Plenty of capable practices still run on AutoCAD, and plenty of Revit migrations have gone badly. The question is not which software is better in the abstract. It is whether the way you work would benefit from a model.
Where a model genuinely pays
Change frequency. If your projects change repeatedly late in documentation, a model earns its keep. Move a wall once and plans, sections, elevations and schedules follow. In CAD you update each one and hope nobody misses a sheet.
Coordination with consultants. If you regularly coordinate with structure and MEP, working in a shared model finds problems before site does.
Schedules and quantities. If you produce door schedules, finish schedules or area calculations by hand, these come from the model as a by product rather than a separate exercise.
Repetition. Multifamily, hospitality and anything with repeating units rewards modelling heavily.
Where CAD remains sensible
Small alteration projects, one off details, quick studies, survey markups and work where you are handed CAD backgrounds and asked to return CAD. Modelling a small internal fit out to produce four sheets is usually slower, not faster.
There is no prize for modelling something that did not need a model.
The costs people underestimate
The licence is the smallest part. The real costs are the learning curve, during which output slows before it improves, the template and family library work that has to happen before production is efficient, and the discipline shift. Revit punishes casual modelling in a way CAD does not: a sloppy model produces wrong schedules, whereas a sloppy CAD drawing is just untidy.
Budget for a genuine dip in productivity on the first project or two. Practices that pretend this will not happen tend to abandon the migration halfway, which is the worst outcome.
A workable transition
- Pick one suitable project rather than converting everything at once
- Invest in the template first, since a poor template makes every project harder
- Keep one experienced person available to unblock others
- Accept CAD output where clients need it, exporting from the model
- Build the family library gradually from real project needs rather than in advance
The middle path
Many practices run both for years, quite reasonably. Model the projects that benefit, draft the ones that do not, and keep the standards for each properly maintained. The mistake is running both badly, with a half maintained template on one side and undocumented layer conventions on the other.
Decide based on your project types and change patterns, not on what the industry says you should be doing.