Would love to add your ideas. Lots to be covered. For now, given the paucity if time, I've frankly just focused on the core kernel of the idea. There has to be simplicity in a truly disruptive concept. And then of course execution and detailing it is of paramount importance.

Thanks kate I know I need to write up the story - what's the deadline? I'm short in time until the weekend

Thanks Kate Rushton . The Kahneman study is a classic, and actually you made me think of the following. If 75K USD is the threshold beyond which extra money doesn't affect happiness levels in general, and below which life's hardship becomes more acute then...
The algorithm could initially be built with the following heuristics. for those above the threshold, improved life experiences will likely depend on directing the money towards what dampens the effect of more money (e.g. Stress, negative expectations, morbidity, and especially loneliness); for those below, the optimal nudge strategy would be to try and reduce the "wasted" Money (income that is spent on things that don't bring happiness), so that they can inch upwards towards the discretionary income of the 75K+ cohort.
But also - 75K per year is the general median - where are the folks aged 50+? what is that threshold in retired vs non-retired people (who likely still have commitments to children etc) - possibly, the freer from commitments you are, the less money you really need to feel happy (also remember studies that show happiness goes up later in life, despite actual income shrinks). Also, to borrow from the spirit of class Modigliani-Miller long term horizon choices, some people will not exhibit happiness levels of their CURRENT income level, but they will experience some of the "future" (I.e. If they know that they will become, or may become, significantly poorer), or perhaps some of their children's etc. In other words, the segmentation of the audience will need to be finer for the recommendations to be accurate. That's why AI and mass personalization can add value - because cookie-cut if-then rules won't work well.