Quite so! And quite right if I may say so!
When I was MD of a computer bureau in 1970-72 we had the problem you set your people. Most of our customers refused to sign things they didn’t understand. They would say “You build the system and if it works and we like it THEN we’ll pay you for it”. Needless to say we couldn’t afford to take on work on these terms because there was no customer commitment. We absolutely had to get something from them in the nature of a test which, if the system passed it, would be grounds for requiring them to pay. This required our systems people to be able to talk to the person in the customer organisation who could authorise the cheque and they, for their part, had to understand what they were buying well enough to know it was worth the money (sometimes the judgement what it was worth was made at a higher level).
And, of course, the impasse was resolved by building something temporary which could be amended till it did do what the customer needed (or at least what he could work with). But since, in those days we were working with COBOL and on machines where a 16K partition was allocated to run each production module, in practice the only work we could get was similar to things we were already doing for other people where a set of modules could be reassembled with minor changes to suit. Payrolls, purchase ledgers, sales ledgers, membership record systems and so on were the bread and butter; almost everything else lost money! Watford Computer Centre, according to 3is (which funded a management buyout 1974ish), said it was the only profitable computer bureau they knew of.
But it is interesting that this fact about the needs of the business made it almost impossible to promote programmers to be systems analysts; they just could not hold their own with the financial directors or administration managers of our customers. The systems analyst job was the most exposed of all; if he allowed a system to be produced which the customer wouldn’t pay for that was curtains for him.
Some years later, when I worked for the BBC, I encountered the situation you describe in another way. I was in charge of a project to produce a system which was supposed to manage the program planning for radios 1, 2, 3 and 4. My boss, realising that the controllers of the networks could not possibly sign off a systems requirement or specification (as they could not envisage the workings of the proposed computer based system), had found a way to fudge it (I didn’t know this when I joined). About a year after I joined I finally established that no controller, no chief assistant and no presentation editor had agreed to use the system proposed and, at a highly traumatic meeting with all these people and the Directorate of Radio the system (something of the order of thirty man-years of work) was abandoned.
Later a totally different approach kitted out each network with its own word processor (does anyone now remember the Wordplex machine?) which allowed them the essential facility to make changes ten times a week during the three month planning cycle (and easily send out the revisions) until the FINAL plan was nailed to the floor and sent to Radio Times. I was able to see this replacement into service, where I’m glad to say it lasted many years.
It cost me my career with the BBC as my boss, who was the real villain of the piece, suffered a loss of face and demotion from which he never recovered – but few of the managers in the BBC were sufficiently confident to take me on when I could no longer report to that boss (at the traumatic meeting I asked all present to ensure that I didn't have to report to him any more).
Wittgenstein would say (and I would agree) that utterances are behavior, and that that’s the beginning and the end of it so far as language is concerned. So the only question is whether it is adaptive for you to utter the words ‘these specifications are correct and complete’ on any given occasion. He would say that the question of whether the specification ‘is’ in some cosmological sense ‘correct and complete’ is a nonsense question.