I am less interested in another window that can chat about everything than in changing a plan without checking five separate places afterward.
Suppose a group of friends postpones a weekend trip. The tickets can be changed and the hotel allows rescheduling, but their available dates may not line up. Some friends have agreed; others have not read the message. The shared document and calendar still contain the old times. Every app works, yet none can answer a simple question: what is actually booked now?
The previous article considered why someone would switch products. Here I want to look between products. Even excellent tools can leave their user as the only person connecting them: moving information, reconciling dates, and remembering what remains unfinished.
I think software can keep multiplying while our everyday entry points become fewer. Connecting several apps to a chat box is only part of getting there.
More tools, the same person coordinating
Tickets, accommodation, and messaging have good reasons to be separate products. They manage different records, transactions, accounts, customer relationships, and revenue. A trip spans those services, but that does not give one provider authority over another's refunds, prices, or availability.
Fragmented entry points therefore reflect how services operate as well as how interfaces are designed. Spreadsheets, browsers, suites, and plug-ins have long brought work together. Models that interpret instructions and call tools offer another way to connect smaller pieces.
Growing software supply makes that question more useful to examine, though a supply count cannot tell us how much work users face. RevenueCat's 2026 report, citing Appfigures, puts monthly launches of mobile subscription apps at roughly 2,000 in January 2022 and above 14,700 in January 2026. The report
Monthly subscription app launches: two snapshots
By early 2026, monthly additions were roughly seven times the level four years earlier.
- January 2022 (about)2,000
- January 2026 (more than)14,700
View app launch data
| Month and qualifier | New subscription apps / month |
|---|---|
| January 2022 (about) | 2,000 |
| January 2026 (more than) | 14,700 |
Two benchmarks from the report's text, not a complete monthly series. The 2022 figure is approximate; the 2026 bar ends at the 14,700 threshold, below the reported actual count. Coverage is iOS / Android subscription apps. The data do not identify how many were built with AI.
SourceThat shows growth in one category. It does not establish that AI caused all of it, or that everyone is overwhelmed by apps. My design question comes next: will the operations a new tool saves be offset by choosing it, configuring it, and repeating the same background elsewhere?
If five services each have an assistant and I must explain the trip to all five, the coordination remains. Copy-and-paste has become five conversations.
Fewer clicks do not necessarily mean less effort
We can time five conversations; the extra effort needed to keep track of them is harder to see. A controlled CHI 2008 experiment offers a reason to examine speed and burden separately. Its 48 participants performed simulated office tasks with no interruptions, same-context interruptions, and different-context interruptions. Interrupted participants spent less net time working, excluding the interruptions themselves, but reported more stress. The paper
Net work time was shorter under interruption
These times exclude the interruptions themselves; they are not total elapsed completion times.
- Uninterrupted22.77
- Same-context interruption20.31
- Different-context interruption20.6
View net work time data
| Experimental condition | Mean net work time (minutes) |
|---|---|
| Uninterrupted | 22.77 |
| Same-context interruption | 20.31 |
| Different-context interruption | 20.6 |
Means from 48 participants who experienced all three conditions; variation is not plotted. Same-context interruptions concerned the main task's subject. This chart does not establish an overall time saving from interruptions.
StudyIn the same experiment, reported stress was higher
Both interruption conditions had higher stress ratings than the uninterrupted condition.
- Uninterrupted6.92
- Same-context interruption9.46
- Different-context interruption9.13
View reported stress data
| Experimental condition | Mean stress rating (1–20) |
|---|---|
| Uninterrupted | 6.92 |
| Same-context interruption | 9.46 |
| Different-context interruption | 9.13 |
Ratings run from 1 (low) to 20 (high); bars show means from a zero baseline. This is a subjective scale, so score differences should not be converted into a percentage increase in stress. It is not a survey of today's consumers.
StudyGerman university students made up 81% of the sample. This was neither a survey of today's app users nor evidence that interruptions shorten the entire job. The useful reminder is narrower: people can work faster while having to push harder.
Back in the travel example, I would include that burden in acceptance checks. Suppose an assistant removes ten clicks but requires three pages of reading before I trust the booking: the saving needs recalculating. Initial permissions, service response times, date checks, and recovering records after an error belong to the same task. Setup reusable on the next trip could reduce those costs over time.
A workspace I can return to
For this rescheduling task, I would want a living arrangement: old and proposed dates together, train and hotel options compared as combinations that are actually available, and friends' confirmations shown separately. If transport is changed but accommodation is pending, that difference should remain visible. Before payment, show the amount and any irreversible changes.
Natural language helps explain why the trip is moving and what unusual constraints apply. Dates belong on a calendar; prices need comparison; frequent controls should stay in familiar places. Regenerating the way everything works can spend the saved time on relearning.
There are already different approaches to this. Google Research's November 2025 generative UI work produced interactive pages from prompts. At the time, generation could take over a minute and outputs could contain errors; the preference evaluation excluded generation speed. Google's research account
Another approach brings existing tools into a common entry point. Claude's interactive connectors, announced in January 2026, put services such as Asana and Figma inside conversations. This demonstrates integration capability, rather than its long-term effect on workload. Claude's announcement
These attempts make me want to inspect what happens after generation. Can I reopen the trip next week, let friends edit it, and export confirmed bookings? The screen needs to retain the latest state, and the information needs to be portable afterward. Otherwise I have to reconstruct the itinerary from conversation history again.
What would make fewer entry points worthwhile?
Providers need reasons to participate. Will a booking platform let another service own the customer relationship? Does query access also permit changes? Do commissions influence the recommendations? A unified screen does not unify the businesses underneath it.
Users also need a way out. Collecting preferences, history, and permissions in one place is convenient. If the arrangements can only be opened there, switching that entry point may become harder than replacing an individual app. Exporting records, returning directly to the original service, and changing providers should be usable features.
If a common entry point can preserve the arrangement, secure providers' participation, and let users take their records away, I am more optimistic about it handling recurring coordination. If each task still demands re-explanation, long waits, and item-by-item checks, I would start with easier-to-assess search or organization. Those are adoption conditions, not evidence for a year when the transition will be complete.
Specialist apps also deserve to remain directly accessible. An experienced spreadsheet user may finish with one drag. Design and editing need precise controls; games and social experiences often make the interaction itself the point. One fewer entry point is not automatically better. Removing a stretch of repeated manual transfer is a much more concrete improvement.
I would watch an ordinary moment: after changing the tickets, can the user put the task down, or must they still remember to chase the hotel, group chat, and calendar? That unfinished work is what I want the next interface to take on.
