Punchcard product updates

What's New

• Structured triage interview after automation runs: When an automation run finishes, CoAudit now opens an interactive interview instead of a block of text. The AI presents the top three to five findings as individual cards — each with concrete fix options and a "handle later" escape — and works through a root-cause sequence: are files in the right groups? are matches correct? are failures genuine or a workflow mismatch? Workpapers created from a task template get an additional note when repeated failures suggest the issue lives in the template rather than individual rows. The same interview runs when you ask CoAudit to "review my analysis," and it won't re-read data already visible in the run card above.

• Preview what the AI will write before you confirm: When CoAudit proposes writing a value to a protected workpaper cell, the approval card now shows the pending value — or Pass / Fail for a testing column — plus the AI's one-line reasoning, before you confirm. Batches show up to eight rows at a glance so you can review the recommendations rather than accepting them blind.

• Citations toggle now governs the AI's citation behavior in workbook responses: The Citations switch in the chat header controls how actively the AI cites workbook cell values. On — the AI treats citing every supported value as mandatory and finalizes citations before responding. Off — the AI cites workbook cells only when you explicitly ask. The ability to add citations is always available regardless of the setting.

Improvements

• Firm Library loads without freezing: Firms with very large libraries no longer experience the browser becoming unresponsive when opening the Library page. The list now renders only what is visible on screen, so the page is responsive immediately regardless of file count.

Bug Fixes

• Workpaper editor suggestion pickers no longer crash: The @ mention, / skill, and # assertion suggestion pickers no longer cause the workpaper editor to crash when document content is updated while a picker is open.

• Clear message when a document group name is already taken: Creating a document group with a name already in use on the workpaper now shows a specific "Name already in use" notice instead of a generic failure message.

• Sidebar updates immediately after joining or leaving an engagement: Adding or removing yourself from an engagement now refreshes the left sidebar list right away, without requiring a page reload.

• Workpaper cells save reliably when multiple changes land at the same time: Simultaneous edits to workpaper cells are now saved reliably — an edit no longer fails when another save is happening concurrently.

• Files with unusual characters process correctly: OCR output and AI responses that include uncommon text characters now save correctly instead of failing partway through processing.

• Client Organizations list is more resilient: The client list now handles brief connection interruptions gracefully, showing your clients instead of an error page.

• "Replace Rows" replaces "Replace Selection Data": The sampling split-button dropdown option has been renamed from "Replace Selection Data" to "Replace Rows" for clarity.

Bug Fixes

• CoAudit's activity feed now shows consistent, readable steps when helper agents are at work: CoAudit sometimes delegates parts of a task to specialized helper agents. Their activity now appears in the same clean format as the main agent's steps — showing actual file names, collapsing repeated actions into a single grouped count, and displaying a compact summary once the work is done. Tapping the summary expands a scrollable view of every step. Previously, this activity could show garbled internal codes instead of file names, repeat the same step dozens of times and flood the sidebar, or disappear from the chat entirely after a page reload.

• Cited workpaper cell references in Excel exports no longer fail on large workpapers: When CoAudit builds an Excel export that cites specific workpaper cells, it can now anchor those citations directly from the workpaper data it already scanned — without going back to re-read source documents a second time. This resolves a failure that previously blocked cited Excel exports when working across a large workpaper.

New in CoAudit

• Clickable citations in AI-produced Excel workbooks: When CoAudit builds an Excel workbook for you, individual cells can now include citations that link directly to the source document pages — the same click-to-open experience you get from a workpaper exported in Punchcard. Ask CoAudit to build you a workbook and it will wire up the citations automatically.
• CoAudit can now delegate complex tasks to worker agents: For requests that require reading many files and authoring detailed evidence across a large workpaper, CoAudit can now hand off that exploration to a focused worker — keeping your conversation responsive while the heavy lifting runs. This capability was previously available only to administrators; it is now part of every task-mode CoAudit chat.
• Filter workpaper rows by review state: CoAudit can now scope its reads and actions to only the rows you've flagged, validated, or haven't reviewed yet. Ask it to "look at only the flagged rows" or "re-run only unreviewed items" and it will target exactly that subset.
• Automation re-runs preserve your reviewed work: CoAudit now correctly understands that re-running a workpaper automation never overwrites cells you've already validated or flagged — so you can safely ask it to refresh the rest of the workpaper without worrying about losing your review work.
• Upgraded AI model: CoAudit chat now runs on an upgraded model with noticeably better accuracy on complex routing and multi-step tasks.

