Context switching at work: how to design a low-friction environment with AI
You open an AI chat to write a draft. You copy it into another tool to summarise a PDF. You download the result, move it to a spreadsheet and, finally, save it in a folder that no model can access. When you finish, you feel as though you have done a lot of work. Yet much of that effort has gone into moving information from one place to another.
In this post, you will see what happens to your attention every time you switch tools and why having lots of scattered AI applications reduces their value. You will also find three principles for designing a unified workspace and a simple formula for measuring friction in your processes.
What happens to your brain when you switch tasks
Human attention is not designed to jump from one thing to another without a cost. When you switch between applications, tabs and language models, the change is not instant. Cognitive psychology calls this the task switching cost. Researchers such as Robert Rogers and Stephen Monsell (1995), and Monsell himself (2003), have studied it.
This cost arises for two main reasons.
1. You have to reconfigure your mind
Each task has its own rules, goals and ways of working. When you move from a document management system to an AI chat and then to a spreadsheet, your mind has to set aside the rules of the previous task and load those of the new one. This includes what you are looking for, what constraints you face and even where everything is on the screen.
This reconfiguration takes effort and time. That is why, immediately after switching tasks, you tend to work more slowly and make more mistakes.
2. The previous task keeps interfering
What you left unfinished does not disappear from your mind. It keeps competing for your attention and makes it harder to properly analyse what is in front of you. This is very similar to the attention residue we discussed in the previous chapter.
The key idea: with traditional software, fragmentation made you less efficient. With AI, it can cancel out much of the benefit you were looking for, because you spend the mental energy that AI was supposed to free up on logistics.
The trap of scattered AI tools
In recent years, dozens of specialised tools and standalone models have emerged. This has created a common problem for businesses and professionals: AI tool sprawl.
A typical workflow often looks like this:
- You write a draft in a tab using a text model.
- You copy the text into another web application to review the style or summarise a PDF.
- You download the result and move it to a third platform to put the data into a table.
- You save the information in local folders that are not connected to the model's instructions.
This setup creates three major problems.
You lose context with every switch
The tools do not share memory. In every new tab, you have to explain again who you are, what you need and what limits the model must respect. It is a kind of platform amnesia that forces you to repeat the same work over and over.
Security and privacy risks increase
If you spread sensitive data across five or six free or consumer services, you lose track of where your information is and how it is used. This makes the data sovereignty we discussed in the first chapter harder to achieve and increases the risk of breaching the data protection regulations that apply to your activities.
The logistics tire you out, not the work
Your mental energy goes into copying, pasting, changing formats and logging in. This is precisely the extraneous load we discussed in chapter 5: effort that does not help you make better decisions and leaves you with fewer resources for what matters.
Three principles for a unified workspace
For AI to work as a mental exoskeleton rather than another source of interruptions, you need an environment with as little friction as possible. That environment rests on three pillars:
Principle 1: make context persistent
Your business knowledge, such as your methodologies, templates, criteria and historical records, should not live in instructions saved in a notepad. It should be organised in persistent knowledge bases that you can easily connect to any agent or conversation.
Example: instead of pasting your style guide every time you request a piece of writing, you save it once in your knowledge base and connect it to the agents that write.
Principle 2: make it possible to switch models
As we saw in chapter 3, no single model is ideal for everything. A good environment lets you use different engines on the same documents and from the same interface:
- Advanced models for complex reasoning.
- Lightweight models for quick classification or filtering.
This way, you choose the engine based on the task, not on which tab you have open.
Principle 3: keep the interface simple
The interface should make it easy to work without interruptions. To do that, it should include:
- Direct access to your documents.
- Editing in the same place where you work with AI.
- Clear references to the sources for each answer.
- The ability to run workflows without leaving the environment.
How to measure friction in your processes
To find out how much tool sprawl affects you, apply this audit to your three main processes, such as preparing a report, an expert opinion or a sales proposal.
Friction index = (no. of application switches + no. of copy-and-paste operations) ÷ time spent on value-adding tasks
How to calculate it
- Choose a process and complete it from start to finish as you normally would.
- Count every time you switch applications or tabs.
- Count every time you copy, paste, download or upload a file to move information.
- Measure the time you spend on value-adding tasks, such as analysing, making decisions or writing up your own ideas. Do not include the time you spend moving data.
- Divide the total number of interruptions by that time, measured in hours, for example.
Example: you prepare a sales proposal with 12 application switches and 18 copy-and-paste operations, and spend 1,5 hours thinking and making decisions. Your index is (12 + 18) ÷ 1,5 = 20 interruptions per hour of value-adding work. In other words, one every three minutes.
The number is a guide. Its value lies in comparing your processes with one another and seeing whether they improve over time.
A warning sign
Friction is critical if preparing an expert opinion, a report or a proposal requires more than three platforms and many manual text transfers. In that case, adding another AI tool will not help. You need to unify your infrastructure.
Conclusion:
Reducing context switching is not about working faster, but about no longer spending attention on moving information back and forth. If you centralise your knowledge, can choose the model without switching environments and carry out actions in a single workspace, AI genuinely helps you think better.
This chapter concludes the second module of the series, dedicated to the relationship between AI and the human mind. In it, you have seen how to prevent AI from weakening your thinking, how to protect yourself from automation bias, how to escape the trap of superficial productivity and how to unify your work environment.
Want to work with AI in an environment with less friction? Visit libertadia.com and sign up in the application.