An AI tool might promise to save time, reduce waste or make a service easier to use. The useful question is what would improve if you introduced it, and for whom.
A community garden, for example, might consider AI to help decide when to water its plants. Less wasted water would be a benefit. So would making the watering schedule easier for volunteers to manage. The decision becomes more complicated if the tool needs expensive equipment or volunteers spend more time checking its advice than they previously spent planning the watering.
This worksheet helps a team decide whether AI would be useful for a particular task. Start by writing what you would ask it to do, such as suggesting when to water the garden. Answer the five questions below with the people who would use the tool or be affected by it. Then decide together: should you test the tool, change how you plan to use it, solve the problem another way or learn more before deciding?
1. What do we want to improve or make possible?
Start by explaining what is difficult now and what you want to make better. For the garden, that might be: “We sometimes water the same area twice because volunteers do not know what others have already done. We want a reliable watering schedule that keeps the plants healthy and avoids wasting water.” Once the problem is clear, you can consider whether AI would help solve it.
Decide how you will tell whether the tool helps. The garden could keep a record of how much water it uses, whether the plants stay healthy and how long volunteers spend watering, checking the AI’s advice and looking after the equipment. Compare these records with how the garden worked before. Using less water would be useful, but not if the plants suffer or volunteers have to do much more work.
The goal could also be to do something the garden cannot currently manage, such as growing vegetables for a community kitchen that needs to offer more meals. If the kitchen needs those vegetables from July, include that in the goal. The group can then ask whether AI would help it plan the expansion and meet that need in time.
Write down: What you want to improve or make possible, who it would help and when it needs to happen.
2. Who would benefit, and whose work might become harder?
Include the people who will use the tool, maintain it, correct its mistakes and rely on its results. Ask them what would change in their day.
Saving time is one useful outcome. People may also value having a say, learning a skill, feeling confident in their work or spending time together. Harvard’s Human Flourishing Program considers several dimensions of wellbeing, including health, meaning and relationships. This broader view helps explain why productivity alone is an incomplete picture of how people are doing. Harvard Human Flourishing Program.
For the garden, AI might save the coordinator time by creating a watering schedule. But if its suggestions often need correcting, volunteers may spend longer checking the soil and adjusting the schedule than they did before. The coordinator’s job has become easier while the volunteers’ work has increased. Ask both groups whether the change is helping.
Write down: Whose work would become easier? Who would have new tasks? Have you asked everyone who would need to use, check or maintain the tool what they think?
3. What resources would it require?
Write down what you would need to buy, what you would pay each month and what people would have to do. Include installing equipment, entering information, learning to use the tool, checking its advice, correcting errors and maintaining it. If the project includes training or adapting an AI model with your own data, include the time, expertise and computing costs for that work too. Ask the company offering the tool which tasks it handles and which would be your responsibility.
For environmental questions, consider the electricity, water and equipment involved in providing the service, alongside any resources it might help save. Our guide to AI’s electricity and water use explains what these figures can tell you.
Record water use, spending and volunteer hours separately. For example, if the garden saves water but volunteers spend more time fixing problems, write down both results. The group can then discuss whether the water saving is worth the extra work and expense. Combining everything into one score would make that trade-off harder to see.
Write down: What you need to buy, what each person would do and which questions you still need the company offering the AI tool to answer.
4. What else could meet the same need?
Consider at least one other way to solve the problem. The garden could start with a shared schedule showing who waters each area and when. Other options might include a watering timer or a device that measures how wet the soil is and switches watering on when needed.
Compare the options by asking the same practical questions about each one: Would it keep the plants healthy? Could it reduce wasted water? How much work would volunteers have to do? For example, a simple watering schedule might solve the problem of watering an area twice. An AI tool would need to offer enough additional help to justify its cost and the time spent checking and maintaining it.
Write down: Another way you could solve the problem, what each option would cost and why you would prefer one.
5. Who will decide, and when will we reconsider?
Agree who will make the final decision and how volunteers can report a problem. If you test the tool, choose a date to discuss how it went. Also agree on problems that would mean stopping the test sooner.
For the garden, that might mean stopping if plants are not getting enough water, volunteers cannot change an unsuitable watering recommendation or repairs and checks take more time than anyone can give.
The test should help answer a clear question: “Are we keeping the plants healthy and using less water without adding more work than our volunteers can manage?”
Write down: Who will make the decision, when you will discuss the results and which problems would make you stop or change the way you use the tool.
A worksheet to copy
Start with a sentence describing the decision. Complete the table together, writing “unknown” where information is missing. Beside any numbers or promised benefits, note where the information came from and whether you have checked it yourselves.
