What we learned
The hours that never show up in any report
A team's time doesn't go into the big tasks: it goes into copying, renaming, forwarding and answering the same thing again. Before automating, you have to see it.

Edmundo SpohrOctober 14, 2026 · 5 min read
In 30 seconds
- 01Lost time hides in small handoffs: copying, renaming, forwarding and answering the same thing again.
- 02Count it before you automate: often it's enough to stop doing something or change a form.
- 03Automate what repeats the same way. What varies or is sensitive stays with a person.
When we ask a team where their time goes, they almost never name a big task. They name a feeling: “we can't keep up”, “we spend all day putting out fires”. And when we look closely, the time doesn't go into the important work. It goes into what happens between one task and the next.
Where the time really goes
At a foundation that works with volunteers, one person ran the whole selection process: reviewing every application, downloading the documents, renaming them one by one, assigning interviewers and sending two rounds of emails. None of it showed up in any report. But it was almost all of their job.
At the same organization, coordinating each program's WhatsApp groups took about two hours a day. Attendance came in with names typed by hand, so it couldn't be added up. And reimbursements were copied from one form to another before being uploaded to the bank.
What the team thinks it does
- Select volunteers
- Support the programs
- Report to the board
What actually takes up the day
- Renaming files
- Forwarding the same message to several groups
- Copying data from one form to another
- Answering the same question again
The three questions that arrive a hundred times
Another silent leak: repeated questions. At that foundation, direct messages were almost always the same three: how do I recover my password, where do I continue my application, why wasn't I selected. They tried an automatic reply on social media and it helped a little. Before that they had tried a chatbot with a fixed menu, and dropped it because it was too rigid.
What helped was realizing they were three questions, not a hundred. Three questions can be answered before they arrive: on the screen where people get stuck, in the confirmation email, in a message that goes out on its own at the right moment.
When the platform adds clicks instead of removing them
Sometimes the problem is the tool itself. At an organization that works with preschools, the team ended up logging their progress in a separate spreadsheet because reaching a preschool's record in the platform took dozens of clicks. Nobody decided to abandon the platform: they simply stopped using it, quietly.
We saw something similar in a reading program. Changing a target from 20 to 21 readings required calling a developer. Every small adjustment turned into a request, and the team stopped asking for them.
Go deeper
We pay for tools nobody uses
Many organizations have more tools than they use. The problem is almost never the tool: it's that nobody adopted it.
When AI doesn't save time
The same volunteer foundation tried using artificial intelligence to rank candidates and check documents. It worked well with standard documents, but failed with foreign documents and badly taken photos. And just testing it cost money for a month. To check a document that is always the same, a fixed rule plus a person reviewing the exceptions was faster and cheaper.
It's not that AI is useless. It's that it's useful for interpreting what varies, not for repeating what is always the same.
What to automate and what to leave with a person
A simple rule
- 1
It repeats the same way
Renaming files, sending the reminder, moving data from one form to another: automate it with a rule.
- 2
It repeats, but varies
Sorting open-ended answers, summarizing conversations: AI can help there, with someone reviewing.
- 3
It's sensitive
Telling an applicant no, supporting someone through a hard situation: it stays with a person.
One week to see it
Before buying or building anything, suggest a one-week exercise to the team. At the end of each day, everyone writes down the tasks they did more than once and roughly how long they took. Nothing sophisticated: a shared sheet is enough.
At the end of the week, put the lists together. Most likely three or four tasks will show up eating a good part of the time. That list is your first project. And sometimes the solution isn't technology: it's stopping something, merging two forms or answering a question before it arrives.

Edmundo Spohr
Applied AI, adoption and impact · Sumadots