Most “AI for documents” pitches generate text you then have to paste into your real file, reformat, and proofread. That is not the tedious part. The tedious part is the mechanical fill: taking a template you already trust — a services agreement, an NDA, a proposal — and swapping in this client’s name, dates, amounts, and scope, correctly, every time.
CloudAxis handles the mechanical part on the actual .docx file, in your persistent cloud workspace. You give it the template and the values; it fills every placeholder in one pass. And because a wrong replacement in a contract is expensive, every fill runs with a dry-run preview and keeps an undo snapshot. You see the diff before it saves, and you can roll back after.
How template filling works
Upload a .docx that uses placeholders — tokens like {{client_name}}, {{effective_date}}, {{fee}} — into your workspace. Then tell the agent the values, or point it at a table or spreadsheet that holds them.
The agent fills the whole set of placeholders in a single, run-aware pass (it understands how Word splits text into runs, so a placeholder that Word broke across formatting boundaries still gets replaced). Two guardrails are built into the flow:
- Dry run first. Ask for a dry run and the agent shows you which placeholders it found, what each would become, and anything it could not match — without touching the file. You approve, then it writes.
- Undo snapshot. Before it writes the final file it keeps a backup, so a fill you do not like is reversible rather than a rewrite-from-memory.
Beyond fill: targeted edits on a document you already have
The same specialist can do more than fill a template. On an existing .docx it can:
Read and summarize
Pull the plain text out of a .docx, or convert it to Markdown, so the agent can summarize sections, answer questions grounded in the document, or compare two versions and tell you what actually changed between them.
Run-aware find and replace
Change a term everywhere it appears — a renamed product, an updated address, a corrected figure — with the same dry-run-then-write safety. You can test on a single occurrence before letting it replace all.
What about PDFs and spreadsheets?
It is worth being precise here, because this is where most tools quietly overpromise.
- PDFs are read-only. An agent can extract text from a text-based PDF — to summarize it, pull line items into a table, or find every mention of a clause across a folder of contracts. There is no OCR, so a scanned document or a photo of a page cannot be read; convert it to text first. And CloudAxis does not fill PDF forms or generate PDFs — the editable format is Word.
- Spreadsheets: create and analyze, from data. An agent can create a new .xlsx or CSV from structured rows, summarize columns, and join two tables on a key. Editing a workbook means reading it, transforming the rows, and writing it back — not authoring live in-cell formulas.
Where it earns its keep
This is not a novelty — it is the boring, repeated document work that quietly eats an afternoon:
- Agencies & consultancies — turn one proposal template into a per-client proposal by filling scope, pricing, and dates from a brief.
- Contracts & NDAs — fill a standard agreement for a new counterparty, dry-run the placeholders, and keep the pre-fill version as your undo point.
- Recurring reports — a monthly report template filled from that month’s figures, produced on a schedule rather than by hand.
- Onboarding packets — fill the same welcome or engagement document for each new hire or client from a row in a spreadsheet.
It fits the way CloudAxis already works
The template and the filled output live as real files in your persistent cloud workspace, so one agent can research or assemble the data, drop it in a folder, and a document specialist can pick it up and fill the template — on a schedule if you want. You open the desktop and review the result, or ask for the dry run and approve before it writes.
The honest summary: CloudAxis will not invent your contract, and it will not read a scanned PDF. What it does well is the mechanical, error-prone fill on the documents you already trust — with a preview before and an undo after. Document skills are available from the Growth plan.
Frequently asked questions
Can it fill a PDF form?
No. PDFs are read-only in CloudAxis — an agent can extract text from a text-based PDF, but it cannot fill PDF form fields or generate a PDF. The editable, fillable format is Word (.docx). If you need a PDF at the end, produce the .docx and export it to PDF yourself.
What does the dry run actually show me?
A preview of the change before anything is written: which placeholders were found, what each would be replaced with, and anything that could not be matched. You review it, then confirm the write. Template filling and find-replace also keep an undo snapshot so a bad edit is reversible.
Does it work on scanned documents?
No — there is no OCR step. A scanned or image-only PDF is just a picture of text and cannot be read. Use a text-based file, or run OCR outside CloudAxis first. A quick test: if you can select and copy the text yourself, an agent can read it too.
Can it create a Word document from scratch?
The Word tools are built for editing an existing .docx — reading it, converting it to Markdown, find-replace, and filling template placeholders. So start from a template you provide rather than expecting a brand-new formatted .docx from nothing.
Which plan includes this?
Reading and editing Word documents, and reading PDFs, are available from the Growth plan. See the pricing page for full plan details.
AI document processing & automation · Docs: Word Documents · Docs: PDF Documents · AI agents for accounting firms