Sometimes, you're in a meeting, and the technical people will say that they have to do huge amounts of integration to get X old system to work with Y new one. The integration, because it involves your old stuff, will be expensive and time consuming. You sigh and wonder how you are ever going to get out of the legacy trap.
You know that, even if you mandate a new systems approach, you're always going to have to spend piles of money doing legacy integration, because that's where the data is. It is data on which the business relies, so it has to be available, and, obviously, the job of moving it all, and the business rules and all the rest, to a new system in one go is far to large and politcally charged.
And then, you wake up one morning and realise on top of everything, you have the "die or retire problem". Everyone who knows the guts of legacy is either dead or about to retire.
You have to do something about legacy, so you sigh, and sign up for truck loads of integration anyway.
it occurs to me, though, that there are solutions to this problem which don't involve piles of expensive technology work. What about, for example, using people as glue? Why can't they look up data in both systems and perform appropriate updates?
The thing is, technologists have spent decades trying to systematise and automate everything. The idea that you'd de-automate is an anathema. "What about error!" they cry whenever you suggest it. "What about security!, How about our data quality, how about... how about... and how about...!".
There are obviously lots of objections to putting people back into a system but they all miss the major point: if you are time constrained, and you need to get something out the door right now you can get almost any system up and running instantaneously with paper and people.
That is certainly true when you have the die-or-retire thing to deal with.
As far as I can see, the only time this absolutely won't work is in those circumstances when a process is entirely automated from end to end with no people in it at all. Like, for example, an ATM transaction or a web purchase online. But whenever you have people intermediating the machine for other people, you can use people as glue.
Now, I'm not suggesting that anyone put people back into systems permanently. But as temporary, good enough, solution while something better is worked out, surely it makes sense to at least consider the option.