Chat & UI

• Drag folders and ZIP files into the chat: You can now drop a folder (or several folders at once), or a ZIP archive, directly into the CoAudit chat composer — all files inside are collected and uploaded as individual attachments. Previously only individual files could be dragged in.
• Scroll position stays put when the AI finishes responding: When the AI completes a reply, your reading position is no longer snapped to the bottom of the conversation. If you scrolled up to read an earlier message, you stay there. The scroll-to-bottom button becomes highlighted with a bouncing arrow to signal that a new reply is waiting — clicking it or scrolling down clears the indicator.
• Redesigned AI loading indicator: The indicator shown while the AI is generating a response is now a slim, solid bar pinned at the bottom of the conversation with an animated Punchcard mark and status text. It no longer lets transcript content bleed through behind it.

Requests

• Simplified request status labels: Client-facing request statuses are now Outstanding, Returned, Submitted, and Accepted. The button clients use to submit work is now labeled Submit Request; when a request is already submitted, it shows Return to Outstanding instead. Both auditors and clients will see a one-time announcement on the board summarizing the label changes side-by-side.

Bug Fixes

• Testing attribute name saves reliably on the memo page: Typing a name and then quickly reloading the page or closing the tab no longer silently discards the change — it now saves the moment you click away from the field.
• ZIP file previews work correctly in CoAudit chat: Files extracted from an uploaded ZIP now preview correctly in the document pane instead of showing a fetch error.
• PDF black-page recovery improved: Black or blank PDF pages are now detected and recovered from the moment a document opens, not only when navigating between pages.
• Collective workpaper columns compute correctly across multiple files: A column configured to produce one shared result across all files in a datasource group now generates a single synchronized value and applies it to every file, rather than running independently per file.
• Excel workbook delivery is more reliable: In rare cases, generating an Excel artifact could fail due to an internal step unrelated to the workbook's content. That step now fails gracefully so your workbook is still delivered.
• Task breadcrumb segments no longer collapse: Breadcrumb text in the task header now maintains a visible minimum width instead of occasionally collapsing to nothing.

Features

• Workpaper exports match your current sort order. Excel exports — Workpaper Only, Files + Workpaper, and Table Data Only — now output rows in the exact order displayed in the grid. If you sorted by balance largest-to-smallest before exporting, the downloaded file reflects that, with row numbers renumbered to match. Your sort is also saved automatically and restored when you navigate away and come back.

• Clickable source chips in the cell panel. When you open a cell's detail panel to view the prompt or formula CoAudit used, values that were drawn from other columns now appear as highlighted, clickable chips. Click a chip to jump directly to the source cell in the grid; a back arrow in the panel header lets you return to where you started. This makes it straightforward to trace exactly which upstream values shaped a given result.

• Skill mentions appear as pills in the main chat view. When you type an @[Skill name] mention in the standalone chat, it now renders as a blue pill — matching the visual treatment in the task chat sidebar — so it's immediately clear the skill was recognized and will be applied when you send the message.

Bug Fixes

• Engagement variable form stays open after a failed save. If you try to create a variable with a name that already exists, the form now stays open with your entries intact so you can correct and retry — instead of closing as though it had saved successfully.

• Bulk file uploads recover from temporary conflicts automatically. Uploading a large batch of files at once (common in payroll workflows) no longer surfaces intermittent "Upload Failed" errors when multiple files finish around the same time. Conflicts are resolved and retried automatically in the background.

• Large PDFs no longer trigger a false re-upload prompt. Files that take 10–15 minutes to extract — such as large payroll PDFs — sometimes showed a timeout warning and prompted a re-upload even though processing was still running and would have completed. That false alarm no longer occurs.

Features

• Import Deep Analysis questions into Skills: Each existing Deep Analysis question now has an "Import to Skill" button that opens the Skills editor pre-filled with the question's content and name. The option to add new Deep Analysis questions has been replaced with a link to the Skills library — existing questions continue to run, edit, and delete as before.

Reliability Fixes

• Files still processing after upload no longer appear as failed: When a large file takes longer than expected to finish text recognition, the workpaper now shows a "still processing" notice rather than marking it as failed. The status updates automatically once processing completes, so there is no need to re-upload.

