Most conversations about technical debt fail before they start, because they are framed as engineering asking for permission to stop delivering for a while. I think the framing is the problem. Technical debt is a decision that was made, usually for a good reason, and the conversation with product is about whether the terms of that decision still hold.
The word itself does not help
Debt is a metaphor that sounds precise and is not. It suggests a number that could be paid down, and product managers reasonably ask what the number is. There is no number. What there is, in my experience, is a list of specific things that are slower or riskier than they should be, each with a cause. The conversation goes better when the list is on the table instead of the metaphor.
What I try to bring
Three things. The specific friction, in product terms, such as a change to the checkout flow that should take a day and takes a week because of how the form state is handled. The thing that would remove it, sized honestly. And what gets easier afterwards, named as features product already wants. Debt work that cannot be connected to something product cares about is a harder sell for a reason, and sometimes the reason is that it should not be done yet.
I also try to say what will happen if nothing is done, without drama. Not a cliff. Usually a slope, where each quarter the same work costs a little more, until someone proposes a rewrite because the slope has become invisible.
Making it a standing decision
The teams I have seen handle this well treat it as a proportion rather than a negotiation, something like a fixed share of each cycle that engineering spends on the list, with product seeing what was done and what moved. The share is argued about once, not every sprint. What is on the list is visible to both sides, and product can push a specific item up when it blocks something they want.
I think the proportion matters less than the fact that it is settled. A team that renegotiates it every two weeks spends the time on the negotiation.
The exception
Sometimes the right answer is to let it accumulate. A product that is being retired, a prototype still finding its users, or a feature nobody is sure will survive does not deserve a clean foundation yet. The honest version of that is to say so, out loud, and write down when the decision gets revisited, so it is a choice rather than a drift.
What I am still working out is how to explain the slope to someone who has only ever seen the cliff. The rewrite proposal is what technical debt looks like from outside, and by then the cheap options have gone.
Photo source: https://photos.robertstowe.com/providence-canyon
