NLP/ML functionality to auto-fill docx

Tech newbie here. I am stuck with this problem. Lets say there are n = 10 versions of output_n.doc created manually from a baseline.doc and based on 10 different sets of input text entries from a Typeform Q/A. Each set of input entries is based on a mix of 20 Q/A. Some of these entries overlap, others are different. Now when a 11th input is filled in via Typeform, based on MLP/NL, how accurately can the corresponding 11th version output_11.doc be created, based on auto-filled text entries. I realize that, certain fields of output_11.doc may be left empty, but how much can it auto-fill? How can I validate above test scenario with typeform, airtable and some quick hacks? Any other tool or plugin or module that can be leveraged to help implement above? Please share any reference links that I can read further. Is there any major technical flaw in my description and assumption above?

Thank you hackers !!