• Failed saves now revert to the last confirmed value: Across status changes, memo fields, review states, task names, and descriptions, a save that doesn't go through now reliably restores the field to the last value the server confirmed, instead of leaving it stuck in the unsaved state.

• Switching tasks no longer mixes up edits between tasks: A save started on one task will no longer affect the task you navigate to afterward. Each save stays in the context of the task where it was made.

• Review status changes now update all related items immediately: Accepting or clearing a review state on one workpaper item in a linked group now immediately reflects on the other items in that group in the same view, without requiring a refresh.

• Narrative card review badges stay in sync: Review status badges on narrative cards now consistently reflect the latest confirmed state, even when a save and a concurrent change arrive at the same time.

Features

Safeguards for files shared across workpapers

When you move or delete a file in a document group used by more than one workpaper, Punchcard now shows a confirmation listing exactly which other workpapers will gain or lose access to that file. This applies to all file-explorer entry points: menus, drag-and-drop moves, bulk moves, file replacements, and deletions.

• Activity stream awareness: file moves and deletions in shared groups now appear in the activity stream of every affected workpaper, attributed to the person who made the change.
• Real-time notices: if a colleague moves or deletes a file from a group your workpaper shares, a brief notice appears so you're aware without needing to refresh.
• CoAudit approval cards: when CoAudit moves a file between document groups, the approval card now lists which other workpapers are affected before you confirm.

CoAudit: smarter PDF-to-Excel conversion

Converting a PDF to Excel now detects when a document's layout cannot be reliably extracted — such as borderless tables or payroll reports where columns are positionally arranged rather than tabular. Instead of producing a useless single-column spreadsheet, CoAudit guides you through a stepped escalation: first upgrading the document's text recognition, then applying geometric layout analysis if needed.

A "Re-run with higher-quality text recognition" option is also now available in CoAudit for any PDF processed with the standard engine — useful when a document's tables came out garbled or collapsed into a single column.

CoAudit: citations in AI-edited workpaper cells

When CoAudit edits a workpaper cell directly in chat, the cell now renders with full citation quality: clickable document highlights, inline citation pills woven through the reasoning, and clearly separated supporting and counter-evidence. Previously, cells written via CoAudit chat showed the value and reasoning but no linked citations.

CoAudit: full citation coverage tracking

CoAudit's citation-quality check now measures against all cells that should have citations — matched document rows across all retrieval columns — not only cells where citations were already attempted. Cells that were never filled and datasources with no matched files now surface as gaps, so the coverage report reflects the complete workpaper rather than only what was attempted.

CoAudit: aggregate and group spreadsheet data

CoAudit can now sum, count, average, and group data from spreadsheet files in a single query. For example, asking "What is the total amount by vendor?" returns grouped subtotals directly, without needing to browse every row first.

Correcting auto-sorted files from the activity stream

When Punchcard auto-sorts a file into a document group, the activity stream now shows a "Correct this action" link on that entry. Clicking it navigates the file explorer to the destination group, pre-selects the moved files, and shows an inline prompt to move them — making routing corrections a guided one-step flow.

Improvements

• Testing formulas — blank cell comparisons: a formula comparing two workpaper cells (e.g. comparing an extracted amount against a budget figure) now correctly shows as failing when one of the referenced cells has no value. Previously, a blank cell could cause the comparison to show as passing (green ✓), potentially producing misleading conclusions in testing columns.
• CoAudit collective columns: when CoAudit edits a collective (merged) column in chat, it now checks all rows sharing that merged value before writing, so changes to any member row never silently overwrite the entire group.
• Financial Statement Review sidebar: the "Open a current year financial statement" prompt now only appears when a non-current-year file is already open. It was previously visible in the empty state where clicking it had no visible effect.

