{"id":79194,"date":"2026-10-21T13:55:50","date_gmt":"2026-10-21T11:55:50","guid":{"rendered":"https:\/\/juvo.be\/?p=79194"},"modified":"2026-10-22T13:32:22","modified_gmt":"2026-10-22T11:32:22","slug":"genai-in-finance","status":"publish","type":"post","link":"https:\/\/juvo.be\/en\/genai-in-finance\/","title":{"rendered":"GenAI in finance: what it does today, and what it doesn’t"},"content":{"rendered":"

[et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row column_structure=”1_2,1_2″ _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”1_2″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

sounds to many CFOs like a chatbot that takes over bookkeeping. That image is both too large and too small. <\/p>\n

Too large, because generative AI does not make bookings, does not execute payments, and does not make decisions that fall under your controller’s responsibilities. Too small, because it does much more than answer questions: it reads invoices, emails, contracts, and bank statements, makes connections, and prepares proposals. In this article, you will read what GenAI is concretely doing in finance today, where its limits are, which three misconceptions stand in the way of a good start, and how to choose an initial use case. <\/p>\n

[\/et_pb_text][\/et_pb_column][et_pb_column type=”1_2″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_image src=”https:\/\/juvo.be\/wp-content\/uploads\/juvo-genai-in-finance-wat-het-wel-en-niet-doet-hero.webp” alt=”Finance director reads a summary on her laptop at her desk, with an invoice, a contract and a bank statement beside her.” title_text=”juvo-genai-in-finance-hero” _builder_version=”4.27.9″ _module_preset=”default” border_radii=”on|6px|6px|6px|6px” global_colors_info=”{}”][\/et_pb_image][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” background_color=”RGBA(255,255,255,0)” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row make_equal=”on” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text admin_label=”Tekst” _builder_version=”4.27.9″ _module_preset=”default” background_color=”#F2F2F2″ min_height=”100%” custom_padding=”25px|25px|25px|25px|true|true” border_radii=”on|6px|6px|6px|6px” global_colors_info=”{}”]<\/p>\n

What GenAI in finance is: language, not calculation<\/h2>\n

Generative AI is strong in language in finance: reading, understanding, and explaining. Calculating and deciding are left to rules and people.<\/strong> Whoever makes that distinction well will obtain a system that can explain itself to its auditor.<\/p>\n

Most of the work in finance is not numerical calculation. Coding an invoice means understanding a description. Balancing a payment means interpreting a message filled with typos. Explaining a discrepancy in the monthly report means reading ten bookings and identifying the cause. That is language work, and generative AI excels at that. <\/p>\n

Calculating is something different. A language model can extract a period from a contract, but dividing by twelve follows a mathematical rule. A cash forecast arises from your outstanding items and a statistical model, not from a language model. Therefore, a well-designed application combines three elements: a language model that reads and explains, rules that compute and verify, and a person who decides. Every proposal contains a reference source, so you can always see what it is based on. <\/p>\n

In practice, such a combination is referred to as an agent. An agent carries out a defined task, such as reading the creditors’ mailbox or reconciling bank statements, and prepares the result as a proposal. The rule that we apply universally is simple: the agent proposes, a person decides.<\/strong> <\/p>\n

Where that concretely helps varies by domain, but the pattern is the same everywhere.<\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row admin_label=”Row” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

Five finance domains, one pattern<\/h2>\n

In every finance domain, the craftsmanship lies in reading and connecting information that exists outside the ERP, and there GenAI provides its value.<\/strong> The table shows per domain where that is located.<\/p>\n

[\/et_pb_text][et_pb_code _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

<\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n
Domain<\/th>\n

<\/p>\n

Where the craftsmanship resides<\/th>\n

<\/p>\n

What GenAI does<\/th>\n

<\/tr>\n

<\/thead>\n

<\/p>\n

Creditors (purchase-to-pay)<\/td>\n

<\/p>\n

Coding, matching, mailbox with reminders and status inquiries <\/td>\n

<\/p>\n

Proposes coding from your booking history, links lines to purchase orders, prepares responses<\/td>\n

<\/tr>\n

<\/p>\n

Debtors (order-to-cash)<\/td>\n

<\/p>\n

Evidence with the invoice, reconciliation without notification, disputes<\/td>\n

<\/p>\n

Collects work orders, finds invoices behind batch payments, recognises disputes<\/td>\n

