How do I get my team to use AI in a way that creates business value?

You bought the licenses. Your team attended the training. People occasionally use AI to write an email, summarize a meeting, or answer a question.

The business still runs the same way.

The same people move information between the same applications. The same handoffs slow things down. Your most experienced employees still spend too much time preparing work and too little time applying their expertise.

Getting more value from AI starts with changing how the work gets done.

Reinvent the workflow with AI, and design it with the people who use it.

1. Choose one workflow and define the business outcome.
2. Map the friction with the people who run it.
3. Redesign the work with clear AI and human responsibilities.
4. Test, review, and improve it together.
5. Measure net effort, quality, and the business result, including review and corrections.

That gives you two things your investment needs: a process capable of producing a better result and a team that has a reason to use it.

Buying access and providing training are useful starting points. The next step is putting that capability to work inside a specific business process.

Start with the people who run it.

Bring together the employees who do the work, handle the exceptions, and receive the output. Walk through a real example from beginning to end.

Where does the information come from? Who copies it somewhere else? Where does work wait? What has to be corrected? Which decisions depend on knowledge that lives in someone’s head?

Mine the process for friction, issues, and blockers. Give people room to explain what actually happens, including the workarounds that never made it into the procedure.

Then redesign it together.

Those who write the plan don’t fight the plan.

People who help build a process understand why it works the way it does. They can point out where a proposed improvement would create more work. They can explain what good output looks like and recognize when something is missing.

Their involvement improves the design and gives them ownership of the change.

An estimating workflow shows what this looks like.

In one pool construction business, the work involved standardizing estimating information, documenting the process, and building reusable templates that showed AI what a good estimate should look like.

The team’s knowledge mattered throughout. Experienced estimators knew which sources to trust, how to interpret the work, and which assumptions needed attention. That knowledge had to become part of the workflow.

With that context in place, AI could support the review of takeoffs and drawings and prepare estimating work for a person to check and approve.

The reported result was an 87% reduction in estimating time, bringing it down to 13% of the time previously required.

See how the estimating workflow was redesigned.

That result came from work on the information, the process, and the role of the estimator. The AI had useful context. The person had prepared work to evaluate.

This is a different design problem from traditional software automation.

Traditional automation works well when you can define the sequence and rules in advance: when this happens, do that. Those rules still have a place. Calculations, permissions, and required approvals should remain dependable.

AI adds the ability to interpret information that does not arrive in a neat, predictable format. It can work with context, compare possibilities, and propose a next step without someone specifying every possible branch beforehand.

That flexibility creates an opportunity to reconsider the sequence itself.

Consider prospecting. A salesperson might spend hours gathering company information, sorting through records, deciding which prospects fit, and preparing outreach.

A redesigned workflow could use AI to evaluate a large collection of prospect information against the team’s criteria, surface a promising group, and explain the evidence behind each recommendation. The salesperson can challenge the selections, add relationship knowledge, and work with AI to prepare relevant outreach.

The person’s attention moves toward deciding where to invest effort and how to start a useful conversation.

Giving AI context makes that work more useful. It also gives the person something concrete to check. A recommendation should come with enough supporting information to judge whether it makes sense.

The design question becomes: how much preparation can the machine handle so the person can spend more time on the work that needs them?

Think of 80/20 as a design lens. Look for the substantial portion of gathering, organizing, comparing, and preparing that can support the human contribution: creativity, relationships, judgment, and decisions. The split will vary by workflow.

The test is whether the new process makes the person more effective.

The process has to produce more value for the end user than it demands from them.

If employees have to maintain another tool, repeatedly explain the same context, and correct most of the output, the process needs more work.

They should experience a practical benefit: less data management, fewer manual tasks, less time spent preparing, and better information when a decision needs to be made.

The business needs a benefit too. Faster estimating might create capacity to respond to more opportunities. Better prospect research might help salespeople spend more time with suitable buyers. Less rework might let a team deliver more with the resources it already has.

Define that outcome before you start. Measure the full effort, including review and corrections.

Keep the people involved as you test.

User meetings help uncover where the redesigned process still creates friction. Show-and-tell sessions let employees demonstrate what worked on actual work and explain where it fell short. Small cohorts give people time to build, test, and improve workflows with colleagues.

Each session should move the work forward. Bring an example, examine the result, and decide what to change.

Someone also needs to coordinate that effort: keep the business outcome in view, help resolve blockers, and make sure useful experiments become part of everyday work. That can be an internal leader or a partner like Flyte.