AI workforce planning sounds like a big-company exercise, but for a small business it can be one page: your team as an org chart, with AI drawn in where it helps and a named person checking each tool.
In this lesson
- How to draw your team as it is today and list each role’s main tasks.
- How to mark where AI fits, using the four outcomes from Lesson 4.
- How to redraw the chart with AI attached to the roles that use it, and a named person checking each one.
- The three settings in our team planner, from assisting the team you have to asking how far the data says AI could go.
- What changes for your people, and how to talk to them about it.
Why draw it at all
An org chart is the quickest way to see a business whole. It shows who does what, who reports to whom, and where the knowledge sits. Drawing AI onto that chart does two useful things. It forces you to say exactly which role each AI tool serves, and it makes you name the person responsible for checking it. A tool with no name next to it is a tool nobody is checking.
Step 1. Draw today’s chart
Start with what you have. A box for each role, the number of people in it, and lines for who reports to whom. Keep it rough: a piece of paper is fine. If one person does two jobs, as owners of small firms usually do, give each job its own box with the same name in both.
Step 2. List each role’s top tasks
For each box, write the five to eight tasks that take up most of the time. Ask the people in the role; they will give you a truer list than the job description. If you get stuck, look up the role on this site: each job page lists the official US task descriptions for that job, rated AI does it, AI helps or needs a human. The AI team planner does this for a whole team at once.
Step 3. Mark each task
Use the five-question test and the four outcomes from Lesson 4:
- Hand it over, with spot checks
- AI drafts, a person checks
- AI helps when needed
- Keep it human
Be strict with the last one. Anything physical, anything where the customer expects a person, and anything a law or regulator expects a named person to do or sign stays human.
Step 4. Redraw the chart
Now draw the same chart again, and hang the AI off the roles it serves. In the diagram below, people are amber boxes and AI tools are grey boxes. Each grey box has a small amber tag naming the person who checks or approves its work.
Before: the owner; an office manager with two receptionists and a bookkeeper; a sales lead with three sales reps; an operations lead with four technicians. 14 people. After: the same chart with one receptionist instead of two, 13 people, and four AI tools, each with a named person: an AI phone answerer checked by the receptionist, AI receipt capture checked by the bookkeeper, an AI drafting assistant checked by the sales lead, and an AI scheduling agent whose actions the operations lead approves.
- Owner
- Office manager
- Receptionist ×2
- Bookkeeper
- Sales lead
- Sales rep ×3
- Operations lead
- Technician ×4
- Owner
- Office manager
- Receptionist ×1AI phone answererchecks: Receptionist
- BookkeeperAI receipt capturechecks: Bookkeeper
- Sales leadAI drafting assistantchecks: Sales lead
- Sales rep ×3
- Operations leadAI scheduling agentapproves: Ops lead
- Technician ×4
A few things to notice in the “after” chart:
- The AI tools sit under people, not beside them. The phone answerer serves the reception role, and the receptionist checks its work: which calls it handled, which it passed on, what it told callers.
- Every AI tool has a named checker. “Checks: Bookkeeper” means the bookkeeper reviews what the receipt capture tool put in the accounts. “Approves: Ops lead” means the scheduling agent proposes and the operations lead says yes before anything is sent to customers or technicians.
- The technicians’ boxes do not change. Their work is on site and hands-on. AI might help with the paperwork around it (quotes, job notes, follow-ups), but the work itself stays with them.
The checkpoint
Every arrow from an AI tool to a customer, supplier or official body passes through a person first, unless the task is one you have deliberately handed over with spot checks. This is the most important design choice in the whole chart.
Flow: a task goes to AI, which drafts it. A person checks the draft. If it is not right, the person sends it back to the AI with a note, and the loop repeats. Only checked work goes on to the customer.
The person in the checkpoint is not a rubber stamp. They can send the work back, fix it, or do it themselves. They also notice patterns: the same mistake every week, a kind of request the tool always gets wrong. That knowledge is how the setup gets better.
Three settings for AI workforce planning
How far you take this is your choice. Our AI team planner has a switch with three settings, so you can see what each would mean for your own roles, from live data.
- Low: assist the team. Nothing is handed over fully. AI helps your people with the tasks our data rates AI does it and AI helps, and the planner shows the hours a week that might free up. The chart stays the same shape, with AI tools attached.
