The four things prep is made of
Pull the numbers. Arrange them into something comparable. Decide what they mean. Write the document the client keeps. Only the third is the work a client pays an MSP for.
At two to four hours per client per quarter, an MSP running thirty clients spends more than a working week rebuilding documents that differ only in their numbers.
Why the rebuild is so slow
It is not the typing. It is the switching: export a report, reconcile a date range, notice a metric moved, go back for the detail, restate it in client language, then check the previous quarter's deck to keep the framing consistent.
Every one of those steps is repeatable, which is precisely why it is worth removing before touching anything that needs judgment.
What not to automate
The decision. A generated document that recommends an action without a person standing behind it is worse than no document, because it will eventually be confidently wrong in front of a client.
The right split is narrow: a tool arranges evidence and drafts the questions; the MSP decides, corrects, and approves before anything reaches a client. Anything that blurs that line is a liability wearing the costume of a time saving.
Standardise the input before buying anything
The fastest saving costs nothing: agree on one export, with one set of field definitions, used for every client. Most of the reconciliation time is spent deciding what a number means this quarter because it was pulled differently last quarter.
Write the definitions down once — what counts as an opened ticket, which alerts are critical, whether a licence is unused after thirty or ninety days — and the export stops being a judgment call. That document is worth more than any tool, and it is a prerequisite for one.
Batch the quarter instead of scattering it
Preparing thirty QBRs one at a time, as each meeting approaches, pays the context-switching cost thirty times. Pulling every export in one sitting and drafting every document in the next pays it twice.
It also surfaces patterns across the book that are invisible client by client: the same recurring incident at four sites, or an unused-licence position that is worth a single conversation with a vendor rather than four apologetic ones with clients.
A practical way to test it
Take one client and one quarter. Prepare the QBR the way you do now and note the elapsed time honestly, including the interruptions. Then prepare the same one from two rows of a fixed template and review the generated page line by line.
The comparison you want is not generation speed. It is time to a document you are willing to defend, which includes every correction you had to make. If that number does not move, the tool is not worth paying for, and you will know within one client.