Moreover, there are already several companies that provide this type of services. Xapo is one of them, Circle has just arrived, and several others, so I don't think it is wise to enter this kind of competition at this stage.
I was thinking of a solution without going the easy path of hinting that Bitcoin will disrupt credit card companies and make them obsolete. I would need thinking out of the box for a solution that would benefit both the credit card company and Bitcoin.
Any other ideas? Brainstorming would also be OK...
dont think about the exchanges such as bitstamp and btc-e, etc.. trying to get bank transfers in and out are a headache. what you should do is find a large whale FIAT investor who wishes to continually buy into bitcoin.
the secret of how most services work these days is that exchanges are just for the little fish and the large whales trade privately.
for instance bitpay got $30milllion and that investor will get bitpays bitcoins in return. without bitpay needing to throw coins at an exchange or withdraw fiat from an exchange.
obviously the credit card company has interest in bitcoins, so maybe they want to be a whale collecting bitcoins, thus solving your problem, where all you ave to do is set the price of your swaps with them.
but if they are just looking for a "gimmick" to continue their fiat profits. then you will need to find a whale.
above was the large picture of the long time goal/strategy(private exchange). but for initial trials you could start with a small reserve to pay the card company to load the cards etc. (reserves are important to deduce delays) and then use exchanges to refill the reserves as customers participate.
reserves are easy to calculate. simply by setting reload limits.
EG $100 load limit and releasing 1000 cards = $100k reserve/investment capital requirement to start.
so having reserves and method to get more FIAT for refill is as you say a difficult thing to securely handle (as other parties are involved). but is achievable.
from the concept:
think of the "gimmick"
what will make customers come to your card as oppose to other bitcoin card issuers. again it brings it back to if its a simple cashout method for customers (0% ATM fee) if its ..umm.. branded to be used for pocket money amounts in starbucks, walmart, 7-11. (0% retail fee)
*i mention 0% obviously because bitcoiners believe in monetary freedoms. so no point branding your service as an ATM card if you have 50c fee per use*
i think the "gimmick" (sales pitch/branding) that works is the hardest part. .. trying to find whales(as envrin points out) and cold storing/swaping coins is easier.
basically at stage 1 the card issuers need to know the 'fees' they will get. thus plan the branding and the fee's around the branding, then later concentrate on how to reload' the cards.