What's new
New: charts for your hypotheses
October 2026
Next to the table, your hypotheses now have a Charts view. You find it in the main navigation under "Hypotheses".
- Quick-win matrix. Every hypothesis with a potential and a rating is placed by its potential against its effort - or, at a click, its complexity. Four quadrants show at a glance which hypotheses are quick wins, which need strategic planning, which are nice to have and which can be deprioritised, each with its number of hypotheses and total potential. Hypotheses that sit close together form one bubble sized by their count, and its colour shows the confidence.
- Hypothesis distribution per data field. Choose a data field, and one column per value shows how many hypotheses are assigned to it - stacked by status, as a number or as potential. Under "Configure data values" you pick up to 10 values, set their order by drag and drop, and decide whether the "Not assigned" column is shown. The setting is saved for you per workspace.
- Same filters as the table. Search and filters carry over when you switch between table and charts, so the charts always show exactly the hypotheses you are looking at.
- From the chart to the hypotheses. A bubble with a single hypothesis opens it directly. Click a bubble of several hypotheses or a column to list the hypotheses behind it, and open that selection in the table view with "Show selection in table view".
New: the Data Explorer - see your data as charts
October 2026
The Data Explorer turns the tables in your workspace into charts, so you can spot where a hypothesis is worth looking for before you start searching. You find it in the main navigation, in the "Find hypotheses" section.
- Build an analysis in a few clicks. Choose what to group by - a data field or a column of a file directly - and which value to calculate: the sum, minimum, maximum or average of a column, per document.
- Bar, pie or matrix. One grouping gives a bar or pie chart; two groupings give a matrix that compares the combinations at a glance - including empty ones where nothing has been recorded yet.
- Always up to date. Charts are calculated in real time from your current data whenever you open them; nothing is stored in between.
- From a number to the rows behind it. Click a bar, a slice or a matrix cell to see the underlying rows, and use "Discuss in expert chat" to hand that selection over to the assistant together with the matching documents.
- Save and share. Analyses are saved, can be described, searched and sorted, and stay private until you share them with your workspace.
New: copy an answer, or fork the chat from it
September 2026
Every completed answer in the Expert Chat now has two actions.
- Copy the answer. Copies its Markdown to the clipboard, so headings, lists, tables, and highlights survive when you paste it elsewhere.
- Continue in a new Expert Chat from here. Opens a new chat that carries over the transcript up to that answer, together with the source chat's documents and settings. The new chat always starts private, and your personal MCP connections come along only when you fork your own chat.
Both actions also work on a chat a colleague has shared with you in your workspace.
New: share a chat with your workspace
September 2026
You can now share an Expert Chat session read-only with everyone in its workspace, so colleagues can follow the analysis without you sending screenshots.
- Share from the chat. Turn sharing on to give every workspace member view access to the running transcript; turn it off again at any time to make the chat private once more.
- Read-only for everyone else. People you share with can read the conversation and watch it update, but cannot write messages, submit options, or trigger actions like promoting or dismissing a hypothesis.
- A lock icon marks private chats. In your chat history, chats you have not shared show a lock icon so you can tell at a glance which ones only you can see.
New: data fields - one structure across all your documents
September 2026
Every document brings its own column names. Data fields put one shared structure on top: you define a field such as "Supplier", "Category" or "Region" once and map the matching column of each document onto it. Its values are then collected across every mapped document, so the same supplier is a single value no matter which file it came from.
- Set up under Data Sources. The "Data fields" tab is where you create a field and map document columns onto it; the values are read from your files automatically.
- The assistant knows your fields. It can refer to them in the chat and offer you the actual values from your own data instead of free text.
- Hypotheses are linked to field values. When a hypothesis is created, the assistant proposes which values it is about. That stays a proposal until you decide: confirm it, reject it, or mark the field as not applicable on the hypothesis.
- Visible in the hypotheses overview. Each row shows how many field values it carries, and one click opens the details per field.
- Filter by field value. Under "Additional filters" you pick a field, an operator ("is any of", "is all of", "is none of") and its values - so you can evaluate your hypotheses per supplier, per category, or any combination of the two. "Not assigned" finds the hypotheses a field was never set on.
The assignment is a curated, incomplete index: a missing assignment is not a statement that the hypothesis has nothing to do with a value.
New: every table you upload is checked for data quality
September 2026
Spreadsheets and CSV files are now checked automatically before they are used in an analysis, so a problem in the source file no longer turns into a wrong number later.
- A quality report for every table document. In the Data Center each file shows what was found, per worksheet and per column, together with what to do about it - plus the detected format, the number of distinct and empty values and a control total for each column.
- Common problems are corrected automatically. A header row that is not the first row is recognised, total rows at the end of a sheet are removed when they really are totals, several tables sitting side by side on one sheet are separated, and pivot output is kept out of the analysis.
- Anything ambiguous is flagged instead of guessed. A missing currency or unit, an unclear day/month order, subtotals in the middle of the data, values that contradict their column header: you get a note instead of a silent interpretation.
- The assistant knows about the findings. Every table query and document analysis is told what to watch out for, so totals are not counted twice and numbers stored as text are still read as numbers.
- Files that cannot be read are stopped immediately. A file whose content does not match its extension is reported with a clear reason instead of being imported incompletely.
- New checks reach documents you uploaded earlier. Those still checked by an older version are marked "Checks outdated"; "Re-run quality checks" brings them up to date.
Updated: document analysis now covers several documents together
September 2026
The document analysis no longer offers "All documents". Instead you pick up to 8 specific documents, and a run analyses them together, so figures spanning several spreadsheets and CSV files are compared directly. Insights that connect documents are saved for future analyses. Scheduled checks that used "All documents" were removed automatically; please recreate them with the documents you want covered.
Updated: how hypothesis confidence is scored
August 2026
We recalibrated how a hypothesis's confidence score weighs its supporting evidence - every factor now counts, not just the strongest one, so a hypothesis backed by a mix of strong and weak evidence may show a different score than before. Existing hypotheses were re-scored automatically; the breakdown shown for older ones is a best-effort estimate, since the original AI assessment behind them was not stored.
New: feature announcements in your notifications
August 2026
From now on you will receive a notification here whenever we release a noteworthy new feature - so you always know what Aiguro can do for you.