AI IN A
WORKING BUSINESS.

Start with one customer journey.

Follow an enquiry from the moment it arrives to the point where someone takes responsibility for it. Where does the customer wait? What information gets lost? Which questions does the team answer repeatedly?

A narrow starting point makes the work easier to assess. For example, you might explore answering routine questions before passing a qualified enquiry to a person. That is a clearer brief than asking a new tool to improve the entire business.

Write down the current process first. Include the people, systems and exceptions involved. AI still needs an accurate understanding of what the business offers and what the customer can reasonably expect.

A conversation needs somewhere to go.

A response is useful when it helps a customer move forward. That might mean finding the right service, arranging an appointment or reaching a team member who can resolve a more complicated question.

Decide which information the system can rely on, what it may do and when it should hand over. If a person needs to take over, make sure the conversation and relevant details reach them. Customers should not have to start again simply because the enquiry has changed hands.

For scheduling, define the actual rules: availability, appointment types and what happens when a booking cannot be made. Test those exceptions before treating the workflow as ready.

Give the workflow an owner.

Someone in the business needs to maintain the information, review mistakes and decide when the process should change. The work does not end when a tool is switched on.

Use only the access and customer information needed for the job. Make escalation straightforward. Review unusual conversations as well as successful ones, because exceptions often reveal what the process is missing.

Look beyond the number of messages.

A busy conversation is not automatically a successful one. Choose measures that reflect the job: whether an enquiry was resolved, whether a valid booking was made, whether the handover was complete and how much rework the team needed to do.

Compare those results with the previous process over a meaningful period. Response time can be useful, but it belongs beside accuracy and the customer’s outcome. Set a review point before deciding to expand the workflow.

Improve one part, then build from it.

Once a workflow is useful and reliable, its lessons can inform the next one. Keep what works, correct the gaps and avoid assuming that a process suited to one part of the business will transfer unchanged to another.

Rami’s businesses provide the context for this practical focus. BlackBull made AI central to its direction in 2022 and launched AIVUS for conversational engagement in 2023. CleanMade combines AI-supported customer service and administration with service delivery by human franchisees.

The question to keep returning to is simple: what does this change make better for the customer and the people doing the work?