Sage Transform 2024 - "end the monthly close" raises the stakes - but does AI bring real-time finance within reach?

Jon Reed Profile picture for user jreed February 28, 2024
Sage Transform 2024 is just at the halfway point, but there is plenty of news to drill into - starting with Sage's Copilot. But I think the issues go deeper, into the jugular possibility of eliminating finance admin. Is that even possible? And does AI account for this year's shift in tone?

Steve Hare keynoting at Sage Transform 2024
(Steve Hare at Sage Transform 2024, by Bob Scott)

Enterprise keynote audiences are fairly predictable - sometimes the applause is loud, with that good-to-be-out-of-the-office vibe.

Other times, the applause is of the muted variety - perhaps acknowledging a lukewarm news announcement. But at the Sage Transform 2024 opening keynote yesterday, I heard something entirely different.

It started with Sage CEO Steve Hare's mention of the rapid adoption of Sage's Accounts Payable automation solution, with 1,400 new customers added in 2023. (Sage leaders credit the product's success to their seven year AI maturity path - arguing that most off-the-shelf AP automation technologies do not match fields and documents all that well today, nor do they understand accounting context. Sage now has 24 special purpose ML models for recognizing/processing invoices).

Is the end of the monthly close (and the annual audit) getting closer?

Hare's casual mention of AP Automation elicited a jugular response from a few audience members. Their joyful shouts indicated something I hardly ever see: signs of a genuine change in the administrative finance workload.

Then, a few minutes later, it happened again. In the lead up to Sage Transform, we heard Sage leaders talk about three main areas of innovation: continuous close, continuous assurance (automating compliance), and continuous insights. Given that the history of finance is mostly about poring over rogue/historical Excel documents and gathering batch data for pressurized filing deadlines, talk of continuous close does not lack ambition. But talk of 'acting on real-time data' too often feels like a brochure.

In Hare's keynote, he took it further. He named the nemesis: 

Our ambition is to eliminate the monthly close.

If continuous close means no more monthly close, then continuous assurance means "no more audits." Yep - Hare went there.

Our ambition is to eliminate the annual audit.

Obviously, Hare wasn't undermining the auditing profession or the importance of compliance. Audits will never go away, not should they. It's about blowing past administrative finance bottlenecks that once seemed insurmountable.

After Hare singled out the top areas of finance duress, some audience members lost their minds a bit again -  the prospects of obliterating the monthly close and the stress of annual audit prep pushes a jugular button.

Sage CTO Aaron Harris said when Intacct first talked about the "end of the close" about six years ago, there was a very different - and more subdued - audience reaction (maybe it was quiet disbelief).

Are these new realities right around the corner? Of course not. Transform even has sessions on the comparatively modest ambition of trimming back some of the monthly close time. And for most organizations, "continuous close" will involve a level of real-time finance and HR (payroll) integration that is not close to reality.

The point is, perhaps for the first time, these keynote statements came off not as pie-in-the-sky, but as things that may someday actually happen.

And for finance leaders, it can't happen soon enough. No, they don't want more time at the water cooler. Finance leaders know the business needs a more strategic, collaborative kind of finance - and they need it now. We can all sense it - but we aren't going to get there with same-old same-old, or burning ourselves out on the weekend reporting push. 

Can AI make real-time finance real? Sage announces its Copilot

Can AI help? Given that the biggest day one announcement, Sage's Copilot for SMBs and accountants, won't be formally available for Intacct customers this year, you could wonder why this mindset shift seems to be happening. It's important to note that some of the generative AI features that wind up in Sage Copilot will likely be rolled out to Intacct customers in 2024 - but I think the answer to the crowd's spontaneous response lies elsewhere.

I think Sage Intacct customers have seen enough impactful AI features from Sage, from AP automation to GL outlier detection to employee time capture. Does that mean embedding AI into finance brings those "continuous" finance dreams into a more achievable timeline? Looks like we're going to find out.

Factor in that I am a pretty notorious AI grouch. I hammer the downsides of AI and condemn the frequently-observed phenomenon I call AI overreach. What do I find more persuasive? Customer proof points. Consider:

Sage Intacct's AP automation toolset learns from 60 million vendor relationships, optimizes about 40,000 vendors, and achieves 97.7% accuracy with common vendors. (Part of the 'learning' here comes from a soon-to-be-patented Sage technique they call the digital fingerprint - more on that shortly).

In Accounts Receivable, AI is being used to predict cash receipts at 7, 14 and 30 day intervals. The tool is basing its predictions on several transactional data points, but also on the payment history of the customer. AI is also helping Sage customers speed up bank reconciliation efforts.

