Introducing AI Assistant: A First-Pass Review on Every Entry

This is part of a series of posts on the new capabilities FloQast announced at TakeControl 2026. For the full picture — including new Transform building capabilities, Detect, Operational Audits, and the COSO AI governance module — check out the complete TakeControl 2026 announcement here.

‍

If you've spent a close staring at a queue of 400 journal entries, wondering how you're supposed to meaningfully review all of them by Friday, you already understand the problem.

Journal Entry review is one of the highest-risk activities in the close. It's also one of the least scalable. Volume grows. Headcount doesn't. And somewhere in that arithmetic, "review" quietly becomes "triage." An entry from a preparer you trust, in a GL account you know deeply, gets a quick scan. An entry from a new preparer, in a less familiar account, at 11pm on the last day of close — it gets the same quick scan, because there's no time for anything else.

That's not a process failure. There's just not enough time to give every entry the same look. And there hasn't been for a while.

What We Announced at TakeControl 2026

At TakeControl this year, we introduced journal entry review by the AI Assistant, built into FloQast's Journal Entry Management. A preparer can run an AI review before submitting an entry for approval, and reviewers can request one during approval. Either way, the Assistant runs structural checks, scores the entry's audit risk, explains what it found, and recommends a fix.

Your team always makes the final call. The Assistant never posts anything, corrects anything, or makes any call on anyone's behalf. It does the first-pass structural work up front, so a preparer can fix what it flags before submitting and a reviewer approves with context already in hand.

What It Actually Checks

The checks the Assistant runs are the ones a thorough reviewer would run if they had unlimited time: duplicate postings, empty subledger fields, missing reversal dates, and GL account and dimension combinations that deviate from historical usage.

That last one deserves a moment. The Assistant isn't working from a generic rulebook. It measures each entry against your posting history: roughly 180 days of trailing activity on the account, and the dimension pairings your team has previously approved. When it flags an account-dimension combination, it's because your team hasn't used that pairing before, not because an external standard flags it as suspicious. The difference matters. Context does too.

The Audit Risk Score

Alongside the structural checks, the Assistant calculates an audit risk score for each entry it reviews. The score draws on signals auditors look for: after-hours entry creation, use of suspense or intercompany accounts, amounts just below thresholds, and manual preparation. Scores fall into three bands: Low (0–29), Moderate (30–59), and Elevated (60–100).

This doesn't tell your team what to decide. It tells them what an auditor is likely to notice, so they can look at those entries with their eyes open before anything posts.

What Your Team Gets

After the Assistant completes a review, your team sees a summary of flags, the reasoning behind each, and a recommended fix where relevant. The preparer can address issues before submitting; the approver sees what was flagged and what was done about it. All of that activity is captured in the chat for auditability.

A human reviewer brings things the Assistant doesn't have: knowledge of the business reason behind an entry, context about a particular preparer, judgment about whether an unusual combination reflects an error or a legitimate exception. The Assistant brings consistency. It runs the same checks on every entry it's asked to review, regardless of the time of day, the preparer's tenure, or how many entries are in the queue.

Together, they catch more issues in the current cycle, before anything posts, rather than in correcting entries after close. The best part is your team avoids the worst type of work: re-work. 

A Starting Point, Not a Finished Product

Journal entry review is the AI Assistant's first use case. We're building this capability into more parts of the platform, and it will evolve as we learn from early customers.

This is currently available to a select group of customers in early access, with broader rollout to follow. If your team is buried in journal entry review volume and wants to see how it fits into your workflow, reach out. 

The journal entry queue isn't getting shorter. But it can get smarter.

‍