- Medium: hand over and reshape. Tasks rated AI does it are handed over with spot checks, and AI drafts the AI helps tasks for a person to check. Some roles end up with a lot of free time, or mostly checking work, and the planner flags them so you can decide how to reshape them.
- Scorched earth: replace everyone you can. The far end. Everything our data rates AI does it or AI helps goes to AI, and the planner shows the smallest team our data supports today. A year setting (today, 2030, 2035, 2040) moves more tasks to AI as each job’s replacement range passes, so you can see how that might change. It is there because people ask the question, and because the honest answer is useful: even at this setting, the planner lists the work in your team that our data says still needs a person, and every AI task keeps someone checking it.
The same illustrative business at all three settings:
The same 14-person business at three settings. Low, assist the team: all 14 people stay, each role gets an AI assistant, and 60 to 90 hours a week are freed. Medium, hand over and reshape: 10 to 12 people's worth of task time remains, 2 to 4 people's worth is freed; one receptionist with an AI phone answerer, the bookkeeper at part of one person's time with AI receipt capture, two sales reps with AI drafts checked by the sales lead, an AI admin assistant checked by the office manager, an AI scheduling agent approved by the operations lead, and four technicians. Scorched earth, replace everyone you can: 6 to 8 people's worth of task time remains, 6 to 8 is freed; the owner, an office manager who also covers reception and books with three AI tools, one to two sales staff with an AI sales agent, and four technicians with an AI scheduling agent approved by the owner. What still needs a human: on-site work, client calls, sign-off, and checking every AI's output. Technicians stay at every setting because their work is physical.
14 peopletoday's team, unchanged
- Owner
- Office manager+ AI
- Receptionist ×2+ AI
- Bookkeeper+ AI
- Sales lead+ AI
- Sales rep ×3+ AI
- Operations lead+ AI
- Technician ×4
Hours freed: 60–90 a week
10–12people's worth of task time, from 14
- Owner
- Office managerAI adminchecks: Office manager
- Receptionist ×1AI phonechecks: Receptionist
- Bookkeeper part of one person's timeAI bookschecks: Bookkeeper
- Sales lead
- Sales rep ×2AI draftschecks: Sales lead
- Operations leadAI schedulingapproves: Ops lead
- Technician ×4
Freed: 2–4 people's worth of task time
6–8people's worth of task time, from 14
- Owner signs off
- Office manager also covers reception and booksAI phoneAI booksAI adminchecks: Office manager
- Sales ×1–2AI sales agentchecks: Sales
- Technician ×4AI schedulingapproves: Owner
Freed: 6–8 people's worth of task time
An illustrative example, not a real business. Figures here are made up to show the shape; the planner computes yours from live data, always as ranges.
What the diagram shows, and the planner’s sums build in, is that the far setting does not take you to zero: every task handed to AI still needs someone to check it. The technicians stay in every column because their work is physical. The people who are left take on new work: checking AI output, handling the exceptions it passes up, and keeping customer relationships. Lesson 10 covers what you risk if you cut too far, and the legal steps that come first if roles change. Lesson 12 has a plan for each setting.
Try it with this business
The planner is below. Open the illustrative business in it with the link underneath, switch the setting to see how the task time moves, then replace the roles with your own.
Or try an example: Trades firm, 12 people Dental practice, 8 people Online shop, 4 people SEO agency, 9 people Web development agency, 11 people Marketing team, 8 people
More example teams (28)
Digital and marketing: Creative agency, 9 people Video production company, 6 people Software startup, 12 people IT support company, 8 people
Professional services: Accounting firm, 10 people Law firm, 8 people Recruitment agency, 8 people Real estate agency, 7 people Insurance broker, 7 people Financial advice firm, 7 people Consultancy, 8 people Translation agency, 7 people Architecture practice, 6 people Customer service center, 14 people
Health and care: Family doctor’s practice, 12 people Care home, 19 people Physical therapy clinic, 6 people
Hospitality and retail: Restaurant, 17 people Hotel, 13 people Hair salon, 9 people Retail store, 11 people
Trades, transport and industry: Construction company, 13 people Auto repair shop, 8 people Warehouse and delivery firm, 18 people Metal fabrication shop, 8 people Cleaning company, 13 people
Education and community: Elementary school, 15 people Nonprofit, 7 people
What to do first
- Pick two or three of the tasks at the top of the list and try an AI tool on them for a month.
- Keep a person checking everything AI drafts before it reaches a customer.
