Finance & billing
Month-end document chase
Each client gets one precise request listing exactly which statements and receipts are still missing, and the close checklist updates as they arrive.
Starts when
The first working day of the month.
What happens
- Read and classify everything each client has uploaded
- Match documents to ledger transactions and bank activity
- List what is missing or unmatched for each client, down to the transaction
- Draft a client-specific request and schedule reminders
In more detail
An automation like this runs on the first working days of the month. It reads each client's ledger in QuickBooks Online or Xero and the files the client has uploaded, works out which bank or card statements are missing, which transactions need a receipt and which sit uncategorized, and drafts one email per client listing exactly those items with dates and amounts. The bookkeeper reviews and sends every email; reminders are drafts too. The firm needs a consistent chart of accounts, a shared upload folder and a receipt policy per client. Ordinary code does the matching, so figures come from the ledger, not the AI. The language work can run on a model on the firm's own computer, which keeps client bank data in the office. Products such as Uncat, Financial Cents and Karbon already offer client requests; Fenrir Labs could design a version built around the firm's own rules.
You stay in control
The bookkeeper reviews each request before it is sent.
Connects to
- Client portal
- Accounting software
- Bank feeds
- Email
Questions we would ask you
- Which ledger does each client use, and how many client files are on QBO versus Xero?
- Where do client uploads land today, and can that location sync to a folder on a firm machine?
- What is your receipt threshold, and does it differ by client?
- Do clients reply with attachments, uploads, or both? Which mailbox receives them?
- How do you mark a client "ready to close" today: a checklist in Karbon or Financial Cents, a spreadsheet, or memory?
Discuss this further with the Live Agent
Finance & billing
Bills and receipts coded into the ledger
Vendor bills and receipts are read, coded the way this client's past transactions were coded, matched to the bank feed and queued for review.
Starts when
A bill or receipt arrives by email, upload or photo.
What happens
- Read the vendor, date, amounts, tax and line items
- Suggest the account and class based on this client's history
- Match the document to a bank or card transaction
- Flag duplicates, new vendors and amounts outside the usual range
In more detail
An automation like this reads each vendor bill or receipt that arrives by email, upload or photo, pulls out the vendor, date, totals, tax and lines, and suggests how to code it based on how this client coded the same vendor before. It looks for the matching bank or card transaction and flags likely duplicates, new vendors and unusual amounts. A bookkeeper approves or corrects each entry before anything is posted to QuickBooks Online or Xero. It needs a few months of consistent coding history per client and one intake address or folder per client. Mature products such as Dext, Hubdoc and the ledgers' own capture tools already do much of this, so for many firms buying is the better choice. Fenrir Labs could design a version that runs the reading on the firm's own computer, or that adds exception checks on top of the tool the firm already uses.
You stay in control
A bookkeeper approves the coding; exceptions are reviewed one by one.
Connects to
- Email
- Accounting software
- Bank feeds
- Document storage
Questions we would ask you
- How do bills and receipts reach you today, and what share are photos versus PDFs?
- Do you already use Dext, Hubdoc or the ledger's own capture? What do you dislike about it?
- How many documents per client per month, and how many clients?
- Do you code at bill level or split lines across accounts and classes?
- Who approves entries, and would they approve in a queue or prefer to do it inside QBO or Xero?
Discuss this further with the Live Agent
Documents & compliance
Tax-season document tracking
You know, client by client, which forms have arrived and which have not. Clients are reminded about the specific documents still outstanding.
Starts when
A client uploads documents, or a reminder date passes.
What happens
- Identify each uploaded form and the tax year it belongs to
- Compare what has arrived with last year's return and this year's organizer
- Update the client's checklist and the firm-wide status board
- Draft a reminder that names the documents still needed
In more detail
An automation like this keeps a checklist for each tax client. It builds the list of expected documents from last year's return and this year's organizer, reads every upload to identify the form, tax year and payer, and marks items received. The firm sees one board showing who is complete and who still owes what, and staff get draft reminders that name the exact missing forms. Staff approve every reminder and decide when a return is ready to prepare. The main tax packages do not offer public APIs, so it works from return PDFs and portal uploads. Because this is tax return information, running the AI on the firm's own computer is the cleanest choice. Drake users should check their 2026 license, which restricts using the software's output with AI. Fenrir Labs could design this around the firm's existing portal.
You stay in control
Staff approve reminders and decide when a return is ready to prepare.
Connects to
- Client portal
- Tax software
- Email
- Status board
Questions we would ask you
- Which tax package do you use, and can it export a list of each client's prior-year source documents?
- If you use Drake, have you read the 2026 license section on automated means and AI?
- Where do clients upload, and can those uploads sync to a folder on a firm machine?
- Do you use an organizer, and in what form do answers come back?
- How many individual returns per season, and how many clients upload versus drop off paper?
Discuss this further with the Live Agent
Reporting
Pre-close anomaly check
Before a period is closed, the ledger is scanned for the things a careful reviewer looks for: duplicates, missing recurring bills and entries that break the pattern.
Starts when
A period is marked ready for review.
What happens
- Compare this period with the same accounts in prior periods
- Flag duplicate payments, missing recurring items and unusual amounts
- Check for uncategorized entries and transactions coded to suspense
- Write a review memo that explains each flag with links to the entries
In more detail
An automation like this runs when a client's month is marked ready for review. It compares each account with prior months and last year and flags what a careful reviewer looks for: possible duplicate payments, monthly bills that did not appear, amounts far above normal, anything left in suspense or uncategorized, and bank accounts that are not reconciled. The checks are ordinary code; the AI only writes a short memo explaining each flag with the entries behind it. The reviewer decides what is an error. It needs about a year of ledger history and the firm's own thresholds. A simple version can work from a general ledger export saved to a folder, with no connection to QuickBooks Online or Xero. Products such as Double and MindBridge offer review tools; Fenrir Labs could design one around the firm's own review checklist.
