baz

On a formal opinion you are buying insurance, not research.

Which is why 464,000 of this spend should keep going out. The addressable part is real, it is smaller than the invoice total, and the difference between those two figures is the difference between a business case and taking on risk your insurer has not seen.

1,115,800
external spend, an in-house team of six
371,850
addressable — 33%
464,000
should stay external
73
instructions a year

What each category is actually buying

Sorted by spend. The two largest are not the two most addressable, which is the point of doing this by category at all.

Routine lookups are the smallest average bill and the largest volume, which is the shape that makes them worth automating. The mixed categories count at 45% — the research half of the instruction, not the judgement — and the two external rows count at nothing.

What actually changes

The first is the largest lever and it moves no risk at all, which makes it the one to start with.

The instruction gets shorter
Most of an external bill on a contested point is somebody reconstructing the background you already had. Sending the groundwork with the question is the single largest lever and it changes no risk allocation.
You stop asking the easy ones twice
The same routine question goes out from three different people across a year. A searchable internal answer with its source ends that, and it is where the volume is.
Panel reviews get an evidence base
Spend by category and by firm, against turnaround, is a negotiating position at renewal. Most teams arrive at that conversation with an invoice total and nothing else.
The residual gets more expensive, correctly
What remains is the hard, high-value work. Average cost per instruction will rise after this and that is the intended outcome, not a failure of the programme.

Four reasons the number is smaller than it looks

The second is the one this whole page is built around.

Two categories are not savings
464,000 of the spend here should keep going out. Counting it would make the case look better and would be advice to take on risk your insurer has not priced.
In-house answers carry in-house risk
An opinion from outside comes with somebody else's indemnity. The same conclusion reached internally does not, and that difference is the product being bought.
It needs somebody to own the answers
An internal research function without a named owner becomes a folder of stale memos. The saving is real and it is conditional on a person maintaining it.
Relationships have option value
Firms you rarely instruct answer the phone slowly in a crisis. Cutting volume to a panel firm is a decision with a cost that does not appear on this page.

371,850 against a 1,115,800 bill is a good year's work. Quoting the 1,115,800 would be easier and it would be advice to stop buying insurance because you have not yet needed it.