If I need to be more specific about the meaning of 'intelligence' then I'm happy to restrict this to the first use of one or other elevator algorithms
All the algorithms mentioned in that post have been implemented by means of relays or even high power contactors way before any processor and even less operating system was used.
Most automatic elevators have an algorithm that optimises travel between floors and users. Presumably this can be tailored to a particular building.
Automatic elevators became already a thing in the 1930s, naturally including logic for decisions what to do when there is no immediate request. This included not only waiting level, but also changing that according to time of day - e.g. not moving to a waiting position during the night, wait at the lobby in morning hours and some higher level during afternoon. All of that was implemented by electro mechanic means (relays, controller cylinders and alike).
I am interested in knowing when the first programmable 'intelligent' elevator was built and what processor and operating system it might have had.
As soon as there are preferences for what to do when there is no immediate request, there must be ways of setup, or as you call it 'programmable' and 'intelligent'. Programming the waiting region for each mode happened by plugging wires or setting pins on controller cylinders. Similar time-dependent settings were put as contact pins on clock driven controller cylinders and so on.
I got interested in this when our department moved into a 12-storey building in 1980. The building was erected in 1972 and had only three elevators. We occupied the upper 5 floors and waiting was sometimes less than satisfying. So our CEO, a engineer by trade, became, let's say, 'interested' in their workings. This included real-time monitoring on his desk and us optimizing the programming ... a quite nifty mixture of TTL, relays and contactors. Capable of doing example #1, #2 and #3 on the list - no need for #4, as Munich isn't New York :) But sorry, no processor or OS.
A Little Rant
(Not implying anything about the OP)
The question reminds of the problems we often have looking into the past, as we automatic apply today's point of view. Today we use microprocessors to control even such a simple issue as the charge of a battery, and more often than not they are way more complex than a classic 8-bit microcontroller and run some OS - heck, some even an off-the-shelf Linux. All for a task that could as well be done with a Zener diode and a resistor.
We automatically assume that everything more complex than an on/off switch will need some processor to perform, and doing so the application will of course need an OS to run on.
But development was different. Tasks like controlling a lift are rather simple sequences of logic that can be easily done using switches and contactors - adding relays only pays when it gets more complex and there's a saving to be made by adding voltage conversion. Replacing this by some TTL logic is neither a mandatory nor even a logical step. There needs to be a saving that makes a redesign necessary - after all, engineering hours to develop new circuitry are way more expensive than just continuing to build the existing one.
Same when switching to a microprocessor. And even if there's a switch, why go for a high end version capable of running an OS, when all the task needs is to check a few signals and perform simple decisions? Any minimalist (8080 or 8042 or alike) system will do it and programming is a few hundred bytes on bare metal.
The fact that today, 40+ years after the first micros, even the most basic systems offer more than some top end system of the 80s should is no reason to project today's methods back into the past.