The schedule that lives in a spreadsheet

Most programmes are not planned in a planning tool. They are planned in a workbook where somebody has merged cells across columns to draw a bar, colour-coded the phases, and put the dates in the header row. Foreweave reads that, not just the tidy case. If your file is tidy — one row per task, real date columns — it reads that faster.

What it can pick up

• Typed start and finish dates, durations, percent complete, and outline levels.
• Bars drawn as merged cells, filled cells, conditional formatting, theme colours, or drawing shapes laid over the grid.
• MS Project predecessor notation — 12FS+3d, 8SS, and the rest — including dependency type and signed lead or lag.
• The date axis, wherever it lives, and how the bars align to it.

You confirm before anything is committed

Extraction is never applied straight to a plan. It comes back as a review, and every item carries its provenance — the sheet, the cell, and how it was read. Three checkpoints:

• The date axis first. Foreweave states how it interpreted the header — which column is which day — and you confirm that before a single date enters the plan. Getting this wrong silently is the one failure that would poison everything downstream.
• Low-confidence items are opt-in individually. Anything read from a shape or a colour rather than a value has to be accepted item by item. It will not be smuggled in with the confident rows.
• Nothing reaches a baseline without your explicit acceptance of the import as a whole.

Re-uploading the same file gives the same answer — the extraction is deterministic, so a second opinion is genuinely a second opinion and not a dice roll.

Limits, stated up front

Files up to 25 MB and 50 sheets, up to 50,000 imported rows. Anything above about a megabyte processes in the background with progress rather than holding the page. Archives are bounded on decompressed size and compression ratio, and extraction is bounded on cells scanned and wall-clock time — a malformed or hostile workbook is refused with a reason, not left to spin.

Working with the agent

An uploaded file also becomes context the agent can use — it can read the extraction to answer questions about the plan, and building from a file is one of the routes that does not require an accepted planning guideline first, because the structure came from you rather than from invention.