Bug Fixes

• Ghost file entries: files whose uploads were abandoned mid-way (tab closed, network dropped) no longer appear as broken entries in the workpaper. A file now only appears after its upload is fully confirmed on the server.
• Unsupported file types: uploading a file type Punchcard can't extract text from (such as .html) no longer fails the entire upload batch. The file is kept and available for download, with a non-blocking warning shown in the upload progress card.
• Clicking a second file cell: clicking one workpaper cell to open a document, then clicking a different cell's file count, now correctly switches to the new cell's file list. Previously the viewer stayed pinned to the first cell's document.
• Word document viewer: when a Word file fails to display (for example, an older format or a file that cannot be rendered), the viewer now offers a "Try again" button and a "Download" option. Previously it only offered a full page reload, which would fail identically for the same file.
• Memo action buttons: the delete, move, and open buttons in the Memo tab's accounting-policy and testing-attribute rows no longer flicker and become unclickable as you move your cursor from the row toward the button.
• Incomplete AI-generated artifacts: if CoAudit's environment runs out of memory while building an Excel or Word file, the incomplete file is now discarded rather than delivered as if it were complete. The failure is reported clearly so you can retry with a smaller scope.
• Accessibility: the "Select All" checkbox now correctly communicates a partial-selection state to screen readers and assistive technology when only some items in a list are selected.

Bug Fixes

• Clicking a file in a request now opens it immediately. When browsing files attached to a request, clicking a file was dropping you to a top-level folder view instead of opening the file directly. You had to click a second time to actually see the file. That extra step is gone — one click now goes straight to the viewer.

• Image files no longer show two close buttons. When opening an image file (PNG, JPG, and similar formats) in the viewer, two close (×) buttons appeared in the header simultaneously. Only one close button now appears, as expected.

Performance

• Large Excel uploads no longer freeze the browser tab. Uploading large spreadsheets previously caused the tab to lock up for a minute or more and sometimes left a file that appeared in the list but could never be downloaded. Uploads now complete in under a second regardless of file size.

AI & Sandbox

• Co-Audit Agent sandbox steps now show what they're doing. When the agent runs code to build a report or analyze data, each step is labeled in plain English — for example, "Building Excel workbook" — instead of a static spinner. The underlying code is still available behind a "View code" toggle for those who want it.
• The Stop button now immediately halts sandbox work. Pressing Stop while the agent is executing code interrupts all further steps right away rather than letting them run to completion.

Workpaper

• Circular-reference detection covers more scenarios. Punchcard now catches circular dependencies that arise when pasting data with a different column mapping, when a match relationship is auto-linked, or when a match mode is switched. When a loop is found, the warning banner displays the exact chain as clickable column pills so you can jump straight to the involved column.
• Batch run approvals now name each column. When the Co-Audit Agent queues several column-level runs for approval at once, each entry in the approval card now shows the specific column name and its document group. Previously all entries showed only the document group, making a batch of runs on the same group indistinguishable.
• Editing a deleted column now shows a clear message. If you try to update a column that a teammate has already deleted, Punchcard now explains that the column no longer exists instead of showing a generic error.

Reliability

• Extraction progress no longer flickers. The Review & Corrections panel in the chat sidebar stays steady while an extraction is running instead of flashing repeatedly.
• Financial Statement Review consistency checks are more stable. A rare scenario where two consistency rules targeted the same entry at the same time could cause the analysis to stop partway through. This is now handled gracefully.
• Logging in from multiple locations no longer produces an error. Opening Punchcard in a second tab or browser at the same time as an existing session previously triggered a server error in some cases; it now resolves cleanly.

CoAudit Workpaper

• Circular column dependencies are blocked the moment you create them: if a formula, column reference, acceptance criteria prompt, or column name change would form a cycle with existing columns, the edit is rejected before it saves — the warning banner appears immediately with clickable column badges identifying every column involved, so you can correct the issue without navigating away. Only the specific edit that would introduce the new cycle is blocked; unrelated changes and recovery edits on a workpaper that already has a pre-existing cycle proceed normally.

• Clear message when file association conflicts with a running automation: if you try to match or drop a file onto records that are currently being updated by a running automation, you now receive an actionable message — "These records are being updated by another operation. Please retry in a moment." — instead of the operation hanging silently for several minutes before failing.

Activity Feed

• Client filter now correctly scopes the feed: selecting a client organization in the activity feed filter now actually narrows results to that client's activity. Previously, choosing a client appeared to apply a filter but continued showing activity across all clients.

AI Chat

• More reliable during long or data-heavy sessions: improvements to how the AI assistant manages context across extended audit sessions reduce the likelihood of unexpected failures or incomplete responses when working through large workpapers or multi-step automations.

General

• Expired sessions redirect to sign-in: when your session expires and cannot be automatically renewed, you are now taken to the sign-in screen instead of encountering a confusing error with no clear path forward.

