through your app. The notification isn’t a pointer to the app. The notification is the app for 90% of interactions.
Build zero permanent analytics pages. Every chart should be generated on demand in response to a question. Kill the “Analytics” tab with its 15 pre-built charts that 3% of users visit monthly.
Every piece of information should end with a suggested action. Don’t show “3 invoices are overdue.” Show “3 invoices are overdue. Payment reminders drafted. Send all three?”
Measure decisions per minute, not time in app. Less time in your product means more value from your product.
The uncomfortable bit
Most B2B products being built today, including products that call themselves “AI-powered,” are designed around Stage 1 assumptions. They have dashboards. They have navigation menus. They have settings pages and configuration wizards. They’ve added an AI chat sidebar. That’s decoration.
The products that win the next decade will be designed around Stage 3 and Stage 4 from the ground up. They’ll feel less like “software” and more like “a team member who happens to be omniscient and tireless.” The interface won’t be something you learn or navigate or spend time in.
It will be something that works for you. Mostly in the background, occasionally surfacing for your judgment, always concluding with the work actually getting done.
The dashboard is dead. Not because screens are dead or because visual information is dead. But because the idea that a human should spend time navigating pre-built screens to find information and then separately act on it. That idea is dead.
What replaces it is better. Faster. Quieter. And, paradoxically, more visual than ever. Because when the system generates exactly the right visualisation for exactly the right question at exactly the right moment, the visual impact is far greater than a permanent dashboard full of charts nobody looks at.
The future of product UI isn’t no interface. It’s the right interface, at the right time, for the right person, generated in the moment it’s needed, and gone the moment it’s done.
We’re building some of this right now. The early results are encouraging. And honestly a little unsettling. When the first thing your client says after using the new system is “I don’t feel like I’m using software anymore,” you know something fundamental shifted.
Navneet Singh is the Founder & CEO of Webority Technologies. He writes about engineering-first approaches to building technology companies.