XBert's Standard XBerts cover the everyday checks across your clients' files - but if there's something specific you want to keep an eye on beyond those, a Custom XBert lets you set one up yourself.
You describe what you want to catch in plain English, ask XI (XBert Intelligence) to build it, and XI drafts a plan, builds the check, tests it against real data and saves it for you!
1. Create a Custom XBert
Go to Automation > XBert Automation and click Create a Custom XBert.
In plain English, describe what you want XBert to watch for and XBert works out where to look. Then click Ask XI.
XI may suggest a clearer version of your prompt. Choose Use improved prompt or Refine - suggest another, then send it.
XI shows you a Plan of how it will build the alert, with Re-evaluate, Accept and Reject options. If you're happy with it, click Accept Plan.
XI then builds, tests and saves your Custom XBert and gives you an overview of what it created. Click the Open your Custom XBert link.
Your new Custom XBert starts as Inactive — XBert never makes it Active on its own. You set it to Active yourself on the Configure tab (next step).
2. Open your Custom XBert → you land on Configure
Click Open your Custom XBert (or open it from the XBert Automation list). It opens on the Configure tab.
Here you set its title, which clients it runs for, subtasks, process, status, role, estimated time, due date, priority, auto-complete, risk level, risk type and business function.
At the bottom is the Active / Inactive toggle. Set it to Active to make it go live.
The other tabs — Approach, Test and Performance — are optional. Open them whenever you like; you don't have to go through them in order.
3. Approach (optional)
The Approach tab shows, in plain English, how XI decided to tackle your request. There's nothing to do here — it's just there if you want to check its reasoning.
4. Test (a dry run before you set the alert to Active)
On the Test tab, pick a client and click Test.
XBert runs your Custom XBert against that client's real data and shows what it would flag — e.g. "I have found 50 XBerts for this Client" — with a short summary for each. (For some areas you can also filter by Tracking Categories.)
Testing is a preview only — it doesn't create any alerts for anyone. It's how you check it's catching the right things, and not too many.
If it finds nothing, you'll see "no results for this Client… try another."
From here you can click Tune Custom XBert (if it's catching the wrong things) or Configure (if it looks right).
5. Performance (once it's Active)
The Performance tab is where you see how well your Custom XBert is doing once it's Active and raising alerts.
It scores accuracy from your team's feedback on the alerts: Accuracy % is how many reviewed alerts were marked as real; Awaiting feedback / Not enough feedback yet means not enough alerts have been reviewed; Confidence (Very early → Building → Good) shows how much to trust the score so far.
It also shows Activity, Version history and Generation history — a log of when it was created, refined and re-tuned.
Tools within the Test & Performance tabs
Tune Custom XBert
You'll find Tune Custom XBert on the Test tab and the Performance tab.
Click it and, in plain English, tell XBert what you want it to do differently — e.g. "Fire less often on amounts under $500", "catch near-duplicate vendor names", or "change the threshold from $5k to $10k."
XBert rewrites your Custom XBert from your feedback and switches to the improved version automatically (it won't downgrade you if the new version scores worse).
Use it when there are too many alerts, it's missing things, or a threshold is incorrect. It changes what it detects — not its title or wording, and not which data it watches.
Auto-tuning & Pause
On the Performance tab, XBert can improve your Custom XBert on its own over time, based on your team's feedback.
A banner shows its state: Working well, Not enough feedback yet, Needs attention, or Auto-tuning paused.
XBert usually prepares a new version for you to review over time if needed (you Approve or Reject it) rather than changing it silently.
Pause auto-tuning stops XBert changing your Custom XBert on its own — use it when you've got it exactly how you want it. Resume lets it start improving again.
Run Custom XBerts (run it now)
Run Custom XBerts is on the Performance tab.
Once your Custom XBert is Active, it checks your clients automatically on a regular basis. Run Custom XBerts is for when you don't want to wait - it checks your selected clients right now and creates real alerts for anything it finds.
Click it, choose your clients, then Run one at a time or Process All to do every selected client at once.
Use it when you want results straight away - e.g. you've just made it Active, or just tuned it - instead of waiting for the next automatic check.
FAQs
How often does my Custom XBert run once it's active?
It runs automatically each time XBert syncs that client's ledger.
You don't need to do anything to keep it running.
I just switched it on — why aren't there any alerts yet?
Switching a Custom XBert to Active doesn't run it straight away — it produces alerts at the next sync. If you want results immediately, use Run Custom XBerts on the Performance tab to run it now against your chosen clients.
When does XBert improve the rule on its own, and what does "Pause auto-tuning" do?
Every Custom XBert improves itself automatically over time, unless you pause it. XBert reviews performance overnight and, once it has enough feedback, may prepare an improved version. Small, low-risk tweaks switch on automatically; bigger changes that alter the rule's logic are parked in an Approve/Reject card for you to decide.
Pause auto-tuning (on the Performance tab) freezes the rule so XBert leaves it exactly as-is until you Resume.
For self-tuning Custom XBerts, XBert reviews performance overnight and, once it has enough feedback, may prepare an improved version. Small, low-risk tweaks switch on automatically; bigger changes that alter the rule's logic are parked in an Approve/Reject card for you to decide — XBert won't make a big change silently. Pause auto-tuning freezes the rule so XBert leaves it exactly as-is until you Resume.
When can I trust the accuracy score on the Performance tab?
Once at least 30 of its alerts have been reviewed (marked Good catch or Not a problem). Below that it's still learning — under 10 reviews it shows "Very early", 10-30 "Building confidence", and above 30 "Good confidence". Accuracy itself is simply the share of reviewed alerts your team marked Good catch.
Does testing a Custom XBert create real alerts?
No — Test is a safe, read-only preview. It shows how many alerts the rule would raise for the client you pick, but it doesn't create any actual XBerts, notifications or tasks. Real alerts are only created once the Custom XBert is saved and switched Active (or when you use Run Custom XBerts).