Maya gets four hours back. Sam gets a nicer email.

Bit: This conversation is written and produced by The Human Bit. Maya and Sam are the guests who share the common workplace experiences and reviewed research with us.

Sam: Maya has AI doing half her Monday before she has finished her coffee.

Maya: I have been here since seven.

Sam: That is not helping your case.

Bit: What exactly did it do?

Maya: It compared four team updates with last week, found two missing sections, flagged three numbers that changed and prepared the first draft of the Monday report.

Sam: See? Half her Monday.

Maya: It also moved our Melbourne project to Melbourne, Florida.

Sam: A small geographical difference.

Maya: Only if you are paying for the flights.

Bit: And what did you use AI for this morning, Sam?

Sam: I asked it to make an email sound less annoyed.

Maya: Did it work?

Sam: It replaced “as discussed” with “as previously discussed”.

Bit: So, more annoyed.

Sam: More annoyed, but in a waistcoat.

Maya: Did it add kindly?

Sam: Twice.

Bit: This is The Human Bit. Today, three people with access to roughly the same kind of technology are having very different Mondays.

Sam: Two people and one Bit.

Bit: I was counting generously.

Maya: Do not encourage him.

Bit: Sam is not refusing AI. He uses it most days. Maya is not running a company from a hammock while twelve agents negotiate with each other.

Maya: The hammock remains on the roadmap.

Bit: But Maya is getting something more useful from it. The question is why. Is she better with technology? Does she have better tools? Is her work easier to automate? Or did she notice something Sam has not noticed yet?

Sam: I have noticed many things.

Maya: You noticed kindly.

Sam: It was used aggressively.

Bit: Maya, what did Monday reporting look like before this?

Maya: Four team updates. Two came through Teams. One lived in a spreadsheet. One arrived as a voice note from someone who believes punctuation is an administrative burden.

Sam: I know who that is.

Maya: Everyone knows who that is.

Maya: I would collect everything, compare it with the previous week, check the numbers, work out what had genuinely changed, chase what was missing, then write the report. About four hours on a good Monday.

Sam: And now an agent does it.

Maya: No.

Sam: A workflow.

Maya: Better.

Sam: Built with code.

Maya: Not at the beginning. At the beginning I copied everything into one conversation and asked for the weekly report.

Sam: And it worked.

Maya: It was magnificent. Beautiful headings. Clear risks. Excellent grammar. Completely unusable.

Bit: What was wrong?

Maya: It trusted every number it saw. One team had copied an old figure. Another had changed the reporting period. The tool blended them together and wrote a confident paragraph about growth that had not happened.

Sam: That feels important.

Maya: It also treated “no new issue” as “issue resolved”. Those are very different sentences in operations.

Bit: So you improved the prompt.

Maya: See, that is exactly how people make this sound easier than it was.

Bit: Go on.

Maya: The prompt was not the main problem. The four teams did not agree on what “on track” meant. One meant the deadline had not moved. One meant the deadline had moved but nobody had told leadership yet. One meant they felt optimistic.

Sam: The recognised corporate definitions.

Maya: Exactly. The AI did not create that disagreement. It just put all three versions beside each other where I could no longer pretend they matched.

Bit: So the first useful thing it did was expose the work.

Maya: Yes. Then the humans had a meeting. A deeply advanced technology. We agreed what the statuses meant, where the source numbers came from and which information the tool was allowed to use. After that, I could build something repeatable.

Sam: How repeatable?

Maya: It prepares the comparison. It shows evidence for each change. It marks missing information instead of guessing. It drafts the report. I check every number, every risk and anything that could affect a person or a commitment.

Sam: And Melbourne?

Maya: Still in Victoria when I last checked.

Sam: This is the bit that makes me think I am miles behind. You are talking about sources, comparisons and repeatable processes. I am negotiating tone with a text box.

Bit: For what it is worth, the evidence is much kinder to you than the internet is.

Sam: Finally.

