Velixo Intelligence, one finance month at a time

Every finance team knows this moment. A formula breaks on a Tuesday afternoon. A number does not tie, and the reason is somewhere in the ERP, three screens away from the workbook you are working in. 

Until now you had three options, and none of them were good. Guess the problem’s source. Export everything and rebuild it by hand. Or send a message to someone and wait, which on a Tuesday in close week means waiting until Wednesday. 

The fourth option is Velixo Intelligence. Our new module puts an AI assistant inside the workbook you use day in and day out, and it reads your live general ledger to answer – not a copy of your data from last night. The live ledger, directly in your ERP. 

It is available to every Velixo NX user, it includes a free usage allowance, and if you have been updated to 2026.7, it is already sitting in your ribbon. Here is what it does across a normal finance month. 

One prompt box, five specialists, three ways in 

You never pick a tool. You describe the problem, and the phrasing of your question decides which specialist answers it. 

There are five. The support agent answers anything about Velixo, Excel, or your ERP – technical or functional – without a support ticket. The guide agent makes changes to the workbook itself, once you approve them. The analyst agent reads your report and tells you what moved and why. The query builder agent pulls data you have not pulled before. The report designer agent turns a described report into a built one. 

One assistant, five jobs. Your personal Excel power user, ERP analyst, and Velixo expert in the same panel. 

The three ways you can work with Velixo Intelligence. 

Click the Velixo Intelligence button in the ribbon, and a side panel opens. Right click an area or a range in your reports to analyze or troubleshoot data, or use the =VX.ASK() function to create dynamic narratives/notes inside the cell. More on that one later, because it changes a recurring task rather than speeding up a one-time one. 

Stage one: standing up the report that does not exist yet 

The request arrives on day two of the month. Something nobody has built before. A new dimension in the P&L. A project profitability view the PM now wants weekly. The report the auditor asked for last year and will ask for again in October. 

Start from a sentence rather than a blank sheet. 

Describe the data you need in the words you would use out loud. The query builder agent picks the right API, builds the query, and hands you something you can edit. 

The middle step is the one that saves the hours. Knowing which ERP object holds the number you want takes years to learn, and every ERP names it differently. You know you need unbilled project costs by manager for the quarter, filtered to one branch. Knowing which API exposes that, and how it joins to everything else, is the part that used to send you to the documentation or to your VAR at four in the afternoon. The right data, pulled the right way, without the archaeology. 

And what lands in your sheet is a query you can open and change, rather than a static export you would have to rebuild next month. 

If you prefer building queries by hand, 2026.7 also ships usability improvements to the Query Builder itself. The agent is an addition, not a replacement. 

Then the step up from a query to a finished report. Describe the report you need, and watch it build. 

The report designer agent assembles four things: the functions, the filters, the formatting, and the logic. The fourth one is what separates this from a chatbot that hands you a paragraph. What appears in your workbook is a working Velixo report, with live functions in the cells, and it refreshes exactly like the reports you built by hand over the last five years. Open any formula. Change it. It is a real report rather than a one-off answer, and it is yours from the moment it appears. 

Then the part that matters more than the first build. 

Keep asking, and it keeps refining. Nobody describes a report correctly on the first attempt. You get a version, you look at it, and you realize the comparison should be prior year rather than budget, the subtotals belong at the branch level, and the two service lines should be split out. So you say that, in those words, and the report converges. Three prompts in, you have the thing you were looking to do when you started typing. 

Stage two: the refresh that breaks something 

This stage is not on the calendar. It happens on whatever day you were planning to do something else. 

You hit refresh and a column of #VALUE! appears. Or the report is correct and ugly, and it is going to a board member in an hour. Or the chart you need does not exist yet. 

Finding the problem was never really the hard part in that case, but it can be when nested errors hide in somewhere in the workbook. Finishing the fix was, because fixing it meant knowing the location of the cell, the exact function syntax, or where the nested reference broke. 

Tell Velixo Intelligence what you need in the terms you would use with a colleague, and it will route it to the Guide Agent on its own. It reasons over your live workbook rather than a generic template, then proposes the changes: the formula correction, the formatting, the chart. You look at what it proposes. You approve. Then it writes. 

That order matters, and it is the reason a controller can run this on a live workbook two hours before a board meeting. Nothing changes in your file until you say so. 

Real prompts that you should try next time you open your Excel Workbook, powered by Velixo: 

• “Refresh this report then freeze it.” 

• “Add a bar chart of the top ten variances.” 

• “Drill down and compare transactions with my P&L expenses.” 

Less fixing, more finishing. 

The other half of this stage is the workbook you inherited. Every finance team has one: the report built by someone who left in 2022, forty tabs deep, with a formula in row 400 that nobody has been willing to touch since. 

New to Velixo, or just new to this report? Ask anything. Functions, errors, how to do a specific thing, why this cell returns what it returns. The support agent is your always on Velixo expert, sitting next to the sheet instead of at the other end of a ticket queue. No context switch, no waiting for someone in another time zone to read your message. 

Stage three: explaining what moved 

Close week. The numbers are in and now somebody must explain them, usually before lunch. 

The analysis used to start after the export. You pulled the trial balance into a fresh sheet, built a pivot, sorted by variance, and forty minutes later you were ready to begin thinking. The thinking was the job. The forty minutes was tax. 

