Picture a finance team of three ending the week with every recurring report prepared before anyone asked, and time left for the owner’s Wednesday question that needs an answer by Thursday. That is a team doing the impossible with the people it has, and it takes AI working two ways at once.
Very few jobs are one repeatable process. A controller, an office manager or an account lead spends part of each week on work that arrives the same way every time and part on work nobody has done before. The repeatable part becomes workflows your people run with AI. The ad hoc part gets better when people make good use of the AI the company already pays for.
Repeatable work becomes workflows.
Repeatable work arrives on a schedule or a trigger, draws on the same inputs and has a result you can write down. A workflow holds exactly that: the steps, the team’s own rules, the checks and the person who decides. Built once and tested on real cases, it runs for everyone in the role, and the best way to do the work becomes the normal way.
This is where connectors matter most. A weekly workflow reads the same business systems every time, so a connection set up once pays back on every run, and the work comes back ready to check instead of ready to start. We use the system’s own connection first, then a proven third party’s, then one we build. Fully Powered Connectors explains how, and how you decide what each one can see and change.
Ad hoc work gets more from the AI you already have.
Ad hoc work is different every time: a question from the owner, an unusual complaint, a notice from a regulator, the first draft of something nobody has written before. A workflow for a task that never comes back costs more than it returns. Here the gain comes from people using ChatGPT, Claude, Copilot or Gemini well. They give it the context it needs, ask it to test their reasoning, check every figure it produces and keep the decision for themselves. The connections your AI platform already offers, such as files and web search, are usually enough.
Watch the ad hoc pile. A question that arrives three months running is repeatable work nobody has written down yet.
The sorting sheet.
List one team’s work from last week with the people who did it, and answer five questions for each item. Copy the sheet and fill in the brackets.
Sorting sheet: [team name], week of [date]
- Does it come round on a schedule or a trigger you can name? Every Monday, every new order, every month-end.
- Do the inputs come from the same places each time? The same system, report, folder or inbox.
- Could you write down what a good result looks like? Two people would produce much the same thing.
- Does more than one person do it, or does it pass between teams?
- Is the same check made before it goes out? The same reviewer, standard or sign-off each time.
One team’s week, sorted in two.
One invented team’s week, sorted by the people who did the work.
Illustrative example with invented data. A three-person finance team at an invented 70-person wholesale distributor. One week of its work, sorted with the sheet.
| Work this week | Sort | How AI fits |
|---|---|---|
| Receivables summary for Monday’s leadership meeting | Repeatable: five yes | A workflow from the weekly receivables summary template drafts it from the aged receivables. The controller checks it. |
| Tuesday’s overdue-invoice follow-ups | Repeatable: five yes | A workflow from overdue-invoice follow-up preparation drafts one message per customer. The credit clerk edits and sends. |
| Short payments and disputed lines | Repeatable: four yes | A workflow from receivables exception review sorts each exception and names its owner. The controller decides any write-off. |
| Supplier bills checked before Thursday’s payment run | Repeatable: four yes | A workflow matches each bill to its order and delivery and lists the mismatches. The controller approves the run. |
| The owner asks what a warehouse rent increase does to next quarter’s cash | Ad hoc: one yes | The controller gives the AI the cash forecast and asks it to test her assumptions. She decides what to tell the owner. |
| An auditor’s one-off question about a revenue entry | Ad hoc: none | The accounts assistant drafts the reply with the AI and checks every figure against the ledger. |
| A new sales-tax notice | Ad hoc: one yes | The AI lists what the notice asks of the company. The controller confirms it with the tax adviser. |
Four of the seven items went into the team’s plan as workflows. All four read the accounting system every week, so that is where a connection earns its keep. The other three stay with the people who handle them, using the AI they already have.
Put the sort in your plan.
Decide where your company is going with AI. Then bring your people there, one workflow at a time. The Fully Powered Planner starts with your teams and about five workflows each team should be practising, offered as suggestions to accept, change or reject. It sorts each team’s work into the repeatable kind and the ad hoc kind, exactly as the sheet does. Teams and roles are named for the work, and nobody is rated or ranked. The plan is annual, and the sort is redone as the work changes.
To choose which repeatable item to build first, score the pile with the opportunity scorecard. To define the first one on a single page, use the sheet in 10X the return on the AI you already pay for.
All of it runs on the platform your team already has: ChatGPT, Claude, Copilot or Gemini. We ask the people who do the work, nothing records or monitors them, and your organization keeps control of how the work gets done. That is how you 10X the ROI of your existing AI.
Bring one team’s week and sort it in your plan.
Seven-day trial with a card, everything unlocked, nothing charged for seven days. See every plan and price. Prefer to do it together? Book a fit call.