If someone on your team spends every Friday copying information from emails into a spreadsheet, ask why the information has to be entered twice. Maybe customers could submit it through a form. Maybe two systems you already pay for can share it. Or maybe the emails arrive in so many different formats that AI could help pull out the details.
Those are three different solutions. To choose among them, watch the work happen before you buy anything.
Start with a task that costs you something
Repetitive work shows up in more places than payroll. A quote goes out a day late. A customer inquiry sits in an inbox because nobody knows who owns it. Someone copies the wrong address into a work order, and a crew has to turn around.
Ask your team which task they repeat every week that creates the most delay or rework. Pick one and watch someone do it from beginning to end. Note where the information comes from, what the finished work should look like, and where the employee has to make a judgment.
Then measure how often the task happens, how much time it takes, and how often someone has to correct it. That gives you a baseline for deciding whether a change is worth making.
Use the simplest tool that fits the work
Suppose you run a service business. A customer emails to request an estimate, and an employee reads the message, enters the details into your system, and assigns the request to the right person.
If customers are willing to use a short form, the form can collect what you need, and a simple rule can route the request. If most requests still come in as emails, AI could help extract the location, requested service, and preferred date for someone to check. Flag missing information so your team knows what to ask the customer.
I’d assign one person to own any price, deadline, or promise that goes out to a customer. Start with sample or redacted messages, and check what information the tool is allowed to receive before using live records.
Measure the whole job, including corrections
Say your team handles 40 requests a week, and each one takes three minutes to enter. That’s two hours of work. If the new process cuts data entry but still takes 90 minutes to review and correct, you’ve recovered 30 minutes a week before counting setup, maintenance, or software costs.
That might still be worthwhile if customers get a faster response or fewer requests fall through the cracks. Decide what improvement would justify keeping the tool, and measure that alongside the time saved.
Assign one person to own the trial. Run it on one workflow for a defined period, track the exceptions, and keep the old process available in case the tool fails. At the end, ask the employee doing the work whether the change made it easier to deliver a good result. They’ll often spot problems a time estimate misses.
The first step is simple: pick one task, time it, and see what you can remove before you automate what’s left. If the trial works, you’ll have evidence to guide the next one. If it doesn’t, you’ll have learned something about the process without committing the whole business to a new tool.
If you’ve picked a workflow to improve and want help figuring out what to automate, submit an inquiry through Tech Services Done Right. Describe the task, how often your team does it, and where it gets stuck so the team can assess whether it can help.
Disclosure: I have an ownership interest in Tech Services Done Right and may benefit financially if you use its services.