Automate your accounting firm, at a fixed price

In brief

Automating an accounting firm in 2026 first means tackling three projects: document collection with automatic follow-ups that know what's missing file by file, generation of recurring documents (engagement letters, presentations, correspondence), and an internal assistant that answers on tax documentation and the firm's procedures.

• Document collection is the first pool of savings: during tax season, client follow-ups can take 3 to 5 hours per week per staff member.
• A staff member commonly manages 40 to 60 files: it's impossible to track manually what's missing in each one without documents falling through the cracks.
• 2026 momentum: all businesses must receive electronic invoices by 1 September 2026, and SMBs must issue them by 1 September 2027. Your clients will have to change their habits: might as well take advantage of it.
• Budget-wise: a simple automation from €440 excl. tax, a complete business agent from €1,760 excl. tax, where agency custom work often starts at 15,000 €.
• Honest limit: the agent follows up, sorts and prepares. Review and advice remain with the chartered accountant.

The problem, concretely

When I talk with firms, the same finding keeps coming back: the peak load from January to May doesn't come from accounting technicality. It comes from the logistics around it. Chasing missing bank statements, purchase invoices, expense receipts. Retyping an engagement letter from last year's Word template, hoping not to forget to change an amount. Answering for the third time this week the same question from a junior about the VAT procedure for a particular case.

Do the math on your own firm: if each staff member spends even 3 hours a week chasing and flagging missing documents during the 16 to 20 weeks of tax season, that's 50 to 60 hours per person per season. In a 6-person firm, that's 300 to 360 hours a year of work with no accounting added value, the equivalent of 8 to 10 full weeks of one person.

And the calendar adds a reason to act now: the rollout of electronic invoicing (mandatory receipt for all businesses on 1 September 2026, issuance for SMBs and micro-businesses on 1 September 2027) will force your clients and your firm to review document circuits anyway. It's the right time to rethink the whole process rather than patch the old one.

Project 1: document collection, with follow-ups that know what's missing

The classic process, before: the staff member opens the file, compares what arrived with what should be there, jots down the gaps in a corner, writes a follow-up email, the client answers partially, and it all restarts the following week. Multiplied by 50 files, this manual flagging falls apart as soon as pressure rises, exactly when it should hold.

The process with a collection agent, after: the agent knows, for each file, the list of expected documents based on the client profile (monthly statements, purchase invoices, cash, expense notes, social documents). It flags what arrived in the mailbox or the DMS, identifies precisely what's missing, and follows up with the exact list: not a generic "please send your documents", but "we're missing your February statements and the vehicle invoice". Follow-ups space out and change tone based on history, and beyond a threshold the agent hands over to the staff member with the full context. No one follows up by hand, and nothing falls through the cracks.

A quantified example, on a fictional but realistic firm. The Vasseur & Associés firm: 6 staff members, about 220 files, micro-business and professional clients. Before: each staff member spent 3 to 4 hours a week on follow-ups and flagging during tax season, and about twenty annual reports went out late each year for lack of documents. After setting up a collection agent connected to the firm's mailbox and DMS: flagging is automatic, follow-ups go out on their own with the exact list of gaps, and staff only handle cases where a human is truly needed, about 45 minutes a week each. The gain: around 15 hours a week for the firm at peak season, and above all complete files earlier, so a smoothed workload instead of a wall in April.

Project 2: recurring documents, generated instead of retyped

Engagement letters, annual presentations, certificates, renewal letters, reports: a firm produces hundreds of documents every year that all follow the same skeleton, with the file data changing. Retyping them from a Word template is slow and, above all, a source of errors: last year's amount left in an engagement letter is a genuinely awkward conversation with the client.

The automated version: the templates are defined once, in your brand style, and documents are generated from the file data, ready to review and send. An engagement letter that took 30 to 45 minutes of copy-paste-and-check is generated in seconds, and the staff member only does the review. Combined with electronic signature, the complete circuit (generation, sending, signature, archiving) fits in a day instead of two weeks of back-and-forth. I detailed the orders of magnitude for this project in how much PDF document generation costs.

Project 3: an internal assistant on your tax documentation and procedures

Third pool of savings, less visible but very real: interruptions. Internal procedure questions, the "how do we handle this case here again?", searches in tax documentation or internal notes. Each question interrupts a senior, and each interruption costs double its duration in refocusing.

A RAG assistant (I explain it simply in what is RAG) indexes your procedures, internal notes and documentation, and answers the team's questions citing its source: the exact document and passage. The non-negotiable point: it's confined to YOUR documents and answers "I don't know, check with a senior" rather than inventing. It's exactly the setup that runs on beforbuild.com, my B2B SaaS, where the assistant answers on business documentation with systematic citation: the same mechanism applies to a firm's procedures.

And the limit, stated plainly: none of these three projects touches accounting judgment. The agent collects, generates, retrieves. Review, arbitration and advice remain human, and no binding document goes out without validation. On client data, the rules are those of any sensitive processing: per-file access rights, adapted hosting, and the framework set by the CNIL on AI and personal data.

How much it costs (and what it brings back)

On the market, custom development of this type at an agency or IT services company commonly ranges from 15,000 to 50,000 €, depending on scope and number of integrations. Existing collection SaaS tools cost less upfront, but they impose their process on your clients and rarely cover your specific cases: it's precisely on specific cases that manual follow-ups fall apart.

My approach sits in between: features already built and proven in production, customized to your firm. A simple automation (a follow-up workflow, a document generation) starts at €440 excl. tax; a complete business agent (multi-source collection, rules per file profile, monitoring) starts at €1,760 excl. tax. Scope and price fixed before we start, details on AI agents and automation.

The return is simple to calculate. Take the 300 annual hours from the example above: at 35 € of loaded hourly cost, that's more than 10,000 € of time per year consumed in follow-ups and flagging. An agent at €1,760 excl. tax pays for itself in the first tax season, and the following seasons are net gain, without counting the annual reports delivered on time and staff finally doing accounting.

Where to start

Three steps, in this order:

• 1. A free 30-minute audit. We review your document and file circuits, and identify the project with the best gain/effort ratio for your firm. No commitment, and you leave with the analysis even if we don't work together.
• 2. A single, measured project. Usually document collection, because the gain shows from the first season. We measure before (follow-up hours, incomplete files) to compare after.
• 3. Extend on proof. Once the first project is profitable, we add document generation then the internal assistant. The method applies to any repetitive process; I list others in 5 processes to automate in an SMB.

Sources

• impots.gouv.fr · I discover electronic invoicing
• economie.gouv.fr · Everything about electronic invoicing (2026-2027 calendar)
• CNIL · Artificial intelligence and personal data