Direct answer
Manual reconciliation means a person compares two lists and decides each pair. Automated reconciliation means software proposes or posts matches according to rules, and a person reviews exceptions. F20 FinanceOps automates the proposal: a ten-pass ladder, then rules that classify leftovers. It does not post to the GL and it does not sign the control. Xero JAX can auto-reconcile bank feed lines when confident. BlackLine and FloQast automate high-volume matching and close workflow. Manual work remains correct for sparse, judgement-heavy breaks. Automation is correct for volume, replay, and consistent match order. Mixing them up — auto-posting judgement items, or ticking 20,000 lines by hand — is the failure.
Automation is not a moral upgrade. A bad rule applied 40,000 times is worse than a slow human. A human without a match order will also call a timing difference a missing receipt because they are tired. The question is which errors you prefer, and whether you can see them.
F20 is deliberately conservative. Findings that look like fraud patterns are worded 'Potential anomaly requiring review.' Duplicate detectors never instruct you to reverse a payment. Unapplied cash stops when several invoices could explain one receipt. That is automation that refuses to be brave.
What manual still owns
Materiality. Whether a $0.01 break is a reconciling item or a nothing. Software can compare to zero exactly — F20's balance-sheet recon does — but only a person decides whether to clear it.
Narrative. A 'consulting' wire that is actually a refund. A round-sum accrual that is business as usual on this account. Journal analyser flags weekend postings; it does not know about your Indian subsidiary's Saturday close.
Evidence. An email from the bank explaining a charge. Automation can attach nothing it was not given.
What automation actually changes
Order: identifier before amount before fuzzy text. People skip around. Ladders do not.
Completeness: every leftover gets a category instead of a yellow cell. F20's three-way match partitions seven states so counts sum to the population. That is a control Excel colouring never had.
Fatigue: the 400th line of a card settlement file. AutoRek and BlackLine exist because volume is a different sport. F20 is still a file-in tool, not a 2.4-billion-row platform — AutoRek's vendor claim on their site, not our test.
Degrees of automation, not a binary
Level 0: paper. Level 1: Excel. Level 2: file diagnostic (F20). Level 3: GL bank feeds (Xero, QBO). Level 4: close platform with sign-off and journals (FloQast, BlackLine, Numeric). You can be level 3 on cash and level 1 on intercompany. That is normal.
Xero's JAX page says it auto-reconciles when highly confident and otherwise suggests. That is level 3 with a human brake. F20 always leaves posting to you. That is level 2 with a hard brake.
Risks of each
Manual: skipped lines, inconsistent criteria, unreviewable history, key-person risk.
Automated: wrong keys (invoice number shared across suppliers — F20's GSTR-2B matcher uses GSTIN plus invoice for that reason), silent many-to-one, over-trust in OCR amounts, auto-journals that post noise.
F20 mitigates some automation risk by not posting and by stopping on ambiguous cash application. It does not mitigate 'you uploaded last month's file'. That remains manual.
Workflow table
Same recon, different labour.
| Manual | F20 | Close platform | |
|---|---|---|---|
| Who matches | Person | Ladder, then person on leftovers | Rules/AI, then preparer/reviewer |
| Who posts | Person | Person, outside F20 | Often the platform, with approval |
| Evidence | Workbook / binder | This run's tables (anonymous runs not stored) | Persistent audit trail |
| Best volume | Low | File-sized | High, ERP-connected |
Pricing
Manual: salary. F20: no charge. Close platforms: custom, NOT VERIFIED as a dollar figure here. Xero JAX needs Growing or above on the US page we read.
Privacy
Manual paper never phoned home. It also never encrypts. F20: in-memory. Platforms: SOC reports live on their trust centres — read those, do not take a converter article as a SOC substitute.
A close-week example
Monday: bank files arrive. Automated conversion and a first-pass match on F20 or in the GL feed. The population that matched on identifier is done. That is automation earning its keep.
Tuesday: leftovers. Timing differences, bank charges, a payment that hit without an invoice number. This is still mostly human, even if the software named the category. Auto-posting Tuesday's pile is how you bury a duplicate payment inside 'matched'.
Wednesday: judgement accounts — suspense, accruals, intercompany in transit. Automation that 'clears' these to make the dashboard green is not a close. It is a disguise. Keep them manual or on a platform whose sign-off is the point, not F20's anonymous run.
The mix is the method. People who say they are fully automated are usually automated on the easy 80% and silent on the rest. People who say they are fully manual are usually automated in Excel macros they forgot they wrote.
Methodology
Product pages cited in Sources. No claim that automation 'saves 70%'. BlackLine's site hosts customer quotes of that shape; they are vendor claims.
Where F20 FinanceOps is a better fit
- You want automated matching proposals on two exports without auto-posting.
- You are moving off pure Excel but not onto a close-platform implementation.
- You care that ambiguous cash application stops instead of guessing.
Where manual work or heavier automation is a better fit
- Sparse, judgement-only accounts: keep them manual, with a checklist. Automating them creates false comfort.
- Bank feeds already in Xero/QBO: use that automation; F20 will not replace the GL.
- SOX close with owners, due dates, and journals back to ERP: FloQast/BlackLine/Numeric as they describe themselves.
FAQ
It automates matching and classification of leftovers. It does not automate posting or sign-off.
On volume, usually. On a 40-line account with messy narratives, a person may still be faster. We do not publish a universal time saving.
No. It changes who does the pairing. The control is still that two sides tie.
It classifies outstanding cheques as carry-forward style actions in bank rec, not as 'ignore forever'. A person still decides.
F20 will match 2B JSON to a purchase register. ClearGST will also file. If you need filing, that is a different product. See Sources for ClearTax's official recon-and-file path.
Sources
- F20 FinanceOps policies — F20 FinanceOps. Verified. Uploaded files are processed in memory and dropped when the response is sent. There is no charge to run a tool. Output is not professional advice.
- Bank Reconciliation Software | Xero US — Xero. Verified. Xero reconciles bank feeds with JAX using Rule, Match, Memory and Prediction. JAX is on Growing and above. US list prices on that page: Early $25/month after intro, Growing $55, Established $90.
- Transaction Matching | BlackLine — BlackLine. Vendor claim. BlackLine Transaction Matching ingests multiple sources, matches transactions with business rules, and flags exceptions. Customer-quoted auto-match rates on that page are vendor-hosted, not F20 tests.
- Account Reconciliation Automation | FloQast — FloQast. Verified. FloQast reconciliation automation assigns preparers and reviewers, matches high-volume transactions, and integrates with ERPs, banks, and cloud storage.
- Numeric | AI-Powered Close Automation — Numeric. Vendor claim. Numeric publishes close checklists, account reconciliation, cash matching and bank integrations. Public list pricing was not on the homepage.
- GSTR-2B vs PR Recon | ClearGST — ClearTax. Verified. ClearGST reconciles GSTR-2B to the purchase register, can connect to the GST portal, and can auto-fill GSTR-3B Table 4.
Run an automated match on two files — You still approve every posting in your own books.