Governed AI
I've spent a good part of my career working with FP&A platforms, financial reporting, analytics and ETL, going all the way back to my time with Cognos, prior and after the acquisition by IBM. Over that time, I've seen some incredibly sophisticated planning and reporting environments. Organizations have invested years building cubes, business rules, hierarchies, and reporting processes that reflect how their businesses operate.
And now we're all trying to figure out what AI means for those investments.
There's a lot of excitement around connecting AI tools like Microsoft Copilot and Claude directly to enterprise data. With technologies like Model Context Protocol (MCP), that's becoming easier to do, and I think there are some interesting opportunities for the Planning Analytics community.
But something has been bothering me about the way we're approaching this.
Just because we can connect AI to a TM1 cube, does that mean we should expect it to understand the business?
Anyone who's worked with TM1 knows that getting the right number isn't always as straightforward as querying a cube.
I can remember plenty of situations where someone would ask for a number, and the first response from Finance would be another question.
Which version? Which hierarchy? Are we including allocations? Do you want the reported number or the adjusted number?
Think about how much knowledge exists within an experienced FP&A or TM1 development team. They know which cubes to use, which versions are relevant, and how the numbers eventually make their way into management reporting.
Some of that knowledge is built into Planning Analytics. A lot of it isn't.
Now imagine giving an AI assistant access to those same cubes and asking it why operating expenses increased last quarter.
It might retrieve the data, perform the calculations, and give you a perfectly reasonable explanation. But would it know that Finance made adjustments before publishing the management report? Or that the forecast version it used was superseded two months ago?
Maybe. But I wouldn't want to assume that.
Here's where I think we're missing an opportunity.
For years, we've been helping customers automate financial reporting. Every month, Finance teams pull information from Planning Analytics, organize it into Excel and PowerPoint, and distribute it to the people who need it.
The end result is often something Finance has spent considerable time preparing, reviewing, and validating.
A management reporting package isn't just a collection of numbers. It represents decisions about which information matters, how it should be presented, and what the business should take away from it.
So why are we so focused on having AI go back to the raw data and figure all of that out again?
If the CFO asks what happened to gross margin last month, and Finance has already prepared a report explaining exactly that, shouldn't AI be able to use it?
I'm not suggesting we stop connecting AI directly to Planning Analytics. I think there's tremendous potential for AI to help users explore data, run analyses, and interact with planning models.
But there's a difference between asking AI to analyze data and asking it to explain the company's official financial results.
Those aren't necessarily the same thing.
One of the more interesting developments has been MCP, which provides a way for AI applications to connect to external systems and information.
I can see a future where someone asks Copilot a question about financial performance, and instead of recreating the analysis from the underlying cubes, it retrieves information from the management report Finance already published.
And if the question requires additional analysis, perhaps that's when it makes sense to go directly to Planning Analytics.
Of course, there are challenges. Security, version management, and keeping information current are all important considerations. But these aren't entirely new problems. We've been dealing with many of them in financial reporting for years.
I've always found it interesting that reporting sometimes gets treated as the less exciting part of a Planning Analytics implementation. Everyone wants to talk about the models, calculations, and planning processes.
But reporting is where much of that work becomes useful to the rest of the organization. It's where Finance takes the complexity behind the scenes and turns it into something management can understand and act on.
I think AI may actually increase the value of that work.
If we can make the information Finance already trusts available to AI, we don't necessarily need to teach an AI assistant everything about every cube, dimension, rule, and business process just to answer a question that's already been addressed in a management report.
That's a pretty compelling idea to me, and something we're spending a lot of time thinking about at Reportworq.
I don't think anyone has all the answers yet. We're still figuring out what AI will look like in the Planning Analytics world.
But I do think we need to be careful not to overlook the investments organizations have already made. We've spent decades building sophisticated financial models and reporting processes, and there's a tremendous amount of business knowledge wrapped up in those systems.
It seems to me that the opportunity isn't just to give AI access to more data. It's to make better use of the information we already know we can trust.
Related: what Finance already approved is the shortest leash for AI, and what governed AI means for a Finance team.