Policy-Driven Accruals AI: 80% Faster Finance Closings

Deep Dive into Hyperbots Accruals Copilot

Jace Kitchen, AE at Hyperbots:
Um. Today's conversation is around a deep dive on functional capabilities of Hyperbots Accruals Copilot and first question for you. So finance teams still chase spreadsheets at month end. And how does Accruals Copilot change that for you?

Dave Sackett, VP Finance at Persimmon Technology:
Hey. Thanks, Chase. Uh, instead of emailing, uh, from plant managers on goods received, not invoice counts. Uh, Copilot queries SAP, NetSuite, Epicor visibility. You know, whichever ERP system you run, to compile an invoice PO or service entry and then books the accrual journal automatically. Closing can take hours and not days, and the variance to actual can actually be under five percent.

Configuring Accrual Policies

Jace Kitchen:
And what kinds of accrual policies can be configured?

Dave Sackett:
Uh, goods, services, recurring expenses, or manual kind of manager estimates accruals, uh, each with its own materiality rules and GL buckets. Uh, most tools provide a single rule set. Hyperbots lets it set hundreds of policy combinations without any coding.

Cutoff Dates

Jace Kitchen:
And on the cut off dates—is that daily? Weekly? Monthly?

Dave Sackett:
Uh, take your pick. Some entities run daily cut offs based on project margins and swing fast. Uh, corporate does a classic month end. Hyperbots can remember each calendar and each holiday table per entity.

Automated Reversals

Jace Kitchen:
Awesome. And reversals are error prone. And how are those automated for you?

Dave Sackett:
Um, you get to decide, uh, blanket day one reversal or real-time reversal when matching invoices post. I think that's an interesting one where you can actually, as soon as the part comes in, it switches it, uh, which is pretty amazing. We use the same journal ID so auditors can instantly find the link.

Discovery of Goods Received, Not Invoiced

Jace Kitchen:
Awesome. Just very customizable. Flexible for your needs. Um, and how does the Copilot discover GRNI—goods received, not invoiced data?

Dave Sackett:
Yeah. It fetches every receipt line with no invoice reference. Uh, matches the quantity, price, tax and accrues the liability. Uh, a competitor's bot, for example, that Hyperbots tested, only used PO headers, uh, which is a huge variance when it comes to partial receipts.

Handling Services Received, Not Invoiced

Jace Kitchen:
And services can be fuzzy. What about service received, not invoiced?

Dave Sackett:
Uh, we pull approved service sheets from the ERP time entry module and accrue the cost. Even if the freelancer bills two weeks later, that prevents most of the late landing expense as a surprise.

Recurring Expenses

Jace Kitchen:
And recurring. No PO costs, rent, SaaS fees. How are those handled?

Dave Sackett:
A cadence engine looks at prior periods and contracts and predicts the expenses and books the recurring accrual. There are competing tools that expect you to load a template in CSV every month, which is not needed in Hyperbots.

Invoices in Approval Queue

Jace Kitchen:
And you had mentioned invoices still in the approval queue. Could you explain that in a little bit more detail?

Dave Sackett:
Sure. Um, at cutoff, the Copilot asks the Invoice Processing Copilot which invoices are pending? Uh, those are moved to the exclude list, shrinking duplicate accrual risk. Manual processes rarely reconcile the two lists.

Three-Way Matching

Jace Kitchen:
And talk me through the three-way matching that your team relies on.

Dave Sackett:
So the AI engine matches the invoice to the PO price and quantity, and then to each receipt line across one hundred fields with numeric reasoning and descriptions. The invoice is processed into the ERP once the receipts are confirmed against the proper PO and line.

Handling Exceptions

Jace Kitchen:
And what happens when something doesn't line up?

Dave Sackett:
Exception workflows. Firing teams. Uh, for example, line thirty, quantity variance five percent, approve or edit might be the question you get. You know, AP clicks approve, journal posts. You know, they're a competing RPA solution that drops them into a manual tab in hopes that you notice.

Self-Service Portal for Manual Accruals

Jace Kitchen:
Sure. And is there a self-service portal for business users to enter manual accruals?

Dave Sackett:
Yes. Uh, any authorized employee can log in a one-off estimate, attach a note, uh, and the Copilot routes it for approval, then posts it. No more emailing accounting with sticky note numbers.

GL Recommender Accuracy

Jace Kitchen:
Gotta love that. And how does the GL recommender decide the account for each accrual?

Dave Sackett:
It studies historical journals plus any human overrides or uh, training. With two closes it can hit ninety-nine percent coding accuracy. Some competitors require a table to be updated every quarter.

Real-Time Journal Posting

Jace Kitchen:
In the actual postings, is that still manual journal entry uploads?

Dave Sackett:
So all journals are posted via the ERP API. Then reads back to confirm that debits equal credits. No CSV imports, no night batch suspense clearing. It's done all in real time.

Notifications

Jace Kitchen:
Sure. And how are those notifications delivered to the specific employees?

Dave Sackett:
Uh, you've got options. You could do real-time Slack and Teams alerts. Book totals by business unit, exceptions needing approval, and reversal confirmations. Competitors to Hyperbots will send a fifty-row Excel sheet after closing.

Audit Trail

Jace Kitchen:
And what about an audit trail?

Dave Sackett:
So every policy override and posting is timestamped and hash-chained in Hyperbots. Auditors can pull the entire quarter's accrual lifecycle in under ten minutes.

Global Process Across Multiple ERPs

Jace Kitchen:
For our instance, can we run a single global process even with multiple ERPs?

Dave Sackett:
Yes. Uh, one rule can drive SAP, NetSuite, and Sage instances. Hyperbots can standardize the data before AI thinks so that US GAAP and IFRS plants still follow the same approval workflow.

Distinct Rules Per Entity

Jace Kitchen:
Or conversely, can entities keep distinct rules?

Dave Sackett:
Uh, yeah. Flip a toggle and Germany keeps a strict GDPR-based policy. The US permits PO-based accruals. Both configurations can live side by side.

Single Sign-On

Jace Kitchen:
And how does single sign-on work when users need data from, say, three ERPs?

Dave Sackett:
So one Azure AD login, Hyperbots routes credentials to the right ERP connector invisibly. AP never needs to log out to switch different ledgers.

Real Deployment Example

Jace Kitchen:
And could you give me a real deployment example from your case?

Dave Sackett:
Sure. Yeah. Like, for example, two entities, two ERPs, SAP and NetSuite. Hyperbots installs two connectors, maps fields once, and can book five thousand accrual lines in the first close. Variance to actual was under four percent, which saved the accounting team their weekends as a result.

How Hyperbots Outperforms Typical Accrual Tools

Jace Kitchen:
Gotta love it. And how does Hyperbots beat typical accrual tools that you've seen?

Dave Sackett:
So others give you a template and a bot that posts. Hyperbots gives you a policy-driven AI, ninety-nine point eight percent field accuracy, real-time posting, eighty percent productivity gains, live audit trails, multi-ERP harmonization. Uh, while legacy software still asks controllers to please fill in a spreadsheet.