Plan the handoff before the automation.
The most useful workflow is not always the one that does the most. It is the one that reliably moves a defined task forward and makes the exceptions understandable.

1. Name the trigger and the destination.
Write one sentence describing what starts the workflow and what a completed step means. For example: when a website inquiry arrives, send the agreed information to the responsible team. Avoid vague goals such as “automate sales” until the individual tasks are clear.
2. Decide what information is necessary.
List the minimum information required to route or respond to a request. Decide what happens when a field is missing or an answer is unclear. Avoid collecting information just because a tool allows it; access and data-handling requirements need to be agreed for the specific use case.
3. Define when a person takes over.
For voice or chat support, outline which questions can be answered from approved information and which need a person. The handoff should include the relevant context so the customer is not forced to start again.
- Requests outside the assistant’s approved scope
- Uncertain or conflicting information
- A customer asking to speak to a person
- A failed action or unavailable connected service
4. Test the awkward cases.
A normal request is only one test. Try missing information, duplicate submissions, a slow provider and a failed connection. If a response is lost, the action may have happened; the workflow should report uncertainty rather than inviting accidental duplicates.
5. Assign an owner.
Someone needs to know when the workflow fails and who can change it. Agree on review responsibilities, access and how approved information stays current. A working demonstration is the beginning of that operational plan, not a substitute for it.
Map a workflow worth improving.
Tell us which task repeats and where the handoff currently breaks down.
Before you redesign: find the friction.
A redesign is easier to scope when you can describe what is getting in the way. Start with the visitor’s task, then decide which parts of the site need to change.

1. Say what the visitor came to do.
Choose a few realistic tasks: understand a service, check whether it fits, compare an approach or make an inquiry. Walk through each task using the current site. Record the points where you have to guess, search elsewhere or repeat a step.
2. Check the message before the layout.
Read the opening of each important page without the images. Can a visitor tell what you offer and who it is for? Replace broad statements with useful service information. A new layout cannot resolve an unclear offer on its own.
- What does this page help a visitor understand?
- Which question should it answer first?
- What information is missing before someone can take the next step?
3. Follow the inquiry path.
Try the contact journey on a small screen and with a keyboard. Check labels, required fields and what happens when something is missing. Review where a submitted inquiry actually goes and who is responsible for it. A visual success message is not proof that the request reached its destination.
4. Separate needs from preferences.
Make two lists: problems that prevent a useful action, and changes to visual style or presentation. Both can matter, but the distinction helps set priorities and avoids expanding a project before the core work is understood.
Bring a short brief.
A useful first brief includes your main services, important audiences, current site, the problems you observed and any timing constraints. Note the tools and content you need to keep. You do not need to choose the final platform before discussing the work.
Turn the observations into a plan.
Share the pages and customer tasks you would like to improve.