What's new
When you create a model, you can now choose Weekly instead of Monthly. A weekly model shows one column per week and rolls up into months, quarters and years, so you see the weekly detail and the bigger picture in one place. It's built for the 13-week cash flow forecast, and works just as well for weekly revenue or weekly KPIs.

Weekly models work with the rest of Abacum. You can enter weekly inputs, pick the exact week your forecast starts, feed weekly numbers into your monthly P&L, and report by week with the new Next 13 weeks and Week to date time frames.

How it works
Weeks follow your organization's calendar: the day your week starts and how weeks group into months. A model's timescale is set when you create it, so use Weekly for cash and short-term planning and Monthly for everything else. Your existing models stay monthly. Formulas work as usual, with t-1 meaning last week: Opening Cash = Closing Cash.t-1.

Try it now

1. Check your week calendar in Company settings → Configuration.
2. Create a new model and choose Weekly.
3. Add your variables, or follow the 13-week cash flow tutorial step by step.

Available for all organizations. Full guide in the Academy.

Attach a file or pick an integration, and the AI Agent sets up the dataset for you: columns typed, dimensions mapped, data loaded.

What's new
Ask the AI Agent to create a dataset from an Excel or CSV export, or from one of your connected integrations. It proposes a type for every column, the columns to ignore and how each text column maps to your dimensions, including a new dimension when none fits.

How it works
Nothing is created until you confirm. Correct the proposal in plain language, confirm once, and the Agent creates the dataset and links you to it. Files load every row right away. For integrations, the Agent asks which months to sync once the dataset exists.

Try it now

1. Open the AI Agent and attach an .xlsx or .csv export up to 4.5 MB.
2. Type "Create a dataset from this file".
3. Review the proposal and confirm.

Available to Admins and Super Admins. Full guide in the Academy.

What's new
Budgeting season just got lighter for Budget Owners. Budgeting starts with decisions, not cells, so now you can tell the Agent what's changing in your plan, and it updates your Variables in the Input Table for you.

How it works
Abacum Intelligence reads your request alongside the data already in your Input Table, including prior inputs, Actuals, and historical patterns, and works out the impact for you. Nothing changes until you approve the proposal. Once you apply it, the updates show up in the table like any other input, and your request is logged with them.

Try it now

1. Open your Input Table, invoke Abacum Intelligence
2. Describe your plan the way you'd explain it to your Finance partner: "Same team as last year, but the Atlassian contract goes up 15%. We're dropping ChatGPT. Adding Claude at €50k starting Q2."
3. Abacum turns that into a proposal: which values change, which new rows are added, and what amounts go where, based on your existing data.
   
   

Available to all users. Link to Academy available soon.

Building a model used to mean knowing the syntax, or knowing the one person who does. Not anymore.

Open the AI Agent and tell it what you need:

"Create a P&L model from our GL, by Department, forecasting into FY26 Forecast"

The Agent scans your datasets and, a few minutes later, shows you a proposal: every variable with its actuals and forecast formulas. Review it, ask for changes in plain language, and click Approve. The model is built right away and opens in the model builder, planning-ready. Built on your data, calculated by the Abacum Engine, never saved without your approval.

What you can do today

• Build any model from a prompt. P&L, Balance Sheet, Cash Flow, Headcount & Payroll, OPEX, Working Capital, ARR Bridge, Cohorts, Sales Capacity, or whatever you describe. Actuals and forecast included.
• Grow the models you already have. Ask for a new variable or a new section in an existing model. The Agent proposes it; you approve.
• Start from a spreadsheet. Attach a small Excel file with the structure of the model you run today, and the Agent proposes the same lines on your Abacum data.
• Understand any model. Open a model and ask "walk me through this" or "how is Gross Margin calculated?".

Try it now

1. Open the AI Agent.
2. Type "Create a P&L model" or describe the model you have been putting off.
3. Review the proposal and click Approve.

Available to Admins and Super Admins. Full guide in the Academy link. Coming next: editing existing variables and building a model from a full Excel model, formulas included.

The Agent knows your numbers. Now it can know the story behind them too.

In the Library's Company Context section, admins can upload documents like board decks, budget narratives, and policies. When a question calls for it, the Agent finds the relevant document and grounds its answer in it, citing the file by name. Ask why a line came in below forecast, and instead of just describing the movement, the Agent can tell you the decision behind it.