• Simultaneous tick mark applications no longer conflict: if two auditors apply the same tick mark at the same time, both operations now complete successfully instead of one of them returning an error.

Workpaper Match Actions

• Adding a match now appears in the grid immediately — clicking "Select Match" in the cell detail panel reflects the new association right away. Previously the grid cell stayed blank until you refreshed the page, even after a successful match.
• Removing a match from the cell detail panel now works correctly — matches cleared through the panel are actually removed and no longer reappear after a refresh.
• Moving a match that partially fails now shows an accurate result — when some records couldn't be moved because they were protected, you now see which records were kept rather than a false "Match moved" confirmation.
• Adding a match no longer causes a brief flash on the match button or flicker in the cell detail tab bar.

AI Cell Reasoning

• Cell reasoning now leads with the decisive basis — the first line states the key finding directly; each following line presents one supporting fact with its source citation. Lines prefixed with ✗ (shown in red) call out discrepancies or exceptions the AI encountered and how it resolved them. This makes it easy to scan what the AI found without reading through a paragraph of prose. Cells generated before this change continue to display as before.

Lasso & Citations

• Drawing a new lasso makes that region the primary citation — the newly drawn bounding box becomes the top citation for the cell, displacing the previous primary. Citations still referenced in the cell's reasoning text are preserved; stale ones are removed. This applies to single-lasso, sum-lasso, and citation-removal paths.
• Drawing a lasso on a match or tick column attaches evidence without changing the column's result — the pass/fail outcome or match state is preserved; only the supporting citation is updated.
• Lassoing any cell in a Collective-mode column syncs all siblings — the same value and primary citation are applied to the whole group. Each sibling keeps its own individually referenced citations.
• Fewer unresolved citation pills on complex financial statements — citations in match and testing attribute columns now reference precise locations in source documents using structural coordinates, virtually eliminating greyed-out (unresolved) pills on dense tables and financial-statement layouts.

Re-Pasting Data into a Workpaper

• You can now explicitly skip mapping a column — the column mapping step (shown when you replace workpaper data that has downstream references) now always offers a clear skip option. For match columns the default is still to remove the old column; for columns referenced by a prompt or formula, the option is available but not pre-selected, so you have to make a deliberate choice to proceed.
• The mapping step shows only when something would actually break — if every referenced column still has a matching name in the new data, the step is skipped automatically and your references carry across correctly. The step now also correctly detects when a match column depends on an entire datasource, not just a single named column.
• Unresolved columns warn rather than block — if a prompt-referenced column is left unmapped, the modal explains what will be affected (including which specific columns) and lets you continue and fix the prompt later, rather than preventing the save entirely.

PDF Viewer

• The viewer recovers automatically when the browser reclaims display resources — if a blank or black page appears due to the browser dropping the rendering canvas (common when switching between many tabs), the viewer silently remounts and restores your current page without any action on your part. A built-in limit prevents this from looping if recovery fails repeatedly.
• PDFs open faster in multi-panel views — files are fetched and cached once per session and shared across all panels that show the same document, eliminating redundant downloads when the same file is open in multiple places.

Other Fixes

• Panel resize handles are much easier to grab — the draggable divider between panels in the task view now has a wider interactive area, so you no longer need to position the cursor precisely on a 2-pixel line to resize.
• Chat accurately reports when a document isn't ready — if a file's text hasn't been extracted yet or extraction failed, the chat names the file and explains the specific status (still processing, no extractable text, or failed) instead of silently skipping it or showing a generic placeholder.
• Financial Statement Review manual footings no longer flag scaled columns incorrectly — manual footings in TEUR or "in thousands" columns are now validated using the same rounding tolerance as automated footings, so legitimate rounding residuals are no longer flagged as discrepancies.

New

• Run AI testing on exactly the cells you select: Drag to select any range of cells in the workpaper grid (or Ctrl+click individual cells), then choose Run from the right-click menu or the panel that appears. AI testing runs only on those cells — fill empty values only, or overwrite existing ones — and the confirmation dialog shows a count per document group before you commit.

• CoAudit fills cells and creates matches with reasoning and evidence from chat: Ask CoAudit to fill in a specific cell value or match a document to a row, and the result now appears with AI reasoning, supporting evidence, and clickable citation highlights in the document viewer — stamped as AI-authored so it is distinguishable from manual edits. Writing over already-reviewed or locked cells requires your approval before the change is applied.