Decision: ____________________
People taking part: ____________________
Date: ____________________
| Question | Our answer |
|---|---|
| What do we want to improve or make possible, by when, and how will we know we have done it? | |
| Whose work would become easier, who would have new tasks and have we asked those people what they think? | |
| What would we need to buy or pay for? What work, data, electricity and water would be needed? | |
| How else could we solve the problem, and why would we choose one option over another? | |
| Who will decide, when will we discuss the results and which problems would make us stop or change the tool? |
Information still needed: ____________________
Who will find out: ____________________
Our next step: Test the tool on a small task / Change how we would use it / Solve the problem another way / Find out more first / Decide not to use it
Reason for that choice: ____________________
Person responsible and review date: ____________________
How the garden might use it
The following is a fictional example of a completed decision.
The garden is considering a service that uses AI to suggest watering times from soil readings and weather information. Volunteers would check the suggestions and decide when to water. They complete the worksheet together:
| Question | What the volunteers write |
|---|---|
| What do we want to improve? | We want to stop watering the same beds twice, keep our plants healthy and spend less time working out what needs watering. |
| Who would it affect? | All of us who water or tend the plots, and our coordinator. A ready-made schedule could help, but we need to ask the people on each shift how much extra checking they could take on. We also share the water supply with other gardeners. |
| What would we need? | We would need soil sensors, an internet connection and a paid account for the AI service. Someone would have to install the sensors, keep them working and check the advice. We still need to ask the company about the full cost, who fixes faults and how much electricity and water its service uses. |
| What else could we try? | We could use a shared watering schedule and mark each bed when it has been watered. We could also look at a simple device that waters when the soil gets dry. We want to compare the cost and work before choosing. |
| Who will decide, and when? | Our coordinator will make the decision after hearing from the volunteers on each shift. If we test the tool, we will meet after four weeks to discuss it. We will stop sooner if plants are suffering or checking and fixing the system takes more time than we can give. |
Their next step: “We will ask the company our questions before buying anything. If the answers address our concerns and we agree the tool is worth testing, we will try it on a small part of the garden. We will keep checking the plants ourselves and be ready to return to our usual watering schedule.”
During the test, the volunteers would record water use, plant health and all the time they spend watering, checking advice and fixing equipment. They would also note rainfall and differences between the plots, so they do not mistake wetter weather for a saving caused by AI. They could use the four-week review to decide what to do next and check again as the seasons change.
When the answer should change
Suppose the garden uses less water during the test, but volunteers regularly have to correct the suggested watering times and one person spends their evenings fixing sensor problems. The coordinator spends less time planning, but the volunteers are doing more work overall.
The volunteers might decide: “We are glad we are using less water, but this is taking too much of our time. We will pause the AI tool and try the shared schedule first. We can reconsider it if the advice becomes more reliable and someone can support the equipment.”
Their decision recognises both the water saving and the work required to achieve it.
Could AI help the garden provide food for more people?
Imagine a local community kitchen wants to offer an extra weekly meal from July because more families are asking for help with food. Its budget for fresh vegetables is already stretched. It asks whether the garden could expand to supply some of what it needs.
The volunteers want to help, but they have limited growing space, water, money and time. They need to work out which vegetables they could grow, when those vegetables would be ready and whether they could keep supplying the kitchen through the summer.
AI could help compare planting plans against those limits, with experienced gardeners double checking the AI recommendations.
The volunteers might write: “More families need these meals, and we want to help. We need a planting plan we can actually manage, with vegetables ready when the kitchen needs them. If AI helps us work out how to do that with the space, water and people we have, it could be useful.”
The group could compare this with a plan prepared by experienced gardeners using a planting calendar. It would judge the result by whether it delivers useful vegetables when the kitchen needs them, within a budget and workload the volunteers can sustain.
Where possible AI interests fit
Interflourishing also asks how people should respond if AI systems develop experiences of their own. Long and colleagues argue for preparing for possible AI welfare while uncertainty remains about whether systems could have morally significant experiences or interests. Long and colleagues, 2024.
If your proposal raises that question, record the concern and the expertise needed to examine it. Keep it distinct from the effects on people and the environment. In the garden example, the group would record that possible AI experience has not been assessed.
Additional question, where relevant: Is there a reason to investigate possible AI interests in this particular system, and who could help?
Using the worksheet well
Keep the completed version, including disagreements, so people can see why a decision was made. Return to it when the tool, the people using it or the circumstances change.
The approach connects with NIST’s AI Risk Management Framework, which covers responsibilities, environmental impacts, alternatives and decisions about whether to proceed. The five questions here bring those practical concerns together with Interflourishing’s focus on relationships and wellbeing. NIST AI RMF 1.0.
This is a proposed discussion tool from Interflourishing. We welcome feedback on whether the questions helped a team notice something important or left a concern out. Share suggestions through our contact page, without including private information about other people.
Related reading: What is interflourishing?, AI’s electricity and water use and editorial standards.