Is AI helping or hindering our decision fatigue?

Firms are already handing AI some of the small work: first drafts of emails, meeting notes, file summaries, the reformatting at the end of a long document. People in law and accounting practices have been working out which of those jobs AI does well, and for many of them it has taken real weight off the day.

Along with the work, AI takes some of the small decisions off their plates. How to word the reply to an invoice query. Which parts of the meeting go in the notes. What order the file summary should run in.

In 1973, Alan Lakein's How to Get Control of Your Time and Your Life sorted tasks into A, B and C: A for the ones that matter most, B for the ones that can wait, and C for the small ones. If AI takes on more of the Bs and Cs, a person's day fills up with more of the As.

So should we be making more A decisions and letting AI take care of the B and C tasks? But what does that do to a person who spends the whole day on big decisions?

If you plan how work runs in a law or accounting firm, these questions may already be showing up in how the week gets organised.

1What decision fatigue means

Decision fatigue is the idea that making decision after decision wears down the quality of the decisions that come after. In its best-known form, each of us has a certain number of good decisions in us each day, and every choice uses some of them up. The idea went mainstream in August 2011, when John Tierney wrote about it in the New York Times Magazine.

In 2012, Barack Obama told the writer Michael Lewis in Vanity Fair: "You'll see I wear only gray or blue suits. I'm trying to pare down decisions. I don't want to make decisions about what I'm eating or wearing. Because I have too many other decisions to make." Two years later, Mark Zuckerberg told a public Q&A that he wore the same grey T-shirt every day because he wanted to make "as few decisions as possible about anything except how to best serve this community", as CNBC reported.

Some of the most cited evidence comes from medicine. In 2014, Jeffrey Linder and colleagues looked at 21,867 visits for respiratory infections across 23 primary care practices in the United States, published in JAMA Internal Medicine. Clinicians prescribed antibiotics steadily more often as a four-hour session went on. By the fourth hour, the adjusted odds of a prescription were 26% higher than in the first.

A study of Australian GPs found a similar pattern on a much larger scale. Mona Maier and colleagues analysed 262,456 consultations with 2,909 GPs and found that for every 15 more patients a GP had seen, the odds of an antibiotic prescription rose by 8.7%, while prescriptions for statins and osteoporosis medication fell.

Neither team measured fatigue directly, and the time of day could also stand for ordinary tiredness, running late, or the kind of patient who books a late appointment.

The laboratory experiments behind the idea of a daily supply of willpower have been repeated in large studies across dozens of labs, with the methods registered in advance, and the effect shrank to very little or to nothing (Hagger and colleagues, 2016; Vohs and colleagues, 2021). In 2025, Swedish researchers tested the idea on 231,076 real calls to a nurse advice line, and the data favoured no change in the nurses' decisions across a shift or after a break. A review of 82 studies of healthcare professionals found that fewer than half of the tests supported decision fatigue. Even the famous study of Israeli parole judges, whose rulings seemed to turn harsher as a food break approached, is still argued over, because the cases were not heard in random order.

So where does that leave it? Long stretches of demanding work do wear people down, and some professional decisions have been seen to change as a session goes on. The tidy version, with a fixed number of good decisions a day, has not held up. A firm can look at how the working day is set up around those decisions.

2A day of A decisions

Say a senior associate in a commercial team gets an afternoon back. The first draft of the letter is done, the file summary is ready, and the notes from the morning's client meeting are already in the matter. What's left on her list are the calls that need her judgement: the advice to sign off, the risk to weigh up, the figure she will put her name to.

The closest evidence on a day of big decisions comes from auditors. Bart Dierynck and Christian Peters ran three experiments with more than 350 professional auditors, published in Contemporary Accounting Research in 2026. In the first two, auditors who started with the easier task generally did worse overall than those who started with the harder one. In the third, where the auditors chose their own order, they tended to start with the easier task, and more so under time pressure.

Stress has its own effect. A 2016 meta-analysis by Katrin Starcke and Matthias Brand found stress tends to push people towards riskier decisions, particularly where the risk is a poor bet. In a review of the research on sleep, Yvonne Harrison and James Horne found that sleep loss hit hardest the decisions that need fresh thinking: dealing with the unexpected, changing a plan when new information arrives, and weighing up risk. Logical, rule-based reasoning held up better. A 2024 meta-analysis of 44 studies found that even one short night left people sleepier and more prone to lapses in attention, though it found no clear effect on working memory.

In a firm with a heavy season, or a big matter running late, the same person can be under time pressure, short on sleep and making the biggest calls of the week.

AI adds a few small calls to that load. Each time a person opens an AI tool, there are decisions to make first: which one to use, what information is allowed in, and how far to trust what comes back. Checking the result is a decision too, and a demanding one. We looked at the research on that checking load in What a four-minute job costs, and at who does the checking in Are you the human in the loop?

When AI makes room in the day, what should the room be for?

3More decisions, or better ones

With AI, should we be making room so people can make more decisions, or so they can make better ones? The research can't settle it yet, and it is probably a bit of both.

Giving each decision more thought adds the most value where one judgement carries the client's outcome: the advice on a dispute, the structure of a deal, the position taken on a tax return. In a randomised experiment with 273 business professionals working on a time-pressured market-entry decision, Matteo Cristofaro and colleagues found that how people used AI shaped the quality of the decision. They saw the strongest thinking and the fewest biases at a moderate level of AI use, where people compared and questioned what the AI gave them. Very light and very heavy use both did worse.

The room can also go to more decisions. A firm with clients waiting, or a team short of people, may need the extra capacity.

Or some of the time could go somewhere else altogether, such as training junior staff: sitting with them on a file, talking through why a call was made, reviewing their work together. As AI takes on more of the Bs and Cs, juniors may see fewer of the routine tasks that build judgement.

What does it look like when a firm gets this right? People know which decisions are theirs and which the firm has already settled. The hardest call of the day gets a fresh hour. The small AI calls are made once, for the whole team. People have time to recover between big decisions, and the firm has decided what the room AI makes is for, whether that is more client work or time back for the team.

Each of those is set by the firm as a whole, through the tools it approves, the way it allocates work and what it decides the saved time is for.

4Planning for decision load

5Where we come in

Our investigation maps how your firm's work runs today and measures how long it takes, including where the judgement calls sit and who they land on. The findings belong to your firm whether or not we go any further together.

If you would like to see the decision load in your firm, and what AI is changing about it, reach out to our team. The first conversation works out whether your firm is the right fit for the AI Uplift Program, and whether we are the right people to help.

Planning for decision load