When firms are looking at automation (or AI) to find other ways of changing and improving what they do), the approach should always be to look at what you do now and how you do it. You can then automate an already-improved version of your processes and working methods.
Those who don’t do this end up automating existing inefficiencies- doing the wrong things faster. I won’t bother to explain that this is at best sub-optimal.
So, carry out a review of your current product (or service) portfolio, of the way that production/delivery is structured and of the various production/delivery processes. In particular note which processes give rises to most errors and rework, most customer complaints and most product returns.
Talk through issues with the management team, with supervisors and with operational staff. Find out their complaints, frustrations, pet hates – and preferences.
Then, work through alternative ways of satisfying the needs of your customers (including currently unmet needs) and start to build a vision of your new factory, warehouse, office or whatever with all (or most of) the negative points removed and all of the positive points retained. Then think of the role(s) that automation might play.
Finally, within your constraints, work towards creating the nearest practical implementation of this vision that you can. You might even find that the improved structures, processes snd working methods are enough to give you a real productivity boost without automation. The potential investment might be better used elsewhere.