Ask how long a job takes and most people will tell you how long the doing takes. Sending the invoice: two minutes. Ordering the materials: five. Replying to the customer: ten.

But that’s rarely where the time goes. Before the ten-minute job comes the hunting: finding the latest version, checking whether someone has already handled it, chasing the person who knows, and rebuilding enough context to act. That part doesn’t show up on anyone’s timesheet. It just quietly eats the day.

Why nobody measures it

Checking is spread thin. A few minutes here, a few there, across every person and every system. No single moment feels expensive, so nobody adds it up. The work gets done, eventually, and the cost stays invisible.

Doing the rough maths

Pick one recurring task. Not a whole department, just one thing that happens often. Then ask three questions:

• How many places does someone check before they can act?
• How long does the checking take, compared with the doing?
• How often does it happen, and how many people does it touch?

The number won’t be precise. It doesn’t need to be. The point is to see where the time is actually going, so you can decide whether it’s worth redesigning.

What changes when the work comes to you

The fix is rarely “work faster.” It’s making the right work visible to the right person, with the context already attached, so the checking step shrinks or disappears. Sometimes that’s a clearer process. Sometimes it’s better configuration of the software you already own. Sometimes it’s automation or a supervised AI agent doing the digging first.

If you’d like a second pair of eyes on the maths, that’s exactly what the free Conversation is for.

Chasing money is awkward, so it gets put off. And when it does happen, it usually starts with a hunt: which invoice, which version, what the customer said last time, whether anyone has already called.

That hunt is often the real reason follow-ups slip. It isn’t reluctance. It’s that every follow-up means rebuilding the story from scratch.

Design the follow-up, not the reminder

A reminder system on its own just sends emails. A designed follow-up process makes sure the right person sees the right invoice at the right time, with the history attached, and decides what happens next.

• Overdue work is visible without anyone going looking for it.
• The history comes with it: what was sent, what the customer said, what’s been tried.
• Routine steps can be prepared in advance, so a person reviews and sends rather than writes from scratch.
• Anything unusual goes to a person. Disputes, payment plans and relationship calls stay human.

Why it keeps relationships intact

When follow-ups are consistent and well-informed, they stop feeling personal. Nobody has to be the bad guy, because the process is doing the remembering. The human conversation, when it’s needed, starts with the full picture.

In most businesses, one person holds the operation together. They know which job is waiting on what, which customer needs a call back and which supplier is running late. Usually, they know it because they went looking.

That knowledge is valuable. The looking isn’t. Every hour spent checking five places to understand one job is an hour not spent on the judgement calls only they can make.

The risk of the person-shaped system

When the process lives in one person’s head, the business works, until that person is sick, on leave or overloaded. Then things slip, and nobody can see why until it’s urgent.

What “the work comes to them” means

• A clear view each morning of priorities, deadlines and missing information.
• Unfinished work organised by owner, blocker and next action.
• Repetitive tasks routed to whoever should do them, with the context attached.
• Their judgement kept where it matters: exceptions, decisions and people.

Sometimes the answer includes extra hands, including offshore support, for work that genuinely needs a person. Often it starts with making the existing work visible. Either way, the goal is the same: less hunting, more doing.

AI is useful for exactly the kind of work that eats time without needing judgement: gathering information, summarising what’s happened, drafting the first version. It’s also capable of being confidently wrong. Both things are true, which is why supervision matters.

Where an agent earns its place

• Investigating a job and summarising where it’s up to.
• Preparing a follow-up with the relevant history attached.
• Drafting replies, notes and updates for a person to review.
• Surfacing work that needs attention, and explaining why.

Where a person stays in charge

• Decisions and approvals.
• Anything involving money, customers or risk.
• Exceptions: anything that doesn’t fit the usual pattern.
• Checking the output before it goes anywhere that matters.

A well-designed agent doesn’t remove the need to check. It changes what gets checked: instead of hunting for information, a person reviews prepared work and makes the call.

AI is one tool, not the answer

Plenty of friction is better solved with a clearer process, better software configuration or a person. An agent makes sense when the work is repetitive, the information is available and the boundaries are clear. Deciding that comes before building anything.