Bit: Pew asked workers who had used AI chatbots at work what they actually did with them. Editing and drafting written content were among the most common answers. Analysis or coding was much less common.

Sam: So I am mainstream.

Maya: You are statistically annoyed.

Sam: I feel represented.

Bit: Most people are not running fleets of agents while they sleep. Many are researching, editing, drafting and trying to work out whether the result actually saved them anything.

Maya: Which should remove the embarrassment. It does not answer whether Sam wants more from it.

Sam: I do. I just cannot see how you get from “make this sound polite” to “prepare my Monday”.

Maya: Neither could I at first.

Sam: You?

Maya: For months I used it for emails, summaries and presentation notes. I had a collection of prompts that made mediocre writing arrive faster.

Sam: Why have you never mentioned this?

Maya: Because nobody puts “spent three months making bullet points slightly better” in the success story.

Bit: Sam, take the email from this morning. What was it about?

Sam: A client project is late. Not catastrophically late. Late enough that the client has stopped asking for dates and started asking whether we are confident.

Maya: Which means they are not asking whether you are confident.

Sam: They are asking whether they should still trust us.

Bit: And you used AI to write the answer.

Sam: I wrote the answer. I used AI to remove the part where I sounded frustrated with our own team.

Maya: Before you wrote it, what did you do?

Sam: I called the delivery lead.

Maya: Why?

Sam: Because the project plan said Friday and his last update said early next week.

Maya: What else?

Sam: I checked the contract. I went back through the last two meeting notes. I found a promise we made verbally that was not in the project plan.

Bit: What promise?

Sam: That we would give them forty eight hours to review before launch. The new timeline had launch on the same day as handover.

Maya: So the project was later than the delivery team thought.

Sam: Yes. Then I had to decide whether to lead with the missed date, the review problem or the recovery plan.

Bit: And after all of that, you invited AI in to adjust the tone.

Sam: When you say it like that, it sounds slightly ridiculous.

Maya: Not ridiculous. Safe. The email is visible. You know what good writing looks like. You can compare the result quickly.

Sam: The rest is sensitive. Contracts, meeting notes, client promises. I cannot just throw it into whatever tool happens to be open.

Maya: You should not.

Sam: That was quick.

Maya: Because “use AI on the whole task” is not the answer. The question is whether there is approved preparation it can do before you make the decision.

Bit: For example?

Maya: If the business has an approved tool, it could compare the project plan with the meeting notes. It could list every commitment and show where it found it. It could mark dates that do not match. It could prepare the questions Sam needs answered before he calls the client.

Sam: It cannot decide what I say first.

Maya: No.

Sam: It cannot know that this client forgives a delay but hates finding out late.

Maya: No.

Sam: And it is not on the call when they go quiet.

Bit: That sounds like the part that remains yours.

Sam: It sounds like the part I am paid for.

Maya: I think so. You have just been spending a lot of the morning assembling the evidence before you can do it.

Sam: There is still a gap here. You knew which tool was approved. You knew what information it could see. I have been told to “use AI more” and given no definition of more.

Maya: That is fair.

Sam: You also had four hours on a Monday that were clearly attached to one report. My work arrives through calls, emails and people appearing beside my desk saying, “Quick one”.

Maya: That is also fair.

Bit: And it is not only the two of you. Gallup found that unclear use cases, privacy concerns and lack of training were among the main barriers to workplace AI use. People who felt actively supported by their manager were about twice as likely to use AI frequently.

Sam: So “use AI more” is not manager support.

Maya: No. It is a wish.

Sam: A wish with performance review energy.

Maya: My manager gave me something useful. She said, “Use this approved tool, start with the weekly report, do not use personal information, and show me the draft before anything changes.”

Sam: That is a completely different instruction.

Maya: Yes. I still did the experimenting. I still made mistakes. But I was not guessing whether the experiment itself would get me into trouble.

Bit: So some of the gap is skill. Some is practice. Some is permission.

Maya: And some is having a task you can safely test more than once.

Sam: Which means I am not allowed to leave this conversation blaming my subscription.

