People often start using AI before anyone has said what it is for. Each team answers for itself, and when the leadership team asks whether AI is working, there is nothing agreed to measure against. The aim decides what gets measured, where money and time go first, and what people are told about why their work is changing.
Choose one or two. Several of them draw on the same saved time: quality spends some of it on review, capacity commits it to more work, and efficiency pays off only once someone names what the time is for.
EfficiencyLink to this section
Efficiency means the work people already do, done in less time. Drafting, summarising and searching go faster, and gains show early. Check that faster work is still accurate.
Easy to miss. Many leaders expect time saved to become money saved. Savings come in small amounts across many people's days, and without a plan they go on other things. A working paper on Danish workers in eleven occupations highly exposed to AI chatbots found that, in late 2024, those who had used chatbots for work reported time savings averaging about 3% of their work hours, and 85% of those saving time expected to put more time into other tasks (National Bureau of Economic Research, "Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI", 2026). That is a little over an hour of a full-time week, and it is worth something only once the organisation decides what that hour is for.
When to choose it. It makes sense where the leadership team decides in advance what the saved time is for, such as fewer overtime or contractor hours, and tells staff, some of whom will wonder whether saved time means fewer jobs.
QualityLink to this section
Quality means fewer errors and a more consistent standard of work. People use AI to draft, check and improve their work, then review what it produces, which is new work needing time and skill. The value, less rework and fewer complaints, appears more slowly than time saved.
Easy to miss. AI writes fluently whether it is right or wrong, so it lacks the awkward phrasing that often shows a draft is weak, and a hurried review can miss errors harder to spot than the ones it removed. The gain depends on a checker who knows the subject, and the checking time comes out of whatever AI saved.
When to choose it. Choose it where errors are costly, today's rate of rework or complaints is measured before starting, and review is a named step with a person answerable for it.
GrowthLink to this section
Growth means new services, or more clients, made possible by AI. Client-facing staff use AI in work clients see, such as proposals and advice, and need the confidence to explain it when asked. The return is new income, less predictable than savings, because a new service takes time to design, price and sell.
Easy to miss. Growth usually means AI in client work, which brings in client information, accuracy and contracts, some of which limit where that information may go. Pricing needs thought too: where the organisation charges by the hour, faster work earns less unless the price changes.
When to choose it. It is worth choosing where the risk appetite allows client-facing use, and someone is responsible for each new service from design to pricing.
CapacityLink to this section
Capacity means more work handled by the people the organisation already has, with AI doing part of it. It shows as shorter waiting times or a smaller backlog, while staff costs can stay level.
Easy to miss. Efficiency saves time on today's work and leaves open what happens to it. Capacity commits that time to more work, often before the savings are proven, so where AI helps less than expected, people have to do the extra work themselves.
When to choose it. It suits organisations where real demand is waiting, such as a backlog or a waiting list. Raise the volume in steps, where the saving has shown, and watch workloads and errors.
Better work for peopleLink to this section
Better work for people means less repetitive work, and more time for the parts of the job people value. People choose, with their managers, which tasks AI takes on first, and some will ask for tools the organisation can't yet provide. Ask regularly whether AI has taken those tasks on, and count the answers at renewal.
Easy to miss. Better work for people has a large effect on whether the tools get used: people keep using AI that takes away work they dislike, and gradually stop using AI that adds work. It shows first in continued use, and later can show in who stays.
When to choose it. It is best chosen alongside another aim, because it gives staff a reason of their own to put time into the rest.
How to chooseLink to this section
- What would the leadership team point to in a year to say AI was worth it?
- Is work already waiting that the organisation can't get to? If so, capacity may fit. If not, efficiency, with a named use for the time.
- When two aims compete for the same hours, which one wins?
Tensions with other decisionsLink to this section
- Growth with Safe first. Safe first keeps AI off work involving client information, which covers most of what growth needs, so the two conflict. Move to Cautious, where client work is allowed with a person checking it, or choose another aim.
- Growth with client and customer information off limits. AI still helps with structure and drafting, while work on a client's own material stays manual. Check that is enough, or allow client information in approved tools with clear policies.
- Capacity at Carefully, over time. Extra work that starts before AI is used for it gets done without AI, in the same working hours. Take on more only where AI is in use, or quicken the pace.
- Efficiency and better work for people with a Budget tied to results. Better work shows slowly, so at each funding decision it tends to get less money than efficiency. Agree how to measure it before any money is allocated, or fund it separately.
When to revisit itLink to this section
- After the first few months, once the organisation's own work gives evidence.
- When the backlog a capacity aim was chosen for is cleared, or saved time has gone unused.
- When the risk appetite or pace changes, or people say the tools are adding work.
If the leadership team would like support with this decision, see Getting outside help.
Something missing or out of date here? Tell us.