Pull every email address out of a mess of text

This job usually arrives sideways. You need to tell forty people something, and the forty addresses are not in a system, they are scattered across a forwarded thread, a spreadsheet somebody exported, and the notes from an event. Nobody planned for it, which is why it always happens under time pressure.

The mechanical part is easy and this applet does it: find the addresses, drop the duplicates, drop the ones that are not addresses at all. What deserves more thought is what happens next, because two mistakes are common and both are avoidable.

The first is putting the list in the To field. Forty addresses in To is forty people who now have each other's contact details, which is a personal data disclosure you did not intend and cannot take back. It is one of the most common causes of reportable breaches in the UK precisely because it takes one click and no malice. Use BCC, and the applet gives you the list in a form that pastes straight into it.

The second is assuming that valid means deliverable. An address can be perfectly well formed and belong to nobody: a typo that happens to parse, someone who left two years ago, a role account that was retired. Validation here is structural, and it is honest about that. If deliverability matters, the addresses need verifying properly before a send, not just parsing.

There is a third thing worth pausing on, which is whether you should be emailing these people at all. An address you scraped out of a thread came with a context, and that context may not have included marketing. Under UK GDPR and PECR the question is not whether you hold the address, it is what basis you have for the message you are about to send. An applet cannot answer that. It can only make sure that once you have answered it, the list is clean.

Where this stops being a browser job is when it recurs. If you are rebuilding the same list every quarter out of the same three places, the list is not the problem. The problem is that the addresses live in three places, and that is a data management question rather than an afternoon with a text box.