Start with one piece of work that is frequent, repetitive and easy to measure, such as client intake, a weekly report or answering staff questions from your own documents. Get that single workflow running, measure what it saves, and use the result to decide what comes next. You do not need an AI strategy or clean data first.
Do you need an AI strategy before you start?
No. A strategy written before anything is running is mostly guesswork, because you don't yet know what AI is worth in your business. One working workflow gives you real numbers, and the strategy can follow from those.
The same goes for data. Most companies' information is spread across email, shared drives and a few core systems. That is normal. Often the first project is the one that pulls it together.
What makes a good first project?
Look for work that happens often, follows the same steps each time, and ends in something a person on your team can check.
Much of this work is now done by AI agents, software that can carry out a task across your systems. Common first projects include answering staff questions from the company's own documents, client or customer intake, a weekly or monthly report assembled from several systems, and checking documents or applications for missing items. Our guide on what to automate with AI first has a scoring checklist.
What are the steps?
- Write the goal in plain words. For example, "we want month-end close to take two days."
- Choose the one workflow closest to that goal.
- Measure it as it is today. How long it takes, how many people it involves, and how often it happens.
- Build and launch it. The people who do the work should test it before it goes into daily use.
- Measure again after a month. Then choose the next workflow.
For the math, see how to measure the ROI of AI automation.
Should you do it yourself or bring someone in?
If you have engineers, you can try a first workflow in-house. Most established businesses don't, and the work stalls between a busy IT team and a tool nobody has time to set up.
Soren AI is an AI-native engineering firm that helps established businesses catch up on AI quickly and at lower cost: our engineers work with AI agents to build, run and maintain AI inside the systems a business already uses. We bring the engineers, do the build and stay on to run it, at a lower cost than a traditional consulting project. We compare every option in the fastest, lowest-cost way to catch up on AI. The usual first step is a 30-minute AI transformation assessment, which ends with a one-page roadmap.
What mistakes should you avoid?
- Buying a tool before choosing the problem.
- Starting with the hardest, most visible process. Save it for later, once the team has seen one workflow succeed.
- Launching without a number to compare against.
- Leaving out the people who do the work. They know where the process really breaks.
We cover what goes wrong in more detail in why AI pilots die after the demo.
Find your first workflow
Start with a 30-minute call with our team. We'll map how your business runs and show you where AI would give the most back.
Frequently asked questions
We're worried about falling behind on AI. How fast do we need to move?
Sooner than a long planning cycle allows, and you don't have to do everything at once. One working workflow in the next month will teach you more than a year of planning. From there you can add one workflow at a time, at a pace your team can absorb.
How long does a first AI project take?
Weeks, if it is one workflow. At Soren the first workflow is typically live within two to three weeks of getting access to the systems it needs.
How much should we budget?
It depends on scope. Start with one workflow so the first commitment is small, and expand once you have seen a result.
What if our team is not technical?
That is the usual case. Your team needs to know how the work is done today. The engineering can come from outside.
Is our company too small for this?
If a task takes several people several hours every week, it is usually worth a look.


