Services
Explore
Free AI assessmentAnswer 15 questions about one process and see where AI could help.
Start an assessmentWhere to start with AI: a practical first-steps guide
One way to begin: choose a suitable task, agree what data and checking it needs, test it on real work and decide from the result.
What to avoid
Common early mistakes, and how to avoid them
A common pattern is to buy a tool first and then look for uses for it. That makes results hard to judge, because nothing was measured before the tool arrived. Starting from a task you want to improve gives you something to measure.
One practical way to start with AI is to choose a specific, repetitive task that takes your team time each week and can be checked easily. Agree what data may be used and who reviews the output, test a tool on real examples, and measure the result against the current process before deciding whether to expand, change approach or stop.
Buying licences before deciding what they are for often leads to low use and no clear way to judge the value.
Running several trials at the same time splits attention and makes it hard to tell which change produced which result.
Without a baseline, such as how long the task takes now or how often errors occur, you cannot tell whether the AI approach is better.
Introducing AI changes how people work as well as which software they use. Involve the people who do the task, as well as whoever sets up the technology.
One approach
A four-step way to start
- List your repeated tasks
Write down the tasks your team repeats each week. Look for work that involves a lot of reading or writing, follows clear rules, or takes more time than its difficulty suggests. These are your candidates.
- Score by benefit and effort
For each candidate, estimate how much time it might save and how hard it would be to change. Also note what data it uses and who would check the output. Pick a low-stakes task with a good balance of benefit and effort.
- Test one task
Agree what data the tool may use and who reviews the output, then test it on real examples. Record the time taken, including checking and corrections, and compare it with the current process.
- Decide what to do next
Share the result with the team. If the task improved enough to justify the cost and checking, keep it and consider the next candidate. If not, change the approach, try ordinary automation or stop.
After the first test
What progress can look like after a few months
A business that follows this approach might, after a few months, have one or two tools in regular use on specific tasks, measured against how those tasks were done before. The team will have a clearer sense of what the tools can and cannot do, and a short list of further tasks worth testing.
The approach is repeatable: choose a task, test it, measure it, then decide. Some tests will show that the current process or simple automation is good enough, and that is a useful result too.
Common questions
Questions about getting started
Do we need a strategy before we start?
A long strategy document is not needed to run a first test. You need one task you want to improve, an agreement on what data can be used and who checks the output, and a way to measure the result. A wider plan can follow from what you learn.
What if we're not sure which process to start with?
Use the scoring in step two to compare candidates. If you want help, an AI discovery workshop with your team can list and compare candidate tasks and agree which to test first.
How long does it take to go from zero to something live?
For an off-the-shelf tool used on a simple task, such as a writing assistant, you can start testing within days. For custom agents and automations, the time depends on the systems and data involved and is agreed when the work is scoped.
Should we use an off-the-shelf tool or build something custom?
Try off-the-shelf tools first where they fit the task. Some tasks are handled well enough by existing tools or ordinary automation. Consider a custom build when the task justifies the extra cost of building and maintaining it.
How do we get our team on board?
Start with the people whose work will change most. Ask them which tasks to improve and involve them in testing. People who are included early are more likely to use the tools and report problems.
What does working with Reformat Labs look like in practice?
We usually start with an AI discovery workshop with your team, looking at how work moves through the business and agreeing which task to test first. We then build and test it with the people who do the work, hand it over with documentation and review the result with you.
Talk to us
Want help applying this to your business?
Bring the task that takes your team the most time. We reply within one working day.