According to Asana's Anatomy of Work Index, published in April 2026 from a survey of more than 10,000 knowledge workers worldwide, 60% of their time goes to "work about work" — communicating about a task, searching for information, switching between apps, managing priorities that keep shifting, and tracking the status of something. Not the work itself. Everything that surrounds it.
Atlassian reached a similar number from the other side of the table: in its State of Teams 2025 report, a survey of 12,000 knowledge workers and 200 executives, leaders and teams said they lose 25% of their time just searching for answers — a question already answered before, a file that already exists somewhere, a decision already made that no one remembers where it was recorded.
And Microsoft, in its 2023 Work Trend Index, measured the same problem from the complaint side: 62% of people say they struggle with too much time spent searching for information during the workday, while the average employee spends 57% of their time communicating — meetings, email, chat — versus 43% actually creating something: a document, a spreadsheet, a presentation.
Three surveys, three different methodologies, the same picture: most of the workday isn't spent on the task listed in the job description. It's spent on the friction around it. And that friction is exactly what most companies overlook when deciding where AI should show up first.
Where the Time Actually Goes
When a company decides "let's use AI", the first question tends to be "what will impress people the most" — an assistant that talks to customers, a system that writes entire pieces of text, an agent that promises to decide on its own. An agent is an assistant that carries out a task in steps, not just one that answers a question — and it can be genuinely useful, but that's rarely where the team's time actually leaks out.
The time leaks out somewhere more mundane: someone opens five tabs to find the right version of a spreadsheet. Someone sends the same question to the team chat, again, because the earlier answer got buried in a conversation from three weeks ago. Someone redoes a report because they didn't know one already existed, built by someone else, sitting in a place nobody mentioned.
None of these moments is dramatic on its own. Together, they add up to the 60% of "work about work" Asana measured and the 25% of searching Atlassian measured — and they explain why 62% of people, in Microsoft's survey, say they feel that weight every day. It's time that shows up in no productivity report because no one logs "looked for a file" as a line item of work — but that's exactly where it disappears.
Why Companies Pick the Wrong Task
The task that becomes an internal headline — "we launched an AI assistant for customers", "AI now writes our proposals" — is almost never the one that solves the problem measured above. It solves a different problem: looking good at a board meeting, answering a competitor who already announced something similar, justifying the investment with a flashy demo.
The effect is that the company spends time and political capital on a visible, showy task, while the routine that actually wears people down — searching, comparing, updating, notifying — stays exactly as it was. In the end, the team feels like "the company uses AI" while still burning a quarter of the day hunting for an answer that already existed.
The boring routine never becomes a case study in a slide deck, because there's nothing spectacular about it: it's just a search that got instant, a form that started filling itself in, a repeated question that now has a ready answer. But that's exactly where the math on lost time is most favorable — because that's where the time was actually going.
What Has to Be in Place
Picking the right routine is only half the job. The other half is giving the company a way to reuse what works without every team rebuilding the same thing from scratch.
Ready-made process templates for common routines. Most of the repeated tasks in a department — reconciliation, request triage, standard replies, document checks — already follow a known shape. Starting from a ready template, instead of designing from zero, is what turns a months-long pilot into a routine solved in a week.
Reuse with permission by person, group, and role. When one person builds a good way of doing something, the gain stays locked to them until another person, or another group, with the right permission, can use the same thing without rebuilding it. Reach is defined by role — it isn't "share everything with everyone".
Visible cost per conversation, to compare before and after. Without a visible number for how much each interaction costs, there's no way to prove the automated routine is cheaper than the manual search it replaced. Visible cost turns a vague sense of improvement into a number any manager can defend.
Human approval when the routine touches something sensitive. Not every boring routine is low-risk — updating a tax record, for instance, can be repetitive and still require confirmation before it goes through. The rule isn't "it's boring, so automate everything"; it's "boring and low-risk, automate it; boring and risky, automate it with approval built into the middle".
That's why Skyller was built this way: a place where one person's boring routine becomes a reusable template for the whole team, with permission, visible cost, and approval wherever the risk calls for it.
The Payoff of Choosing the Boring Routine
The payoff from starting with the boring routine is different from the payoff of starting with the flashy task — and it lasts longer. A flashy automation impresses once and then becomes one more AI project that "didn't take off". A solved boring routine saves time every single week, quietly, and that time disappears from the complaint list before it ever becomes a headline.
There's also a trust effect that builds without effort: when AI fixes the file-searching problem that irritated everyone for months, nobody needs to be convinced to trust it for the next routine. Adoption grows from the bottom up, pulled by the person who felt the relief, instead of pushed from the top down by an announcement.
For a small team — and most companies in Brazil and Latin America run lean, with each person covering more than one role — recovering even a quarter of the time lost to searching and rework isn't a footnote. It's the difference between hiring one more person or giving the current team the time it should already have had.
A Roadmap to Get Started
Before deciding where AI should show up first, it's worth asking the team itself where the time actually goes.
- Ask each person to log, for three days, how much time they spent searching for something that already existed. The number tends to surprise more than any vendor demo.
- Identify the question that keeps repeating in the team chat every month. If the same doubt comes up again and again, it already has an answer — it just needs one fixed place to live.
- Pick the most boring routine before the flashiest one. It's counterintuitive, but that's where the recovered time is largest and the political risk of getting it wrong is smallest.
- Measure the cost before and after, per conversation or per task. A comparable number is what turns "feels better" into "is saving X per month".
- Let the solved routine become a template someone else can use. One person's gain only becomes the company's gain once someone else can reuse it, with the right permission.
