If Product Features are Built in a Forest… Our New Release Process and What's in 25.4

Last week, we shipped our 25.4 Fall Product Updates—and honestly, we're almost as excited about how we did it than what we shipped. (Though we’re pretty darn excited about that too.)

Don't get me wrong, the features are great (Multiple invoice detection and auto splitting, anyone?) but here's the thing: building fantastic product improvements is only half the job. If our customers don't know about them, can't find the documentation, or don't understand how these updates make their lives easier, did we get a return on all the effort we put into getting them out the door?

That's why we completely reimagined our release process this quarter. We built new processes, took a templatized but flexible and detailed approach, added some tools, and spoiler alert, leaned into AI. 

Why We Adopted a New Approach

For years, we followed the SaaS playbook everyone knows: ship features continuously, deploy to production as soon as they're ready. It's fast. It's agile. 

The problem? Customer enablement matters just as much as customer delivery.

We'd pour months of engineering effort into a feature, release it into the wild, and then watch as user adoption flagged or support tickets spiked from confused users. We wanted to do better. 

So we built a new process with three core principles:

1. Continue shipping features on demand (because delaying the delivery of customer-value doesn’t make a ton of sense)
2. Bundle communications into quarterly releases (so customers get routine, clear, cohesive story about what's new without overwhelming them)
3. Invest in enablement that scales based on the release items (documentation, training materials, public facing materials, etc.)

Think of it like the difference between someone handing you a pile of puzzle pieces versus showing you the picture on the box. Same pieces, dramatically different experience.

How We Scaled Up Activities Without Adding Work

We realized early that our new process would require a significant amount of content: product wiki pages, enablement slides, customer emails, support call scripts, help center articles, internal training materials—the list kept growing exponentially as we added more to the release.

That's when we leaned into AI (specifically Anthropic’s Claude) as a core part of our workflow. The results have been transformative.

Our team now uses AI to take raw Jira tickets and engineering specs and transform them into detailed internal wiki and enablement content, and uses it again to turns those into polished, customer-facing content. We feed it technical details and a page template, and it gives clear and concise wiki articles that require minimal editing. A few bullet points from a feature description and some persona documentation, and it helps distill why that feature matters to somebody processing 500 invoices a month vs 50,000.  

We're not talking about automated content production that publishes without human oversight. We're talking about a process that takes what used to be a dozen hours of digging through six systems, compiling background materials, and several alignment and demo meetings, into 2 hours of refining and perfecting something that's was 80% of the way there when we started.

The efficiency gains are real. We created comprehensive internal product wiki pages, external help center articles, slides, call scripts, customer-facing release notes, in-app pop-up guides and more in a fraction of the time it used to take. Which means we can spend more time on the strategic work: understanding our customers and ensuring every piece of content serves them.

What's in PredictAP’s Fall 25.4 Release

Now let's talk about the good stuff that you really care about. What did we actually release?

25.4 is all about helping you complete work more efficiently with fewer clicks, streamlined approvals, and better access to the information you need. Here are the highlights:

AutoRoute: One Inbox to Rule Them All

If you manage multiple properties across different databases or ERP systems, you know the vendor email problem. Every property needs its own submission address, and getting vendors to send invoices to the right place feels like herding cats.

AutoRoute solves this. Give vendors one central email address. We automatically route invoices to the correct system based on intelligent detection. No more vendor confusion, no more mis-routed invoices, no more manual forwarding. Just one address that works for everything.

Multiple Invoice Auto-Splitting: Stop Playing Document Surgeon

You know that moment when a vendor sends you a single PDF with five invoices in it? And you have to download it, manually split it in Acrobat, upload each piece separately, and pray you didn't miss one? You’re not the only one who hated that.

So we fixed it. PredictAP now not only automatically detects multi-invoice documents, but suggests where to split them, allows you to drag split points and pages around as you see fit, confirm and submit them, all with a few clicks on one screen. The tedious manual work vanishes. The risk of missing invoices disappears. You get clean, separated invoices ready for processing without ever leaving the platform.

Yardi Credit Memo + Adjustment Support: Smarter Document Classification

This one's for our customers who've been processing a lot of credit memos and adjustments and having to fix the amounts manually in their ERP or AP workflow tool. You might be surprised how many clicks it can take to change +$50 to -$50 dollars. But no more. 

PredictAP now automatically identifies whether a document is an invoice, credit memo, or adjustment, and submits it correctly the first time. Less manual rework. Fewer errors. More time back in your day.

Day One Predictions: No More Cold Starts

Here's a problem we didn't fully appreciate until customers told us: when you first onboard or add a new vendor, PredictAP had no history to learn from. Predictions were... not up to the accuracy our customers had gotten used to. 

With Day One Predictions, we now combine what we see on the invoice with information from your vendor database to make intelligent predictions even for brand-new vendors. Better accuracy from day one. Faster onboarding. Fewer invoices flagged for manual review whether you're just getting started or working with new vendors or new properties.

Improved Delivery Failure Messaging: I’m Sorry, I Don’t Speak Computer

Ever get an error message like VENDOR_PO_ASSOCIATION_FAIL and think "well, so what exactly does that mean, and what do I do about it?” These somewhat cryptic error codes aren’t always helpful in getting the problem resolved. 

We've built a translation layer that converts technical jargon into clear, actionable guidance. When something goes wrong, you'll now see messages like "This purchase order is assigned to a different vendor. Please verify the PO number or update the vendor on the invoice.

Email Context in the UI: Stop Losing Important Details

When vendors email invoices with coding instructions, sender details, or important notes in the email body, that context was obvious to users once the invoice entered PredictAP. You'd open a document and wonder "wait, did they tell me something about what account code to use…? Hold on, let me dig through my emails… maybe I’ll have just have to call them…"

Now you can view the complete original email right alongside the invoice. No more lost context. No more guessing. Everything you need to process the invoice accurately, right there on one screen.

The Results (So Far)

It's early days, but the signs are promising. Our engagement data confirms what our CS team is telling us—customers are actually reading the release notes. Support tickets about "how do I use this new feature?" have dropped. And perhaps most importantly, we're seeing increased adoption of new capabilities.

The new process isn't perfect. We're still refining and looking for that goldilocks zone for comprehensive documentation, proactive communication, and overwhelming people with information. But the foundation is solid, and we're committed to iterating and improving.

What's Next

We're planning quarterly releases moving forward, which means you can expect the next major update bundle in early 2026. 

Want to learn more about any of these features? Check out the complete 25.4 release notes in our Help Center, or reach out to your Customer Success rep to schedule a walkthrough. If you're not yet a PredictAP customer and want to see these features in action, schedule a demo here and we'll show you how we're making invoice coding faster, smarter, more accurate, and a whole lot less work.

At the end of the day, building a great software company isn't just about what you ship, it's about what customers use. This new release process is about becoming more customer obsessed. It’s gotta be more about you, than about us!