- Write down the time it saves and what it gets wrong.
Reshaping roles
- Roles that end up mostly checking AI’s work could be combined or given new work.
- Plan around natural turnover and retraining before anything else.
- Tell your team early what you are trying and why.
Before you go this far
- The smallest team leaves no cover for sickness, holidays or a busy month.
- What your people know about your customers and your work leaves with them.
- AI mistakes reach customers faster when fewer people check.
- Letting people go has legal steps (notice, consultation, fair selection) that differ between US states and the UK. Take advice first.
How we date the replacement range · Talking about AI at work
At any setting
- AI makes mistakes that look right. A person checks anything that reaches a customer.
- Privacy and data rules apply to AI tools too.
- These are ranges from stated assumptions and typical tasks for each job, not a forecast for your business.
The assumptions behind these figures
- Low: AI saves 50–80% of the time on tasks rated AI does it, and 20–40% on tasks rated AI helps, while people keep doing them.
- Medium: tasks rated AI does it are handed over, with spot checks of 5–15% of the time; on AI helps tasks AI saves 40–70% and checking each draft takes 10–20%.
- Scorched earth: AI does both kinds end to end; a person still reviews 30–60% of AI helps tasks and spot-checks 5–15% of the rest.
- Scorched earth in a later year: each job’s Still needs a human score falls towards 15 along the timeline on its job page, and the same share of each label’s time moves one step towards AI (needs a human to AI helps, AI helps to AI does it). A person still checks all of the AI’s work.
- A 40-hour week. Ranges are rounded outward and never below a role’s share of tasks that need a human.
Open this business in the team planner
The roles used, each with its own page and task list:
What changes for your people
Adding AI to a chart is the easy part. The people in the boxes are what make it work.
- More time on the human tasks. The point of handing over routine work is to give people time for the work only they can do: the difficult customer, the new client, the job that needs judgment. Decide in advance what that time is for, or it will fill with more routine work.
- New checking and supervising work. Someone has to review AI output, spot-check the handed-over tasks and handle what the AI passes on. That is real work and needs time in the week. Count it.
- Training. People need to learn how to brief an AI tool, how to check its work and when not to trust it. The less experienced staff may gain the most (Lesson 3), but only if they also learn to spot its mistakes.
- Roles that shift. If one role ends up mostly checking work, talk to the person about what the role could become. Natural turnover, retraining and combining roles are all options. Lesson 12 has a plan for this.
Talking to your staff
Your team will have questions, and many will be worried. Our guide How to talk to your manager about AI is written for them; it is worth reading to see the questions they are likely to bring you. From your side of the table:
- Say what you know and what you do not. “We are trying AI on two tasks for six weeks to see if it saves time” is better than silence, which breeds rumor.
- Involve the people who do the work. They know which tasks are tedious and where mistakes hide. Ask them to help pick the first tasks and to be the checkers.
- Be clear about the rules. Which tools are approved, what data must never go in, and who checks what. Lesson 10 has a policy you can adapt.
- Talk about roles, not just tools. People want to know what their job will look like in a year. If you do not know yet, say when you will.
- Share the results. If a pilot saves time, say what the time went on. If it did not work, say that too.
Key terms
- Human in the loopA named person who checks or approves AI’s work before it takes effect, and answers for the result.Lesson 10: Risks and rules · More in the glossary
- AgentAn AI given a goal, tools and permission to take several steps on its own, such as booking, filling in forms or updating records.Because agents act, their mistakes have effects. Start with a person approving every action.Lesson 7: From chatbot to agent · More in the glossary
- Task timeHow a job’s working hours divide between its tasks. Tasks that matter more and come up more often get a bigger share.More in the glossary
- Task splitA job’s working time sorted three ways: tasks AI does, tasks AI helps a person with, and tasks that need a human.The three shares add up to 100%. Amber is always the human share.More in the glossary
Every term is explained in the glossary.
Try this
Draw your team as it is today, on one page: a box per role, with the number of people in it. Under each box write the role’s top five tasks, and mark each with one of the four outcomes from Lesson 4. Then draw it again with an AI box under each role that would use one, and write in amber (or circle) the name of the person who would check it. If you cannot name a checker for a box, that tool is not ready to use. Then enter the same roles in the AI team planner and compare its task lists with yours.
Sources
This lesson uses no outside figures. The example business is illustrative. Task ratings in the planner and on the job cards are our own, printed live from the current release.