• Switch between files when reviewing a collective cell: When a workpaper cell aggregates values from multiple files, the filename label in the document viewer header is now a dropdown listing every contributing file. Click to switch — and any citation you draw after switching is saved against the file you are viewing, not the original.

Improvements

• More citations resolve to clickable document highlights: Citations on complex financial statements now resolve to highlighted regions far more reliably, including values in non-contiguous table cells (such as a label on one row and its amount on another) and scattered multi-value fields (such as a list of dates or amounts across different lines). Previously many of these appeared as unlinked text with no highlight in the PDF viewer.

• The free-text "Something else" option is always present on AI questions: When CoAudit presents a multiple-choice question in the chat, the open-ended "Something else…" text field is now always shown — you are never limited to only the options the AI offered.

• Number and date formats always match the engagement's firm: Workpaper cell values, pasted data, and sampling imports now consistently apply the number separators and date order configured for the engagement's firm — even when your account is pointed at a different firm. Previously, a European firm's amounts could display or import incorrectly in some sessions.

• Accepted and rejected cell states update the grid immediately: Accepting or rejecting a cell value is now reflected in the workpaper grid right away, without needing to refresh the page.

• Scoped runs show only the relevant progress: When CoAudit runs AI testing on a specific column or document group rather than the whole workpaper, the in-progress indicator now lists only the items being processed.

• Auto-generated chat titles match the conversation language: When CoAudit automatically names a chat, the title is now written in the same language as the conversation.

Bug Fixes

• Citations drawn on a sibling file now save correctly: After switching to a different file in the collective cell viewer and drawing a citation, the annotation is saved against the file you were viewing — not the original file the cell pointed to.

• Fill-empty runs now reach downstream document groups after a manual match: After manually accepting a file match in an upstream document group, a subsequent fill-empty run now correctly processes the newly eligible downstream rows instead of skipping them.

Bug Fixes

• Citations inside table-formatted content now link correctly. When a chat response presented information in a table, citations embedded in table cells previously failed to link back to the source document. They now resolve as expected.
• Workpaper automations apply your firm's preferred language throughout. Each step in a multi-step workflow analysis now uses the firm's configured language, so AI evaluation responses are consistent end-to-end rather than reverting to a default.
• Workpaper automations are more resilient. An optional lookup that previously could abort an entire automation run on a transient error now recovers gracefully, keeping the rest of the run on track.

Features

• AI responds in your language. CoAudit chat, cell reasoning, match agent, testing attributes, review checklists, and math validation notes all now reply in the language you write in. Write in French, get French back. English is the fallback when the language is ambiguous.

• Sampling handles European and regional number formats. Pasting EU-formatted amounts like 1.234,56 (dot as thousands separator, comma as decimal) is now recognized and parsed correctly without manual reformatting. MUS, Random, and Journal Entry threshold inputs — performance materiality, tolerable misstatement, floor amounts — also accept your firm's regional number format. Dates in day-first firms (e.g. DD/MM/YYYY) resolve correctly when the firm's format is configured; format tokens are case-insensitive, so dd/mm/yyyy and DD/MM/YYYY both work.

• Sampling PDF receipts use the firm's currency and date format. Currency amounts and dates in sampling reports now display in your firm's configured format. Currency symbols that can't be embedded in standard PDF fonts (₹, ₩, ₪, ₺, ₽, ₫) are shown as their standard three-letter codes (INR, KRW, ILS, TRY, RUB, VND) so values remain readable.

• Currency declared in source documents is preserved in the workpaper. When the AI extracts a dollar amount from a document and your firm is configured for euros, the cell now shows the dollar sign while the firm's separator style still applies. The source document's currency no longer gets replaced by the firm's display currency.

Workpaper improvements

• Multi-question AI cards now show a visible Next button. When the AI asks a series of questions, the primary button reads Next on each question (enabled once the current question is answered) and morphs to Submit on the final question. The previous keyboard shortcuts still work; the button makes the forward path discoverable.

• Arrow and Tab navigation no longer changes the file viewer or side panel. Moving between cells with arrow or Tab keys now updates the cell detail panel only. Opening the file viewer or switching the right sidebar still requires an explicit click.