If enterprise AI vendors want to outperform the problematic output of the big consumer gen AI bots - and stay away from the PR circus about hallucinations and AI gaffes - then they must infuse their AI results with customer-specific data, without compromising on customer data privacy or quality. Enterprise vendors that have a leg up on their peers on their AI pursuits all have this in common. I don't care how "intelligent" your AI tooling is - enterprise AI success requires a well-thought approach applying customer data to AI, typically using a combination of established LLM output refining techniques like RAG, along with the vendor's own enterprise AI tech innovations.

Our contributor Brian Sommer is not on the ground in Las Vegas this week, but he participated in virtual briefings prior to Transform. Sommer is just as much of an AI "prove it" grouch as I am. As he wrote to me:

Sage is being smart with its AI choices. Specifically, they are using highly-specific (narrow) AI tools to help users more accurately post transactions. In Accounts Payable, Sage is using AI to improve the accuracy of invoices, identify the correct vendor, identify the correct general ledger asset or expense account to book the charge to, etc.  AI appears to be helping their professional services users save billable time by automatically generating much of an employee’s timesheet entries.

Regarding the success of Sage's already-live AI solutions, Hare told Sage Transform attendees:

Sarah at Cambio Communities went from processing about 90 invoices manually every day to over three hundred. Or General Ledger Outlier Detection, which uses AI to capture unusual transactions. Eric at Veracity Research Company told us they caught an overstatement of revenue of about $350,000 before it even hit the books. Or Sage Inbox, which handles communications between your business and your customers and your suppliers, intelligently suggesting responses in natural language. So it's getting more powerful through AI services that deliver real benefits.

Though I had big questions about Sage's acquisition of Intacct, I never doubted Sage could help make Intacct's approach to finance more global in scope. Hare added:

We now have teams not just in the US, but in Canada, UK, France, South Africa, Australia, supporting Intacct customers all around the world. And here's the first big announcement. We weren't going to introduce Intacct into Germany in August. But the great news is, it's already there today, launched in Germany - the second country in Europe following our launch in France.

When Brian Sommer assessed Sage's pre-conference AI news reveal, the role of the digital fingerprint stood out. He wrote:

Sage has applied for a patent on its digital fingerprint. This is a cool idea that utilizes the insights Sage can learn from its commerce network. How it works is this:

  • Sage’s commerce network is where buyers and sellers transact business. If a vendor invoices one buyer, Sage’s AI tool interrogates the invoice and creates a map of the document. It notes the placement of key fields on the (digital or actual) document, address used by the vendor and a number of other values that create a unique way of identifying what a legitimate invoice of this supplier contains.
  • That ‘fingerprint’ is used to validate other invoices entered into the system regardless of who the customer might be.
  • The fingerprint helps reduce losses by flagging potentially fraudulent invoices. Likewise, it can also increase a buyer’s confidence that the invoice is legitimate.

My take

I'm about to press the "publish" button, but there is still plenty of Sage Transform still ahead. A few open questions I'm still pursuing:

On Sage Copilot, we won't get a hands-on customer view for a while. But this much is clear: this is not just a FAQ-trained digital assistant, but a much more pro-active/comprehensive type of companion. As Sommer put it:

Sage’s AI tools, most specifically Copilot, can not only help automatically process transactions but they now can follow up with employees who have yet to respond to requests to approve (or deny) certain transactions.

However, I do have questions about whether I'd want even a copilot as sophisticated as Sage's writing emails for my finance team. I got into that in my interview with Aaron Harris, as well as what his team is doing to combat hallucinations (Harris said today that of the many AI-related patents being filed by Sage's twenty member data science team, the majority of those patents involve reducing hallucinations).

I also had questions about whether Harris' goal to reduce hallucinations is compatible with Sage's use of some of the big external GPT-type language models. Wouldn't a more focused, finance-specific LLM work be more accurate - trained on a finance-specific data set? That's a question for another time, but when you consider yesterday's news that "Sage and AWS will launch the first domain-specific accounting Large Language Model (LLM)," Harris is clearly thinking about these issues.

Sage's vertical initiatives also bear watching. I'd like to see more of their ISV partners building out vertical solutions, but Sage is clearly busy on this front as well, with the January 2024 acquisition of Bridgetown Software being the latest example. Then we have Sage's carbon accounting moves to consider, in the context of Sage's overall sustainability push.

But to me, the biggest question is: if we succeed in reducing the administrivia and compliance chores of finance, how should finance leaders evolve? Being more strategic and collaborative sounds pretty sexy, and it's clearly necessary. But how will that get done? I've heard customers talking through this on the ground; I also took that question up with Hare. Stay tuned.

diginomica contributor Brian Sommer made important contributions to this article. Robert Scott, who took the feature photo, attended the conference on the media/analyst track as well.

A grey colored placeholder image