I'm finding myself in agreement having spent many hours playing with a virtual dice that seems to have a particular liking for rolling 3s. I don't know how it all fits in with the D&D rules as I'm not a player but it feels like the dice roll being by far the major deciding factor in any given succeed/fail situation with character skills being only a very minor influence seems completely the wrong way round.
This is to be expected in lower levels. Currently the max bonus you can get to a roll is +6 (+2 proficiency, +4 stat modifier). Assuming a DC of 10 (easy) you will get an average success of 55% (no bonuses) up to an average of 85% (+6 bonus, would show up as a roll against 4). Those modifiers can get pushed quite far. In RAW dnd you could get a rogue on level 4 to +8 (+2 proficiency x 2, +4 stat modifier) on a skill via expertise (not in bg3 as far as I know). This would get the success change up to 95%. On lvl 5 that rogue would have a modifier of +10 (+3 proficiency x 2, +4 stat modifier) and basically could never fail an easy (DC 10) ability check.
The modifiers still play a decisive roll, even on lvl 4. The table shows the average success chances for a level 4 character with attribute 8 (-1), 10 (0), 16 (4) and 16 + proficiency (6):
Code
Modfiers -1 0 4 6
DC 5 75.00% 80.00% 100.00% 100.00%
DC 10 50.00% 55.00% 75.00% 85.00%
DC 15 25.00% 30.00% 50.00% 60.00%
DC 20 0.00% 5.00% 25.00% 35.00%
DC 25 0.00% 0.00% 0.00% 10.00%
So the bonuses play a role, which only increases with proper character growth.