• Escape key follows a clear priority order in the Workpaper. If the CoAudit chat is expanded full-screen, the first Escape collapses it. If a cell is selected, the next Escape clears it. If either side panel is open, the next Escape closes both.

• Workpaper load failures now show a Retry button. If the workpaper fails to load (for example, during a brief interruption), the page now shows a friendly message and a Retry button instead of a loading spinner with no way to recover.

• Re-run now always forces fresh results. Previously, re-running automations could silently produce no changes when the underlying data appeared unchanged. The Re-run action now always triggers a full re-evaluation.

• Circular column dependencies surface an immediate error. If two columns reference each other in a cycle, automations now fail right away with a clear message identifying the involved columns, instead of silently retrying.

• Bulk validate and flag now work correctly across multi-cell selections. Using Ctrl+click to select cells across multiple rows and then validating or flagging them in bulk now processes all selected cells, not just the first group.

• Citation highlights appear immediately after a match automation run. Clicking a citation in a freshly matched cell now highlights the correct location in the document viewer without needing a page refresh.

• Testing attribute results appear immediately after running. Pass (✓) and fail (✗) verdicts, evidence bullets, and citation highlights now appear in the grid as soon as a testing attribute automation completes, without requiring a page refresh.

• Validated cells no longer show anomaly indicator badges. Cells that have been accepted/validated no longer display the amber count badge. Rejected cells retain the badge since the anomalies are typically the reason for rejection.

• More citations now correctly highlight in the document viewer. The AI sometimes returns citations containing formatted text (HTML table markup, ellipsis spans, punctuation) that previously couldn't be located in the source document, leaving citation pills with no highlight. These citations now resolve correctly for a wider range of document formats.

Fixes

• Deleting a testing attribute in the Memo now confirms the result. If the server was unable to remove the attribute, the row is automatically restored and an error message is shown, rather than appearing deleted until the next page load.

• Sampling receipts are no longer downloaded automatically when saving. Clicking Save Data no longer triggers an immediate Excel and PDF download. The receipt is now included in the workpaper export. The per-datasource "Download sampling report" button still provides on-demand receipt downloads.

• Opening Notification Settings no longer freezes the page. Previously, opening and then closing the Notification Settings dialog could leave the page unresponsive until a full reload. This is now fixed.

• @mention suggestions no longer cover the word being typed. The suggestion popover now positions itself so it doesn't obscure the text you are typing, and the list scrolls correctly for long lists of results.

• Evidence and citation edits in the cell detail panel no longer silently fail. Editing or deleting an evidence bullet on a cell whose citation was matched from plain text (rather than a bounding box region) now saves correctly.

• Downloading a deleted file from the activity feed now shows a clear message. If a referenced file has been removed, clicking its download button now shows a "File unavailable" notice instead of failing silently.

• Double scrollbar in the chat sidebar on Windows and Linux is fixed. The CoAudit chat sidebar no longer renders its own vertical scrollbar alongside the message list's.

Skills

• First-time Skills intro: New users now see a compact, one-time prompt anchored to the chat composer's "+" button introducing the Skills feature — "Save prompts as Skills: write a prompt once, reuse it in any chat, for yourself or your whole firm." Dismissing it in any chat silences it everywhere.
• Typing / always opens the Skills picker: The / shortcut in the chat composer now opens the picker even when you have no skills yet, showing a "No skills yet" message with direct links to browse or create your first skill. Previously the picker stayed hidden entirely, leaving no path forward.

Workpaper

• Redesigned "Informed by" section: The panel section explaining how an AI-produced cell value was derived now shows a concise bullet list — one bullet per contributing document group — instead of an expandable recursive tree. Each bullet names the group and describes in plain language how that group's data shaped the result, with inline document citations you can click to jump to the supporting page.
• Value references open the source document in place: Clicking a referenced upstream value in the "Informed by" section now opens the document viewer at the page where that value came from — without navigating away from your currently selected cell. You can verify the evidence behind each input while staying in context.
• "..." file menu now matches the right-click menu: The three-dot button on each file row in the file explorer now offers the same options as right-clicking — including Move to another datasource and Undo Split for multi-segment files, which were previously available only via right-click.
• Undo Split refreshes the file list immediately: After merging a split file's segments back into a single document, the file list now updates automatically — no page refresh required.
• Document viewer recovers gracefully from rendering failures: If a file can't be displayed (for example, due to a corrupted PDF), the viewer now shows a clear message with a Try again button and a Download option. The rest of the workpaper — the grid, sidebar, and all other panels — stays fully functional. Previously, a viewer failure could take down the entire workpaper page.
• Workpaper Excel export uses hierarchical row numbers: The "#" column in exported Excel workpapers now shows the same hierarchical address (e.g. 1.1, 1.2, 2.1) as the workpaper grid view, rather than a flat sequential counter.