• Upload once, benefit everywhere: every user's conversations use the documents; only admins manage them.

• Built for real documents: a 200-page board deck is indexed and searched, never pasted into the chat.

• Pair it with the company description field: a short description of your business shapes every answer, so a healthy margin for your industry is read as exactly that.

Supported formats: PDF, TXT, and MD with readable text, up to 10MB per file.

Learn how to set it up in the Academy.

Every dimension tends to accumulate items that you've stopped using, like old vendors, renamed accounts, or other leftovers from past mappings. Over time, these can pile up, crowding the lists your team selects from and slowing down your entire workspace.

Now you can clear them out in seconds. Abacum spots the items that aren't used anywhere and removes them in one pass. This means your vendor list shifts from a long, cluttered column to a concise list of the vendors you actively work with.

Faster everywhere, because lean dimensions load and calculate quicker across every table, report, and input. And safe, because it offers only genuinely unused items, and removes nothing until you say so.

Learn more about how to Clean up dimensions

Adding a table or chart to a space normally means building it by hand: the variables, the dimensions, the timeframe, the comparison. Now you can skip that and just describe what you want. Abacum Intelligence builds the block inline on your real data, and you drag it straight into any space.

✨ What's new

• Generate tables, charts, KPI cards, and input tables from a prompt, or embed an existing dataset version. 
• Open a space and the AI Agent, describe the block you want (e.g., "revenue by location for Jan to Feb 2025 vs prior year"), then review it inline and drag it into the space or click insert. Adjust it from there as you would any block.
• Point it at the right data: type @ in your prompt to reference a specific variable, version, or dimension. The more detail you give (the breakdown, the timeframe, the comparison), the closer the result.

Need a whole space rather than a single block? That's AI Spaces: describe the report you want and Abacum Intelligence builds the entire space in one go. Block generation is the same capability that powers AI Spaces and is available on its own whenever you just need one more thing.

Learn more in the Academy.

The Agent can now read the context of the documents where your decisions actually live. Attach a board deck, meeting transcript, offline spreadsheet, or policy doc to any conversation, and the Agent combines it with your Abacum data to answer as your real business context partner.

✨ What's new

A lot of the "why" behind your numbers sits outside the platform: the re-forecast discussed in a board deck, the hiring pause approved in a meeting, the covenant defined in an operating plan. Now you can bring that business context straight into the conversation.

Click the new attach button (+) in the Agent input, add one or more files with your question, and the Agent uses their content alongside your models and datasets, not instead of them. Each file shows as a chip on your message and stays part of the conversation for follow-ups.

📈 Impact

• Cross-check documents against live data, attach your FY2026 board operating plan, and ask "What's the EBITDA burn covenant, how much have we burned to date, and how much headroom is left?" The Agent reads the covenant from the deck and pulls actual EBITDA from your model.
• Validate reports at a glance, attach a P&L screenshot, and ask the Agent to compare it against the P&L in your Space.
• Ground answers in your source documents, meeting transcripts, policy docs, and offline spreadsheets that become part of the Agent's context.

⚙️ How it works

Click + in the Agent input (side panel or expanded view), pick your files, and ask. Supported types: pdf, csv, xlsx, html, txt, md, and images (png, jpeg, gif, webp). Files are validated at upload, so if a file is too large or unsupported, you get a clear message upfront. You can always upload just the relevant tab or section. Attachments are scoped to their conversation only and never influence other conversations or users.

Some questions are faster to say than to type. The Agent's chat input now accepts voice.

✨ What's new

Click the mic in the chat input, ask your question out loud, and the transcript lands in the input as editable text. Review it, fix a word if you need to, and send. Nothing is sent until you decide.

Dictation understands finance language. Terms like EBITDA, CapEx, ARR, or rolling forecast are transcribed correctly, and you can speak in English, Spanish, French, or German. The language is detected automatically.

⚙️ How it works

While you speak, a live waveform confirms the Agent is listening. Confirm to insert the transcript, or cancel to keep what you had typed before. Your browser will ask for microphone permission the first time you use it. Audio is processed for transcription only and is never stored by Abacum.