An agency white labels an assistant, presents it as its own and charges a monthly fee with a healthy margin. For six months this is excellent.
Then a client asks a technical question the agency cannot answer, a second client wants to take the service in house, and a third asks who processes their data for a compliance form.
None of these are unreasonable questions. All of them were answerable at the start, and none of them were asked before the arrangement was agreed.
1. Establish exactly what is white labelled
Interface, branding, domain, emails and documentation. Partial white labelling is normal and the parts that remain branded are where a client eventually notices.
2. Accept that clients find out
Through a support conversation, a document or a search. This is fine when the agency positioned itself as the builder and the manager rather than the inventor, and awkward when it did not.
3. Decide the honest positioning early
We build and manage this for you is a defensible and durable position. We wrote this software is a claim that will be tested and will cost credibility when it fails.
4. Price the support you are taking on
Every client question routes to the agency. This is the largest practical consequence of white labelling and the one most often absent from the margin calculation, as set out in pricing assistant builds for clients.
5. Establish what happens when the client leaves
Who holds the account, what transfers, what the client keeps and what it costs. Agreeing this at the start makes departures professional instead of acrimonious.
6. Establish the data processing position
Who is the controller, who is the processor and which terms the client's compliance team will actually need. This question arrives eventually and a good answer is a competitive advantage.
7. Establish the security answers you can give
Hosting, encryption, retention and access. Clients ask, and an agency that cannot answer looks like a reseller rather than a partner, which is the standard in ai chatbot security questions to ask a vendor.
8. Check what happens on an outage
Whether you are told, how quickly, and what you can say to clients. An agency learning about an outage from its own client is in a very poor position.
9. Check the escalation path to the vendor
Response times, named contacts and what is covered. Your support promise to a client cannot be better than the one you receive.
10. Check the pricing model as you scale
Per client, per seat or per volume, and what happens at ten clients rather than two. Margins that work at small scale sometimes do not survive growth.
11. Check the constraints on your own pricing
Any restrictions on what you may charge, bundle or present. These are worth knowing before you build a price list around them.
12. Check what happens to configurations
Whether a client's setup can be exported, transferred or must be rebuilt. This determines how painful any future change is for everybody.
13. Decide whether white labelling is worth it at all
A named tool that you configure and manage is an easier sell in many markets, and it removes most of the questions on this list.
14. Get the scope and the client expectations right regardless
The commercial arrangement does not change the need for a defined project, which is the ground in scoping a client assistant project properly, and the same first contact questions clients receive appear in clinic websites and the admin questions patients ask first. See pricing, run the numbers on the chatbot ROI calculator, and the sector view sits under agencies.