You stay in control
The reviewer decides what is an error and what is simply unusual.
Connects to
- Accounting software
- Bank feeds
- Review notes
Questions we would ask you
- What does your reviewer check today, and which errors have slipped through in the last year?
- Do you have a written review checklist we could turn into rules?
- Which ledgers, and are any Xero clients large enough that the Advanced tier is already justified?
- Would bookkeepers accept exporting the general ledger to a folder as a first step?
- Where should the review memo live: Karbon, Financial Cents, a shared drive, or the ledger's own notes?
Discuss this further with the Live Agent
Documents & compliance
Vendor tax forms collected before year-end
Every vendor that will need a 1099 is identified early, asked for their W-9 and followed up, so January is not spent chasing paperwork.
Starts when
A new vendor is paid, or the quarterly check runs.
What happens
- Find vendors who have crossed the reporting threshold or are likely to
- Check whether a current W-9 is on file
- Send a request with a secure upload link and remind as needed
- Read returned forms and update the vendor record
In more detail
An automation like this looks at each client's vendor payments every quarter and finds vendors who have reached, or are on pace to reach, the 1099 reporting threshold, which is $2,000 for payments made in 2026. It checks whether a W-9 is on file and drafts a request with a secure upload link, plus reminders. When a W-9 comes back it reads the name, tax classification and address and proposes an update to the vendor record. Staff send every request, confirm each classification and approve the final filing list. W-9s carry Social Security or employer numbers, so reading them on the firm's own computer is a plus. Filing and IRS TIN matching are best left to established services such as Tax1099 or Avalara, or the free IRS IRIS portal. Fenrir Labs could design the early-warning and reading steps around the firm's ledger.
You stay in control
Staff confirm vendor classifications and review the final filing list.
Connects to
- Accounting software
- Email
- Secure upload
- Document storage
Questions we would ask you
- How many clients do you file 1099s for, and how many forms in total last January?
- Which tool do you file with today, and does it already collect W-9s?
- Do you or your filing service have an IRIS Transmitter Control Code yet?
- Are vendors set up consistently in the ledger, with payment method recorded?
- Where are W-9s stored now, and who can see them?
Discuss this further with the Live Agent
Documents & compliance
Tax notice response preparation
A scanned notice is identified, matched to the client and return, and paired with the relevant figures and a draft response for the preparer.
Starts when
A client forwards or uploads a notice from a tax agency.
What happens
- Identify the notice type, tax year, amount and response date
- Pull the related return and the figures the notice questions
- Draft a response or a list of what must be gathered first
- Put the response date on the calendar and open a task
In more detail
An automation like this reads a tax notice the client forwards or uploads, often a phone photo. It identifies the agency, notice type, tax year, amount and reply-by date, matches it to the client, and finds the filed return for that year in the firm's files. It lines up the figures the notice questions against the return and prepares a packet for the preparer: a summary, the comparison, a list of what to get from the client, and an outline to start from. It proposes a calendar date and task for a person to add. The preparer reviews everything and signs every response; nothing goes to the agency or client automatically. It works from return PDFs, so no tax software connection is needed. Notices and returns are tax return information, so running it on the firm's own computer fits. Fenrir Labs could design this around the firm's file layout.
You stay in control
The preparer reviews the analysis and signs every response.
Connects to
- Client portal
- Tax software
- Calendar
- Document storage
Questions we would ask you
- How many notices does the firm handle a year, and which codes are most common?
- Where are filed returns stored, and are they PDFs you can find by client and year?
- How do notices reach you today: client email, portal upload, mail to the firm under a 2848?
- Who owns notice work, and how do you track reply-by dates now?
- Is a fact summary and gather-list enough, or do you want a draft letter outline?
Discuss this further with the Live Agent
Customer service
Client questions answered from the file
Routine questions such as whether a document arrived or when a return will be filed are answered from the actual status of the client's work.
Starts when
A client emails or messages a question.
What happens
- Recognize routine status questions and separate them from advice questions
- Look up the client's checklist, deadlines and recent activity
- Draft an accurate reply for routine questions
- Route anything that needs judgment to the right person with context
In more detail
An automation like this reads client emails or portal messages and sorts them into routine status questions and questions that need judgment. For status questions, such as whether a document arrived, what is still missing or when a return will be filed, it looks up the client's real checklist, job status and deadlines and drafts a short, accurate reply. Questions that need advice are routed to the right person with a summary. Staff approve and send every reply; the firm decides which question types may be drafted at all. It depends on practice-management status being kept current, and needs a connection to the mailbox and to a tool such as Karbon, Canopy or Financial Cents. Client messages hold financial and tax details, so running the AI on the firm's own computer is a good fit. Fenrir Labs could design this around the firm's own list of routine questions.
You stay in control
You choose which question types may be answered directly and which are drafts only.
Connects to
- Email
- Client portal
- Practice management
Questions we would ask you
- What are the ten questions clients ask most, and roughly how many a week in season?
- Where do client messages arrive: one shared mailbox, individual inboxes, the portal?
- Which practice-management tool holds job status, and is status kept current?
- Which question types would you allow to be drafted, and which must always go to a person?
- Would you ever allow some answers to be sent without review? Under what conditions?
Discuss this further with the Live Agent
