Dawn has always been good at answering questions about code, tickets, and plain text. But teams do not only work in those formats.
Someone uploads a spreadsheet of support cases and wants them categorised by severity. A team lead attaches a slide deck and asks Dawn to tighten it up. An ops engineer drops a zipped bundle of reports and needs Dawn to find the relevant file, summarise it, and send back a clean version. Until now, all of those requests hit a wall because Dawn could not read or produce document files.
That changes with this release. Dawn can now open the most common document formats, extract structured content from them, reason over that content, and generate new files that the team can download and use directly.
What You Can Read
Dawn now understands the structure of the documents your team actually works with.
PDFs are extracted page by page, preserving the document’s layout so the agent can reference specific sections. Word documents are read as structured blocks — paragraphs, headings, tables — rather than flat text. Excel spreadsheets come through with their sheet structure, column ordering, and cell data intact, so the agent can work with tabular data the way you would in a spreadsheet application. PowerPoint decks are read slide by slide, with titles, body text, tables, and speaker notes all preserved as separate elements.
RTF files and Outlook MSG exports are supported too, which covers the common case of “somebody exported this from email or an older system and wants Dawn to make sense of it.”
The key difference from before is that Dawn treats each format according to its actual structure instead of flattening everything into plain text. A spreadsheet is understood as rows and columns, not as a wall of comma-separated values. A slide deck is understood as titled slides with content sections, not as a long paragraph with line breaks.
What You Can Generate
Reading documents is only half the problem. Teams often need the output as a file they can share, attach to a ticket, or forward to someone who does not use Dawn.
Dawn can now render new PDFs, Word documents, Excel workbooks, and PowerPoint decks. That means the agent can go from reading a source document to producing a polished output without the team needing to copy and paste from a chat message into a file editor.
A few examples of what that looks like in practice. You can upload a spreadsheet of raw data, ask Dawn to clean and categorise it, and get back a formatted workbook. You can point Dawn at a set of connected sources and ask it to produce a summary report as a PDF. You can ask Dawn to assemble a presentation from extracted material across several documents, with proper slide layouts, speaker notes, and images.
PowerPoint rendering in particular goes beyond what most AI tools offer. Dawn produces real slide decks with native text, tables, images, and multiple layouts — not plain-text slides exported into a .pptx wrapper.
Working With Archives
Document-heavy workflows often involve bundles. A team might receive a ZIP of monthly reports and need Dawn to find the one that matters. Or an agent might generate several output files that should be delivered together.
Dawn can now inspect archives, pull out specific files, and create new archives from generated outputs. The supported formats are ZIP, TAR, and gzip-compressed variants. The agent can look at what is inside an archive, pick the entries it needs, work with those files individually, and optionally repackage results into a new archive for download.
That means Dawn can handle the kind of file-based workflow that used to require someone to manually unzip, find the right file, open it in the appropriate application, and then do the actual work.
What This Changes for Teams
Before this release, any request that involved reading or producing a document file was a dead end. The agent could answer questions about code and tickets all day, but the moment someone uploaded a PDF or asked for a formatted report, the workflow broke.
Now those requests work end to end. Upload a file or point Dawn at one in a connected source. Ask a question about its contents, or ask Dawn to transform it into something new. Get the result back as a downloadable file.
That is a meaningful expansion of what Dawn can do, and it removes one of the most common friction points teams hit when trying to use an AI agent for real operational work.