Ask Velixo Intelligence what moved, and it will automatically pick the analyst agent. It will tell you what changed this month. What is the key information here. Summarize this sheet. Do not export it, do not pivot it, just ask. 

What comes back is variances surfaced, anomalies flagged, and recommendations attached, all traceable to the rows they came from. The agent is reading the same live general ledger your workbook reads, which brings up the most important sentence in this article. 

The number the AI explains is the number on your GL. 

No cube. No data warehouse. No overnight sync. Only live ERP data that Velixo Intelligence knows and understands. Plenty of AI tools will summarize a snapshot of your financials and sound confident doing it. The difference here is that when the analyst agent tells you which cost center drove the variance, it read the ledger to find out, and you can drill from its answer to the underlying transactions the same way you drill from any other Velixo cell. 

The practical effect is a change in timing. You walk into the review with the variance already explained rather than being asked about it and promising to look into it. The insight arrives before anyone had to ask. 

Stage four: shipping it 

The last two days of the month. The board pack, the audit request, the distribution list. 

The number is easy. Explaining it every month is the work. 

You have produced a correct number by ten in the morning and spent until four writing the paragraph underneath it. Then next month you do it again, with the same structure, the same three drivers, and different figures. 

=VX.ASK() is a function you put in the cell where the commentary belongs, pointed at the range that holds the numbers. It writes the explanation there, next to the number it is explaining. In the cell, not in a sidebar. Commentary in a sidebar has to be copied somewhere before it is worth anything to you. Commentary in a cell is already in the deliverable. 

Why we think that function is so great? 

It refreshes. Next month, you refresh the workbook and the commentary refreshes with the numbers. The monthly rewrite stops being a task. 

It travels. Send the file to the board or to your auditor and the commentary goes with it, because it is a cell in an Excel file rather than a chat log someone needs to be granted access to. 

It cites the source rows. A generated paragraph that nobody can trace is a liability in a board pack. A generated paragraph with its source rows attached is a working paper. That is the whole difference between AI commentary you would defend in a review and AI commentary you would delete before sending. 

We all know that handwritten variance commentary is one of the most reliable sources of error in the monthly pack, especially when it’s an artifact from the last board deck. The number changed on refresh and the sentence below it did not, and nobody caught it until someone read it aloud in the meeting. 

Why Velixo Intelligence is more accurate than the AI in your browser 

You could paste a range into a general purpose chatbot today. Some of your team already has. Here is what is different about asking inside Velixo. 

It was taught your ERP. A general model has read the internet, which means it will confidently invent a Velixo function signature or an ERP field name that does not exist, and it will do so in a tone that reads like certainty. Velixo Intelligence is grounded in Velixo, ERP, and Excel knowledge: the real object model, the real function reference, the actual product behavior. Grounding is why it is less prone to hallucination on exactly the questions you care about. 

It reads live data. The answer comes from your general ledger at the moment you ask, so there is no gap between what the AI describes and what your report shows. 

Every answer can be traced. Findings point back to the rows they came from, and =VX.ASK() cites its source rows alongside the commentary it writes. Traceability is what turns AI output into something you can sign. It also answers the question this audience asks first, which is not “will it ever be wrong” but “how would I know.” 

Nothing writes until you approve it. Agents propose. You decide. Then the workbook changes. 

What you should know, though, is that Velixo Intelligence is AI. It is probabilistic, and it can make mistakes. Treat its output the way you would treat good work from a capable new analyst: read it before you use it, and use the approval step as the place where you do that. The traceability above is what makes checking fast rather than theoretical. 

You already have Velixo Intelligence 

That is, if you are a Velixo client. There is nothing extra to install beyond the update you already received. 

Velixo Intelligence is available to every Velixo NX user from version 2026.7, and a generous usage allowance that resets weekly if included in your subscription. When Velixo Intelligence reaches general availability later this year, an paid options will also be available for teams that need more.. 

If your organization would rather not enable it, you can opt out, and the documentation explains how. We would rather tell you that here than have you find it in a release note. 

Questions we are getting 

Do I need to change how my existing reports are built? 

No. Velixo Intelligence reads the workbooks you already have, including reports built years ago.  

Will it change my workbook without asking? 

No. Agents propose changes only asked, you approve them, and then they are written. 

What happens when it gets something wrong? 

It is AI and it is probabilistic, so it will occasionally be wrong. Read its output as a draft, and follow the source rows when a number matters. 

Does it work on reports I did not build? 

That is one of the cases it is best at. Ask the support agent about any workbook you inherited. 

Does Velixo store prompts and data? Is it used to train AI? 

Prompts and responses pass through the Velixo AI Gateway, which keeps a transient diagnostic log for a limited rolling period, after which it is purged. Access is restricted to a limited number of Velixo engineers. Model providers may retain data for a limited abuse-monitoring period, which cannot be opted out of. Neither Velixo nor the model providers use customer data to train or fine-tune any LLM. Full details: AI Technology Terms. 

How do I know about my usage consumption? 

You can check your organization’s usage and how much is left to consume until it resets the following Monday, You’ll find the Usage hover-card in the Velixo Intelligence side panel in the top right-end corner. 

Try it on your own reports 

That Tuesday afternoon formula is still going to break. The difference is that the answer is now one panel away, reading the same ledger your report reads. 

Don’t have Velixo and want to know what having an Intelligent ERP layer can do for you?  Book a demo to how Velixo Intelligence can work for you.