I agree with the OP. Unless Ainge can pull off a marvellous off season in which he signs two all stars to the team there is no point in putting or losing assets (draft picks, cap space) into the current core of Thomas-Bradley-Crowder-Horford.
That's just not a championship group, not even close. Don't let the achievement of getting to the Conference Finals (very nice though) throw dust in our eyes. The Celtics are no better than teams like the Raptors, Wizards, Hawks, Rockets or Jazz.
Hayward is not a big difference maker and signing him (and or) Thomas for a max contract would keep the Celtics into that second tier for a long time. Of course you could be happy with being guaranteed to reach the playoffs every year and sporadically overachieve like this year, but to me the goal should be to become the best team in the NBA. Especially considering the luxury position the Celtics are in long term.
That means patience and develop our younger players like Fultz (hopefully), Rozier, Smart, Brown, Olynyk (?), Zizic and other incoming draftees. I'm perfectly all right with not making abrupt changes and steadily grow as a team while still being competitive, but it has to be clear that the most important players for the Celtics are Fultz and Brown.
Do nothing that hampers the opportunity for them to fully blossom. So trade Thomas or Crowder whenever that's needed. That doesn't have to happen all at once. But don't sign players for max contracts that aren't superstars. No Hayward and let Thomas walk if he demands a 20+ million contract.
If we only give out a couple of average contracts (for maybe Rozier, Smart, Bradley, Olynyk) the next few years than in 2020 when Horford's contract ends we'll have the chance to sign a top free agent and then we can go over the cap again by resigning Brown (and Fultz, Zizic in 2021).
The Celtics peak should therefore be focussed on for approximately 2021 and onwards. In the meantime I'd happily surrender some small successes. I hope Ainge feels the same way and acts accordingly because Stevens probably doesn't.