A company with nineteen branches debates whether enquiries should go to a central team or to the branches.
Central handling was tried and abandoned, because the team could not answer questions about local stock, local staff or the services a particular branch actually offers.
Branch handling is what happens now, and it means enquiries wait until somebody is free, get answered differently everywhere, and disappear entirely at busy times.
What central handling does well
Speed, consistency, measurement and cover outside branch hours. Everything that depends on process rather than on local knowledge.
What central handling does badly
Anything local. Stock, staff, services, access and the specific practical detail that turns a prompt answer into a useful one.
What local handling does well
Knowledge, relationship and the ability to resolve something in one conversation because the person knows the situation.
What local handling does badly
Availability, consistency and reporting. Nineteen branches answering from nineteen sets of accumulated habit produce nineteen slightly different versions of the same policy, and nobody can measure any of it.
The split that actually works
General questions answered centrally and instantly, local questions answered from local data, and only genuinely local conversations reaching a person at the branch.
Which questions are general
Prices, policies, processes, product information and anything true across the network, which is usually the large majority of volume.
Which questions are local
Stock, availability, specific staff, access, parking and anything about that site, which needs branch data rather than a branch person.
Which questions need a person at the branch
Complicated situations, existing customers with history, and anything requiring judgement about a specific case.
What the branches should stop doing
Answering the general questions, which is most of their enquiry load and none of their value.
What head office should stop doing
Attempting to answer local questions from a central script, which is what made centralising fail the first time.
What the routing depends on
Knowing where the customer is and what they need, which is the playbook in routing enquiries to the right location automatically.
What both models currently lose
The enquiries arriving outside branch hours, which neither approach handles and which frequently exceed the daytime volume.
What the reporting should show
Volume, response time and conversion by branch, which most networks cannot currently produce under either model.
What the branch staff should still own
The customers in front of them and the relationships built over years. Taking the general enquiry load away is what gives them time for that, rather than replacing anything they are actually good at.
What the model should look like from outside
The customer should not be able to tell. A question answered centrally and one answered at the branch should read as the same company, which is only possible when both draw on the same written answers.
What to measure once it is split
Which questions reached a person and why. That list is the gap in the documented answers, and reviewing it monthly is what keeps the split working rather than gradually collapsing back to the branches.
Answer generally where you can and locally where you must
An assistant that answers the network wide questions instantly, uses each branch's own data for local ones and passes only genuinely local conversations to the branch gives the speed of central handling with the knowledge of local handling. The head office problem is described in local enquiries arriving at head office, the launch case in launching a new location and the enquiries that follow, and a worked example in ai assistant venue enquiries. See it working on the demo, compare on pricing, and the sector view sits under b2b.