The hardest commercial work usually starts once you have a preferred HCMS and implementation partner.
The SI needs access to test assumptions, confirm the scope and build a credible delivery plan. Every workshop and solution decision also makes that SI harder to replace. The implementation SOW, the price and parts of the licensing position are still open while this is happening.
Skipping detailed scoping is not a sensible answer. It leaves the client negotiating from assumptions made during selection, before the product and delivery approach have been tested together.
I have found it can be reasonable to engage the preferred SI for a defined scoping exercise. That service should have a capped fee, clear outputs, client ownership and use of the resulting artefacts and no automatic entitlement to the implementation.
That work may also change the licensing requirement: module set, environments and user counts. Closing the licensing BAFO before the scope settles means pricing assumptions you may be about to replace.
Use the same period to turn the implementation proposal into a contract-ready SOW. Data, integrations, testing, environments, cutover and client resourcing need to be covered at a level that can be priced, assigned and accepted.
The RACI needs the same discipline. Broad statements such as “client responsible for data migration” or “SI responsible for testing” leave too much unresolved. The parties need to know what each responsibility involves, where the boundary sits and what the client has to provide.
The preferred SI is helping define work it hopes to deliver and price. That's not a criticism of the SI. It is why someone with no commercial interest in winning the implementation should test whether the scope, price, assumptions and responsibilities tell the same story.
By this point the client has usually invested significant time with the preferred SI and may be working to a fixed go-live date. Walking away gets harder if the final SOW is weak or the commercial position shifts.
So set the terms of the scoping exercise before that work begins. Keep the implementation appointment conditional on an acceptable SOW, price, delivery model and contract.
Some SIs will not accept that arrangement. Better to find out while changing SI still only costs you time.
None of this is easy. But it is easier than negotiating change requests against a loose SOW, and cheaper than resourcing responsibilities you didn't realise were yours.