Maya: You can blame it a little. It is expensive.

Sam: Can I ask the question nobody enjoys?

Maya: Go on.

Sam: If I show the business how to remove two hours from my work, what happens to the two hours?

Bit: What do you think happens?

Sam: Best version, I spend it on clients. Realistic version, I get two more hours of work. Bad version, someone decides the job needed fewer hours because it needed less of me.

Maya: All three are possible.

Sam: Thank you for the warmth.

Maya: I had the same thought. The first time the Monday process worked, I showed my manager. She looked at it for about thirty seconds and asked which other reports we could automate.

Sam: Of course she did.

Maya: I remember thinking, maybe I should have enjoyed the four hours quietly.

Bit: What did you say?

Maya: I said we could prepare more reports this way. I also said the preparation was not the decision. The tool could show us the risk had appeared for six weeks. It could not make anyone deal with it.

Sam: Did anyone deal with it?

Maya: Eventually. Because I used some of the saved time to trace it properly and put the right people in a room.

Bit: So the time became better work.

Maya: In that case. Do not turn it into a promise. Saved time is just capacity. Management decides whether capacity becomes better work, more work or fewer people.

Sam: That is the first honest thing I have heard anyone say about productivity in months.

Bit: There is another uncomfortable part. Keeping the repetitive work because it fills your day is not reliable protection either.

Sam: I know. I just dislike how sensible that sounds.

Maya: You should. Sensible things are often very badly timed.

Sam: So the goal is not to prove I am busy. It is to be clearer about where I change the result.

Bit: And clearer about what the organisation intends to do with the capacity it asks you to create.

Sam: That second conversation sounds harder than the prompt.

Maya: It is.

Bit: What would be a sensible first experiment for Sam?

Sam: Not the live client problem.

Maya: Agreed. Use something finished. A client update you have already sent, with information you are allowed to use in the approved tool.

Sam: And ask it to rewrite the email.

Maya: No. Look behind the email.

Sam: There is going to be homework, is there not?

Bit: Three questions.

Sam: Of course there are.

Bit: Before you could send that update, what had to happen?

Sam: Gather the notes. Check the dates. Find the commitments. Talk to delivery.

Bit: Which of those parts is repeated preparation rather than judgement?

Sam: Comparing dates. Pulling every commitment into one place. Marking what is still unanswered.

Bit: And which decision or responsibility must remain yours?

Sam: What the client needs to hear first. What I am willing to promise. The call itself.

Maya: That is the experiment. Not build an agent. Not transform client services. See whether the tool can prepare the evidence without pretending it owns the relationship.

Sam: And if it does a terrible job?

Maya: Then you learn on a completed example instead of during a client crisis.

Sam: You sound irritatingly experienced.

Maya: I sent Melbourne to Florida so you would not have to.

Bit: One week later. Sam, what happened?

Sam: It found three unanswered questions in the meeting notes.

Maya: Useful?

Sam: Very. It also found a commitment I had forgotten because it was buried halfway through the previous meeting record.

Bit: Did you trust it?

Sam: I checked every item against the notes. One was wrong. Two were right. The forgotten commitment was definitely ours.

Maya: And the email?

Sam: Shorter. Clearer. I called the client before I sent it.

Bit: What changed?

Sam: I did not use AI to hide the difficult part in better language. I used it to find the difficult part before I spoke to them.

Maya: Did it say the project remained on track?

Sam: It suggested that. The evidence disagreed.

Bit: And which one won?

Sam: The evidence. I am new, not reckless.

Maya: Did the final email include kindly?

Sam: No. Some decisions should remain human.

Bit: That feels suspiciously prepared.

Sam: I have been waiting a week.

Maya: Fair.

Bit: Maya did not get here by learning every tool. Sam did not move forward by becoming Maya. They each looked more closely at the work they were already doing.

Sam: I still only have one useful workflow.

Maya: That is one more than last week.

Bit: And that is enough for a Monday. This is The Human Bit. AI helps. You decide.