Concept

A useful learning decision connects a technology to a task, its costs and a manageable next step.

Issue

A good demo doesn’t tell you whether learning the tool is worth the effort for your work.

Solutions

Know what to compare before you commit your time, then choose your next learning step.

You open a tutorial for a new tool while a report sits unfinished. The demo looks useful. Before you’ve tried anything, you’re wondering whether to learn the whole system or close the tab.

You don’t have to make that whole decision yet. Start with a smaller question: what would make the next bit of learning worth it? That means looking at the task, the time it would take and what you’d put aside to do it.

This module is for working adults thinking about an optional tool for a specific work task. It applies when you can undo a trial and put a limit on what it costs.

What you’ll get from this module

In this module

By the end of this module, you’ll know how to choose a manageable next step in learning a tool. You’ll see what research on choices, workplace tools, training and adoption shows, along with one company’s decision to stop developing a product. Then you’ll pick an everyday situation and use a written exercise to decide: adopt for a trial, watch for a change or ignore for now.

What makes learning worth the effort?

A review of many experiments: seeing what else you could do.

Allegra Maguire, Emil Persson and Gustav Tinghög’s 2023 meta-analysis combined the results of 39 experiments from 12 published papers. The experiments compared choices made with and without a reminder of what else the same money or time could pay for. That alternative is called the opportunity cost: what you give up when you choose one option.

On average, the reminders changed people’s choices. The effect was small once the earliest experiments were left out. Most of the choices were imagined rather than real, and the results differed from one setting to another.

A new option is only one side of a decision. The other side is what you give up to pursue it. Seeing that alternative can change the choice, though choosing fewer options isn’t automatically better.

Why it matters: For learning, this splits one question into two. Does the tool look attractive? And is it worth the time you’d otherwise spend on something else?

Read the opportunity-cost meta-analysis.

A workplace study: a tool can bring real gains.

Erik Brynjolfsson, Danielle Li and Lindsey Raymond’s November 2024 paper studied an AI assistant that reached 5,172 customer-support agents at different times. By comparing how agents did before and after they got it, the researchers estimated that agents resolved 15% more issues per hour on average.

The gains weren’t the same for everyone. Newer and less skilled workers got faster and did better work. The most experienced and most skilled workers got a little faster, and the quality of their work dropped slightly.

Workers were trained before they got access. The rollout wasn’t random, so crediting the gain to AI depends on how well the researchers allowed for other differences. The study shows a useful gain in that workplace, with different results for different workers.

Why it matters: That makes the task and the result important. Finishing faster matters, but so does whether the work is good enough. An average gain can’t tell you how much one particular worker will be helped.

Read the customer-support study.

A field study and its update: feeling faster isn’t the same as finishing sooner.

Joel Becker and colleagues at METR studied 16 experienced open-source developers in early 2025, working on software projects they knew well. Across 246 real coding tasks, each task was set at random to allow or forbid AI.

When AI was allowed, developers took 19% longer. Afterwards, they believed AI had made them 20% faster. For this group, in this setting, feeling faster did not mean finishing sooner.

These results describe the tools available at the time. In February 2026, METR published an update describing problems with who took part in its newer work and how time was measured. Those problems made the newer estimates unreliable. The update doesn’t show that the result has reversed, and it doesn’t give a figure you can rely on for today’s tools.

Why it matters: The customer-support and developer studies involved different people, different tasks and different ways of using AI. Together they show why a general impression of a technology can’t settle a decision about your own work. You need to look at both the result and the effort it took to get there.

Read the early-2025 developer study. Read the follow-up measurement update.

Does a good first look tell you enough?

A lab study and a critique: how you learn matters.

In 1987, John M. Carroll and colleagues compared two ways of learning an unfamiliar word processor, in a study of what they called the Minimal Manual. In the first experiment, 19 office workers in a mock office used either a short manual built around tasks or the standard self-teaching manual.

The group with the task-based manual reportedly needed 40% less time to learn. Its manual combined several features, including guidance for real tasks and help with mistakes. This was a small study of an old system, so that figure isn’t a forecast for learning today’s software.

In 1992, Thomas R. Williams and David K. Farkas published a critique that helps explain what the comparison can and can’t show. The shorter manual mixed exploring with other kinds of help. Because those features changed together, it was hard to tell which one made the difference.

Why it matters: So the help on offer for learning belongs in your judgment of a tool. Struggling with a poor introduction may tell you little about what you could do with good guidance. The result also doesn’t prove that exploring without guidance is the best way to learn.

Read the Minimal Manual study. Read the critique of the learning-design evidence.

A factory experiment: who pays for the learning?

David Atkin and colleagues’ working paper, revised in July 2016, looked at a cutting tool that wastes less material, used by soccer-ball makers in Sialkot, Pakistan. An experiment with rewards involved 31 active firms that already had the tool.

