October 2026

18 October 

• Agents - you can now save a set of instructions as your own agent and run it whenever you need. You can review each run afterwards and continue it as a chat.
• Artifacts - the assistant now writes reports beside the chat, with sections, tables, key figures and charts. They stay after the chat ends and are listed in the Artifacts tab.

11 October 

• Chat list status - you can now see at a glance which chats are running and which need your input.
• Follow-up messages - sending a message while the agent is still working now joins the turn already running, instead of starting a second one.

4 October 

• Copy tables - you can now copy a table from the table data view straight to your clipboard.
• More chart options - charts now support area and donut types, show several series at once, and let you choose colours per series or pie slice.
• Pick up where you left off - when you reopen a chat while the assistant is still working, you now see its reply as it continues, and you can stop it.

The October release brings faster set-up of new document types in AIDP, realistic record previews in data prep, more ways to investigate across linked reconciliations and richer data in your reporting tools.

Ingestion

Extract checkboxes without training a model

In AIDP, zero-shot extraction now picks up checkbox state - ticked or not ticked - just like any other field. Until now, checkboxes could only be extracted with a trained model, so a single tick box on a form meant going back to annotating documents and training a model before the document type could go live.

You can now set up and take live forms that include tick boxes, such as confirmations, account opening forms, KYC forms and agency notices, using zero-shot extraction alone - no annotation or model training needed.

More accurate agentic extraction

We've improved agentic extraction in AIDP so it produces more accurate results straight away, including on documents with complex layouts. A new document type can now return useful results from the moment you set it up, without waiting for a model to be trained first.

Preview your data prep process with real sample data

You can now upload a real, full-volume sample file to any structured data or semi-structured input in data prep, and Duco will use it to build the full preview of your process - map and transform, aggregation and output. 

Previously, previews were based on a small 20-row sample, which made it hard to check how your rules, and especially aggregation, would behave on real data.

Samples are read using the same parsers as your production runs, so what you see in the preview reflects what you’ll get when the process runs. Files of up to 50 MB are supported, and zip files are accepted.

Records from your sample file will be visible in the new ‘Preview’ tab, as well as the errors will be reflected in the ‘Errors’ tab, showing you straight away where the fixes should be applied.

What that unlocks day-to-day:

• Upload, replace or drop a sample directly from the data input card, or simply drag a file onto the card.
• Open the full parsed sample from the Preview button in the Fields column, and filter, sort and page through it.
• Spot issues early - up to 100 parsing errors are listed alongside the sample, and the Preview button shows a live error count.

Please note that uploads of data samples are not yet available for AIDP inputs.

Reconciliation

Clearer workflow rules configuration

The auto-matched status is no longer available as a match status condition in workflow rules in single-sided, two-sided and cash reconciliations. 

Workflow rules only apply to exceptions, and auto-matched records are not considered exceptions, so a rule based on this status would not take effect. Removing it makes it clearer which rules can  be applied at the moment when you build them.

3 way reconciliation view: more ways to investigate and set up

We’ve continued building out a 3 way reconciliation view, which brings exceptions from linked reconciliations together in one place. This release adds:

Investigating exceptions

• A Details and History panel in the results screen, so you can review a record’s details and workflow history without leaving the multi recon view.
• Filter by breaks, so you can focus on records breaking on a specific field.
• Cell commenting, so you can add a comment against the exact value that needs attention.

Setting up and managing

• Field mappings now use a dropdown of your actual fields instead of data types, making it quicker to map fields across the underlying reconciliations.
• Reported field names must now be unique within a 3 way recon view, so similarly named fields can’t be confused.
• The inputs configuration page now links straight to each underlying process.
• The configuration of each underlying reconciliation now shows that it’s part of a 3 way recon view.
• Multi recon views are now fully supported on the Processes list, with a note that makes them easy to identify.

Early access - contact your Duco Client Success manager  to get access or submit your interest directly to our Product team via the Product Portal.

Reporting

Richer data on data prep activity in your reporting tools

Following last month's release of the data_prep_inputs and data_prep_submission_processings views, Duco reporting database now includes three snapshot views completing the Data Prep metadata set:

• data_prep_snapshots, 
• data_prep_snapshot_to_matching_runs 
• data_prep_snapshot_to_cash_submissions

All five views are generally available to every reporting database user, so you can now answer questions about data prep from your own BI tools rather than the Duco UI: which snapshots were created today, whether they ran on schedule or were triggered manually, which ones are still pending or failed, and who created them.

More importantly, the mapping views connect data prep to the rest of your Duco data. By joining snapshots to the existing runs and cash_submissions tables, you can trace any reconciliation result back to the exact snapshot and source files that produced it, confirm that every snapshot actually fed a downstream process, and build end-to-end operational dashboards covering ingestion, preparation and matching in one place. 

This makes audit trails, SLA monitoring and root-cause analysis of data issues significantly simpler.

Cash account labels now available in your reporting tools

Labels you add to cash accounts on the Accounts page of a cash process are now available in the reporting database. This means you can group, filter and report on your cash accounts by label in your own BI tools, alongside the rest of your cash data, without exporting it from Duco first.

Agentic workspace

Notice on important upcoming changes

New Data deleter role in cash processes

From the November release, removing data from a cash process will require a new role called Data deleter. Today, any user or group with the Config admin role can delete data from a cash process. Once this change is live, the Config admin role on its own will no longer allow this. 

Instead, only users and groups with the Data deleter role will be able to:

• delete statements,
• delete submissions,
• reset a process or delete all transactions,
• import accounts with the Overwrite option,
• delete accounts.

This gives you more control over who can remove data from your cash processes, and lets you separate the people who manage configuration from those who need to remove data as part of day-to-day work. Every deletion will also be recorded in the audit log, showing who removed what, when and how.

Before the November release, we recommend reviewing which users and groups currently have the Config admin role in your cash processes and deciding who should be given the new Data deleter role once it’s available. As this role allows data to be removed, we suggest granting it only where it’s needed. 

We will let you know in further release notes when the new role becomes available.

If you have any questions, please get in touch with your Client Success representative.