A CFO asks why the architecture function should be funded when it ships no features. You have five minutes. What do you say?

Do not answer with activities

"We produce standards, run a review board and maintain the technology radar" describes what the function does, not what the business gets. It confirms the premise of the question.

Answer with the cost of the decisions that were not made well

Architecture's value is concentrated in a small number of decisions that are expensive to reverse. The argument is made concretely, with the organisation's own history:

• The system chosen three years ago that now takes a team of four to keep running.
• The integration built point to point that has to be rebuilt for every new partner.
• The database everything shares, which is why no team can deploy independently.
• The vendor whose renewal doubled because there was no exit path.

Each was a decision made in a week and paid for over years. That is what the function exists to change, and a CFO recognises the pattern immediately because it is a capital allocation argument.

Then give numbers that already exist

Pick the two or three you can evidence:

• Duplication removed. Applications retired, licences cancelled, with figures.
• Delivery speed. Lead time before and after a specific structural change.
• Avoided cost, stated carefully — a migration not needed because a change was made compatible; a vendor negotiation improved by having an assessed alternative.
• Incident reduction attributable to a design change.

Avoided cost is the weakest form of evidence and it is worth including with an explicit caveat, because overstating it is what makes the whole case look promotional.

Accept the underlying point

The question is often fair. An architecture function that reviews everything, decides slowly and produces documents nobody reads is a cost. If that is closer to the truth than not, the honest answer is a proposal: what the function will stop doing, what it will automate, and what outcome it will be measured on next year.

That answer earns more credibility than a defence, and it is usually the change the function needed anyway.