Both groups got a refresher and an offer of a demonstration. One group was also told about a problem with rewards: the people who bore the cost of learning weren’t the ones who got the benefit. Its workers were offered bonuses if they showed they could use the tool.

Being given that offer raised the chance of taking up the tool in the short term by an estimated 32 percentage points. This small experiment measured whether firms took up the tool, and the offer combined an explanation with bonuses. It can’t separate the effect of the money from the rest of the offer.

The authors’ explanation is backed by their field evidence. For a learning decision, it points to who has to spend the time and effort, as well as who might benefit.

Why it matters: A tool can be useful while the person learning it has no time or support. Asking about paid learning time, guidance and existing duties helps bring those obstacles into view, without blaming the learner.

Read the manufacturing adoption experiment.

A company announcement: big backing doesn’t guarantee a product will last.

On 4 August 2010, Urs Hölzle announced that Google would stop developing Wave as a standalone product. Google also planned to reuse its technology and let people export their content.

This old case shows that even a large backer can’t promise that a standalone product will keep being developed. It also shows that a product ending isn’t the same as everything connected to it disappearing.

Why it matters: For your learning decision, ask what you could still use if support ended. Some knowledge or work might stay useful; other work might depend on the product. You can think about what you might lose without guessing how long the product will last.

Read the Google Wave announcement.

Know what you keep if a tool ends

What this means for you

Decide how much to commit next, based on the work you value and what you could gain or lose. That might be a small trial, a supported spell of learning or a deliberate pause. You don’t have to decide whether you’ll use the tool forever.

Start with the work and an acceptable result

Name a specific task, how you do it now and what better would look like. Include the standard the work must meet. Saving time only helps when the result is good enough for its purpose.

Think about when you’ll next use what you learn. Work that repeats gives you many chances to use it. A valuable one-off task can also be worth the effort; explain its value without claiming savings you’ll repeat.

Exploring because you’re curious is fine too. Name that purpose and decide how much time and money to give it. Enjoyment is a different reason from expecting the tool to save work time.

Count the learning effort and what it replaces

Work out the smallest useful thing you need to learn. Include guidance, setup, access, money and checking, plus anything you need to know first. Ask whether you’ll get support for learning or have to fit it around your existing duties.

Then name the best other use of that time: necessary work, rest or something else you want to learn. Include your longer-term development. Otherwise, work with a quick result can keep pushing aside learning that matters later.

A colleague’s experience may answer a question cheaply. Compare their task, skill, support and costs with yours before you rely on their result. Those details show whether their experience answers what you need to find out.

Judge the work after practice

Where privacy or consequences for your organization matter, use access your workplace has approved and practice material that isn’t sensitive. Never use this optional learning exercise to get around required processes for high-stakes work, or tools your workplace requires everyone to use.

Use good guidance before you judge an unfamiliar tool. Following a tutorial tells you something different from doing similar work without copying each step.

Say you’re trying a reporting tool. Finishing the tutorial’s example shows you can follow its instructions. A similar practice report shows what you can do when nobody hands you the next step.

Count setup, help, fixes, checking and tidying up in the total effort. Compare work of acceptable quality with your usual method. If you need to time your usual method, include that time in your budget, or make the budget bigger before you start.

If the whole task won’t fit, choose a typical part of it. Your conclusion then covers that part, not the whole job. One attempt can show up a problem without showing what you’d usually gain. When you can afford it, compare several similar tasks on different days, and switch which method you use first.

If you’re missing access or skills you need first, you can’t make a fair attempt. Choose watch and name what would let you try fairly. For a skill that’s costly to learn, arrange time, funding and good guidance for a first learning goal before you judge typical work.

Limit what you could lose

Think about what knowledge, method or work would stay useful if support ended. Check the real support and export options: can you keep useful work outside the product? A well-known company name doesn’t answer that.

Mark what you don’t know as unknown. Useful work now can justify a small commitment, even when future support is uncertain or you can’t tell whether the learning will help elsewhere. Decide how much time, money and work you’re willing to risk.

Choose adopt, watch or ignore for now

Adopt for trial when you have a likely useful task, costs you can accept and a way to judge the result. You’re choosing the next limited step. You don’t need to change your whole way of working or start a long course.

Watch when the tool seems likely to help but a specific obstacle remains. Perhaps you lack access, need to learn something first or don’t know whether you can export your work. Name an event or a review date that could change your choice; watching doesn’t mean checking every day.

Ignore for now when you have no task you value and no reason to explore, something else deserves the time and money more, or a serious limit makes a trial a bad idea. Add a reason to look again later if missing a change could matter.

Sometimes a permitted trial you can undo costs less than more thinking. If you have a useful task and can live with the loss, try it. Think it through first when setup is big or you can’t contain what might go wrong.

Set a limit before you start and stick to it. Extend once, and only if you can say what the extra work will show you and why that’s worth its cost. Set the new limit first.