Co-Audit Chat

• Working indicator stays visible for the full response: The activity indicator now remains on screen for the entire duration of a Co-Audit response — including brief pauses between steps — instead of disappearing early while the Stop button was still showing.
• Escape key correctly collapses the expanded chat: When the chat panel is open in full-screen mode, pressing Escape now collapses it back to the sidebar. Previously, Escape navigated away from the task instead of collapsing the chat.

File Processing

• OCR progress updates in real time: Files being processed for text recognition now show live status updates, replacing a frozen spinner that previously gave no indication of progress.

Reliability

• Login fixed for users at organizations sharing an email domain: Users at firms where multiple accounts shared the same email domain could encounter login failures on every attempt. This is now resolved.

New: AI Skills

Skills are named, reusable sets of AI instructions you save once and reference anywhere in your audit work — a Personal skill is private to you; a Firm skill is shared with everyone at your firm.

• Use a Skill in CoAudit chat by typing / or selecting one from the "+" Skills menu in the composer. The skill's instructions are automatically included when you send — you can still add your own context before hitting send.
• Reference a Skill in workpaper automation columns by typing / in any retrieval, match, or deep-analysis prompt field. Firm and Punchcard skills are available here; the skill's instructions are applied automatically at run time across every record.
• Browse and manage all your Skills in the Skills library, opened from the "+" menu in CoAudit chat. Search by name or description, filter by Personal / Firm / Punchcard scope, and view, edit, or delete skills you have access to. Saving a change to a Firm skill shows a confirmation before it affects your whole team.
• Co-Audit can author and refine Skills directly from the chat — ask it to save the current approach as a reusable skill without leaving the conversation. Saving a Firm skill from chat requires your confirmation first.
• Sent messages display skill names as inline badges; clicking a badge opens the skill's full detail view.
• The Skills editor supports full formatting — Enter, undo, bold, italic, and list indentation work as expected.

Memo

• Testing attribute descriptions in the Memo now use the same rich-text editor as the Audit Procedure field on the workpaper — format text and include assertion references without the description reverting to plain text or showing markup on reopen.
• Co-Audit now treats Testing Attributes as part of the memo. Asking it to "complete the memo" or "fill in the memo" will include the testing attributes alongside the narrative sections, rather than treating them as a workpaper-only concern.

Request Table

• The amber acceptance-criteria badge on a failing file upload is now a clickable button that opens the system validation notes for that file directly — no need to navigate through the Files tab first.
• Auditor-internal notes written against specific workpaper values no longer appear in the shared client-facing columns of the Request table.
• When a cell contains multiple files, the warning badge now correctly identifies the specific file that failed acceptance criteria rather than defaulting to the first file in the cell.

Board

• The selection checkbox in board list view now appears when you hover anywhere on a row, not just directly over the small checkbox icon.
• Tooltips in board list view no longer clip behind the sticky row header.

Engagement Settings

• The Private engagement toggle is now clearly shown as disabled — with an explanatory tooltip — for auditors who don't have permission to change an engagement's privacy setting. Previously the toggle appeared enabled, allowed a click, and then silently reverted, with no explanation.

Bug Fixes

• Opening an engagement link you don't have access to now shows an access-denied message immediately instead of displaying a loading screen that never resolves.
• Inviting a team member or client no longer stays stuck on "Sending…" indefinitely. If there's a delay, the form surfaces an error message so you can try again.
• The task chat panel and task overview no longer go blank during certain navigation or AI-streaming transitions.
• The @-mention suggestion popup in text editors no longer causes a crash when you close a dialog while the suggestion list is still open.
• When asking Co-Audit to create a custom file in the chat, it can now load any task file using the same references it uses for reading and searching — files that were visible but couldn't be processed before will now work correctly.