Roles are bundles of tasks
Look at any role closely and it splits into tasks. Some are repetitive and rule-based: re-keying details, chasing missing information, formatting the same report each week. Others need judgement, relationships or accountability: pricing a tricky job, handling an unhappy customer, deciding what to do when the rules don't fit.
Current AI is useful on the first kind and unreliable on the second. A workflow that respects that line frees time without removing the person who knows the business.
Introduce it with the people who do the work
- Ask them where the time goes. The people doing a process know which steps are tedious and which ones matter.
- Automate the tedious steps first. Keep a person approving anything that leaves the business or can't be undone.
- Decide in advance what the freed time is for. Following up quotes, better service, training: say it out loud.
- Train them to run and change it. A workflow your team can't adjust becomes a dependency. One they can adjust becomes a skill.
Warning signs of bad automation
- Nobody on the team can explain what it does or why.
- It makes decisions about customers or staff with no one reviewing them.
- Mistakes are discovered by customers rather than by a check.
- Changing a simple rule needs an outside provider every time.
Each of these is a design choice, and each is avoidable. Try the delegation boundary builder to see how the choices play out.
What to tell your team
Be specific: which process, which steps, what stays with them, and what the freed time is for. Uncertainty does more damage than the technology. People who helped design a workflow tend to trust it, and they're the ones who'll notice when it needs changing.