Check a time-saving claim with a simple comparison

For a rough check of time alone, divide the time it takes to learn and set up by the time each acceptable use saves. Take ongoing checking and upkeep off the savings first. What’s left is the net saving.

Say learning and setup take 120 minutes, and each acceptable use saves 10 minutes after those ongoing costs. Dividing 120 by 10 gives 12 uses to win back the time you put in.

Those numbers are made up. You still need to ask how often you’ll really use the tool and how much time it will save. If each use saves no time, or takes longer, this sum never pays back.

Better quality, a new ability, enjoyment or a duty may still justify the choice. Compare those benefits with the best other use of your time. You don’t have to turn every benefit into minutes saved.

Is this tool worth your time to learn?

How about your daily life?

Is there an optional tool you’ve bookmarked for a report, a spreadsheet task or a creative project at work? What work would you actually use it for? What other work or learning would compete for the same time?

List situations you’d like to explore. One good idea is enough.

A task or optional technology I’m consideringWhat I’d like to improve or understand
________________________________________
________________________________________

My chosen situation: ____________________

Choose an optional work task where you can undo a trial and limit its costs. You can change your choice. Take this situation into the exercise; picking it doesn’t mean the tool is worth adopting yet.

Choose your next learning step

Start with your chosen situation. If you came straight here, choose an optional tool for a specific work task where you can undo a trial and limit its costs.

Where privacy or consequences for your organization matter, use access your workplace has approved and practice material that isn’t sensitive. Never use this exercise to get around required processes for high-stakes work, or tools your workplace requires everyone to use. Use your organization’s proper process instead.

  1. Define useful work. Name the task, how you do it now, the quality it needs and when you’ll likely use it next. A valuable one-off task, or exploring on a set budget, is fine; name that purpose without claiming savings you’ll repeat.
    Task, quality and purpose: ____________________
  2. Check the five parts of the decision. Use information you can easily look up. Set a time limit for this check and mark anything you can’t answer as unknown.
QuestionMy answer
What task will this help with, and when will I use it?____________________
What must I learn, and what guidance, access, setup, money and checking will I need?____________________
What specific work, rest or longer-term learning would this replace?____________________
What knowledge, method or output could I still use if the product disappeared?____________________
What support and export options can I verify, and what would support ending cost me?____________________

If a useful task, permitted access, a loss you can live with and a test you can undo are already clear, weigh testing against more thinking. Skip the rest of the thinking when the test would cost less.

  1. Choose the next commitment. Adopt for trial if the task is useful, the costs are acceptable and you can judge the result. Watch if a named obstacle remains; record the event or date that could change your choice.
    Ignore for now if there’s no purpose you value, something else deserves the time and money more, or a serious limit makes testing a bad idea. Record a reason to look again later when missing a change could matter.
    Choice, reason and any revisit trigger: ____________________
  2. If trying, set up a fair, limited attempt. Before you start, state the quality you need, your limits on total time and money, and when you’ll stop. Use recent results from your usual method as a starting point, or time it and check its quality within the budget. Make the budget bigger beforehand if you need to.
    Follow good guidance and practice, then try similar work without copying each step. Record setup, help, checking, fixes and tidying up as part of the total effort.
    Stop and choose watch if missing access or skills prevent a fair attempt. If there’s a lot to learn, arrange a funded learning goal with good guidance first. If the whole task won’t fit, choose a typical part and limit your conclusion to it.
    Limits, stop condition, guidance and comparison: ____________________
  3. Decide what the attempt warrants. Compare the quality of the work and the total effort with your usual method. Adopt the next limited step, watch for a specific open question, or ignore for now.
    One comparison doesn’t give a reliable average or prove what caused a difference. When you can afford it, use several similar tasks on different days and switch the order of the methods.
    Extend once, and only when you can name what you’ll learn and why it’s worth the extra cost. Set a new limit before carrying on. If you’re claiming time savings, check that the net saving is positive once you count what you put in at the start and how often you expect to use the tool.
    What I observed, what remains unknown and my next step: ____________________

Ask Jing

Become a Jingius.

Decide how far to go with a new tool

Loading interactive tool…

Your next step

You’ve finished when you have a decision with a reason, and either a limited next commitment or a specific obstacle and a trigger to look again. A clear reason to wait or say no is a useful result.

A note on the evidence

These sources answer different questions. Most opportunity-cost experiments used imagined choices. The workplace studies looked at particular tasks and ways of using the tools. The training and factory studies help explain the conditions for learning; Wave is an old example of a product’s development ending.

Use your own comparison to judge the attempt you made. When deciding how much more to commit, keep in mind what you don’t know about future use, what you’ll need to learn and support.

How we prepared this module

This module connects research on choices, workplace performance, learning design and adoption with a company’s product announcement. It turns those ideas into a written decision exercise. The reporting example and the time-payback numbers are made up for illustration.

Sources