Where did the hours go? Your first AI job is working. Picture a 24-person business where the AI job drafts the weekly progress update to customers, and your sales team checks and sends each one.

Your sales manager can see the task has got shorter. What she can’t see is where the time went. It slid into longer email threads, extra report tweaks and whatever landed on the desk first.

Unless someone claims them, saved hours get absorbed by whatever is nearest, and you can’t show the team or yourself that the job paid. What follows is the method: count the net hours honestly, send them to one named place, and give that decision an owner.

Count the net hours, not the headline

Count the hours after the new checking work is included. The useful figure is the time really returned to the business, not the time the tool takes to produce a draft.

Start with the old task as people did it: gathering the updates, writing the email, correcting it and sending it. Then time the same task with the AI step in it. If someone in the sales team reads every draft against the order notes, that time belongs in the new figure.

Here’s how the sum works. Say six people in the sales team each spent 90 minutes a week writing customer progress updates, which is 9 hours across the office. If checking and sending the drafts now takes each of them 30 minutes, the net saving is 6 hours a week, not the 9 a demonstration might suggest.

Measure a few weeks of ordinary work so the awkward cases are in the figure, such as the customer who wants every detail spelled out. Use records the business already keeps where you can, because a new counting system adds work of its own. Measuring a soft win helps when the benefit is visible but the figure isn’t, and keeps a quality check beside the time figure.

Write the net saving against the job and note how you worked it out. That stops a rough estimate turning into a promise that grows every time someone repeats it.

Give the hours one destination

Give the saved time a destination while the change in the week is still visible. When a repeated task gets shorter and nobody sets a new priority, the space fills with whatever is closest. Emails get longer, meetings spread and small requests move into the gap. Six weeks later nobody can say whether the AI job created any lasting capacity.

Put the allocation where the work is managed. If the saving sits with the sales team, change their diaries or the call list. If it sits with the accounts assistant, reserve part of her Thursday for the chosen work. A note in a project file won’t change anybody’s Tuesday.

Choose one destination first. Splitting those 6 hours three ways makes each share easy to lose. Once the first allocation has produced something you can see, you can change it.

Send the hours to customers when replies wait

Send the time back to customers when response or follow-up is the constraint you can already see. If the real problem is too few new enquiries, spare hours on call-backs won’t fix it. That constraint might be website enquiries waiting all morning for a call back. It might be customers who haven’t heard from anyone in a fortnight, or quotes nobody followed up after they went out.

Name the customer action and write it into the process. For customer update calls, decide which customers are on the list, who calls them, how often and what counts as done. The hours then have a queue of their own rather than a general instruction to improve service.

Keep the measure close to the action. Check call-back times from the call log and completed customer calls in your CRM (customer relationship management system). Don’t claim the time won you more orders, because price, the market and the person making the call all play a part too. Report what you can see: calls made that weren’t being made before.

Send the hours to quality when checks slip

Send the time to quality when important checking or tidying is being skipped because the week is full. You’ll know where these gaps are. In a busy office they tend to show up as a price list that doesn’t match what customers are being quoted. Others are duplicate customer records, or order files missing documents someone meant to chase.

Choose a named quality job rather than telling the team to be more thorough. Check the price list against live quotes once a month, clear the duplicate records, or chase the missing documents on open orders. Give the work a definition of done and a regular slot in the diary.

Protect the checking the AI job itself created before you allocate anything else. Reading drafts against the order notes and handling the customer who complains are part of the new process. Only the time left after that is free for a separate quality job.

Give waiting work a first finish line

Use the hours for waiting work when the business has a job that matters but has never had anyone with time to start it. This suits a business with a real backlog, not a list of ideas nobody has chosen between.

That might be the quotes that never turned into orders, the past customers who once asked about more work, or template letters that still describe last year’s process. Match the work to the shape of the saving. An hour a week from each person in the sales team can work through old quotes steadily. It can’t fund a project that needs whole days unless someone gathers the time into one block.

Write down the first finish line before moving any hours. “Follow up old quotes” gives nobody a place to start. “Call every quote from the last six months that didn’t turn into an order, and log the reason” is done when the last call is logged.

Don’t turn every saving straight into another AI job. The hours may do more good fixing the information around the first one. Once that job is settled, moving on to the next job gives you the test for choosing another.

Make one person own the hours

The person who runs that part of the business decides where the hours go. Here that’s usually the sales manager or the operations lead. The people in the sales team whose task changed confirm whether the plan fits the real week.

That owner records the destination, changes the diary or the call list, and checks that the chosen work is happening. Without one name against it, the allocation drifts back into the general week.

The sales team need a say because the hours may not arrive in one useful block. A spare 20 minutes here and there across five days can’t fund the same work as a free Friday morning. They’ll also be first to notice if the saving shrinks.

Review the allocation alongside the AI job itself, once a month to start with. If checking time rises, the net saving has changed. If the customer call list is clear, the hours need a new home.

Answer the redundancy question out loud

Once a job gets shorter, someone on the team will wonder whether their role is next. Answer it before they have to ask. I believe AI adoption should be about efficiency, not headcount, and the clearest way to show that is to name where the hours are going.

The tool can’t decide whether freed time becomes better service, cleaner records or fewer people. That choice belongs to you, and avoiding it is still a choice, because the time gets absorbed somewhere. Be precise with the team: name the shorter task, the checking that remains, the net time released and the work now receiving it.

If a role could be affected rather than a task, say so plainly and follow the proper process. The redundancy question covers that conversation and the difference between a task and a role.

Put the hours in the diary this week

On Monday, sit down with your sales manager for 45 minutes. Time the old task and the new one, including the checking, and write the net figure against the job.

Then pick one destination (customers, quality or waiting work) and write its first action in a sentence. Before Friday, put the time into the sales team’s diaries or onto the call list. Tell the team where it’s going, and book a review in four weeks.