<\/tr>\n

<\/p>\n

Month-end closing (record-to-report)<\/td>\n

<\/p>\n

Accruals, intercompany, reconciliations explained<\/td>\n

<\/p>\n

Reads periods from documents, seeks the cause of discrepancies<\/td>\n

<\/tr>\n

<\/p>\n

Controlling (FP&A)<\/td>\n

<\/p>\n

Collecting figures, explaining deviations<\/td>\n

<\/p>\n

Writes an initial commentary with source, calculates cash based on actual payment behaviour<\/td>\n

<\/tr>\n

<\/p>\n

Risk and compliance<\/td>\n

<\/p>\n

Post-hoc checks and sampling<\/td>\n

<\/p>\n

Identifies changes in banking details and anomalous patterns, with explanation<\/td>\n

<\/tr>\n

<\/tbody>\n

<\/table>\n

<\/div>\n

[\/et_pb_code][et_pb_text _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

What stands out in the table is that no application is included in the ERP itself. The sources are emails, PDFs, Peppol messages, bank statements, and contracts. This has an important consequence, and it brings us to the first misunderstanding. <\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row admin_label=”Row” _builder_version=”4.27.7″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

Three misconceptions that hinder a good start<\/h2>\n

The three misunderstandings we encounter most frequently concern autonomy, timing, and savings.<\/strong> Each of them leads to a wrong first decision.<\/p>\n

AI makes the decisions<\/strong> That is not so, and it is also not desirable. A well-designed agent does not book, does not pay, does not approve, and does not change master data. He proposes, with a certainty score and a source. You determine the threshold above which a proposal proceeds to approval. This makes the system defensible against your auditor and acceptable to your team. <\/p>\n

We are waiting for the new ERP to be available<\/strong> As the craftsmanship is outside the ERP, it survives the migration. After the go-live, the same emails arrive from the same suppliers. Agents who work on top of the ERP are already assisting today and will continue to do so at the go-live: you are changing the source, not the agent. <\/p>\n

It brings immediate savings<\/strong> Available time is capacity, not cash savings. At JUVO customers, more than 90% of incoming invoices are coded automatically, and at one client, the reconciliation of supplier statements fell from six to less than half a FTE. This is valuable, but your labour costs only decrease when you consciously decide what to do with that time. So do not count on savings that you do not see back. <\/p>\n

With those three misunderstandings out of the way, the question remains where you begin.<\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row admin_label=”Row” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

How do you choose your first use case?<\/h2>\n

A good first use case has a lot of volume, many exceptions, a measurable result, and exists outside of the ERP.<\/strong> Four criteria, and the more that apply, the better the candidate.<\/p>\n

– Volume<\/strong>. A task that recurs daily yields more than a task that occurs once a quarter. Invoice coding and reconciliation score highly here.
– Exceptions<\/strong>. Rules work well for the standard case. GenAI makes a difference with deviations: an unknown supplier, a spontaneous notification, a modified price.
– Measurability<\/strong>. Choose a task where you can determine if the proposal was accurate. A bank rule is either correctly reconciled or not. A coding is accepted or adjusted.
– Outside the ERP.<\/strong> A task that exists in jelle@jellevanlanduyt.be, documents, or bank files can be initiated without altering your ERP. <\/p>\n

In practice, the creditors’ mailbox, invoice coding, and the reconciliation of payments are often among the first to be addressed. Following this is a short cycle: a baseline measurement, a prototype using your own data within a few weeks, and a decision moment. If it does not meet the business case, you will know quickly. <\/p>\n

Sometimes the honest conclusion is that you should not start yet. If you process several hundred invoices per month from a fixed group of suppliers and your customers pay you neatly with a structured message, then the rules in your accounting package do most of the work. And if your master data is incorrect, a cleanup will yield more than an agent. We will tell you this as well, if that is the answer. <\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row admin_label=”Row” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text _builder_version=”4.27.9″ _module_preset=”default” global_colors_info=”{}”]<\/p>\n

What you can do tomorrow, even without us<\/h2>\n

Three figures tell you where GenAI delivers the most in your finance team, and you can gather them yourself.<\/strong><\/p>\n

Tell how many emails come in per week in your creditors’ and debtors’ mailboxes that are not invoices or payment proofs. Note how many bank transactions per month are not automatically reconciled. And ask your controllers how many hours the last month-end closing took in searching and explaining, aside from the accounting. With those three figures, you can see where the manual work lies, and thus where AI can make a difference. <\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” background_color=”RGBA(255,255,255,0)” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row make_equal=”on” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_cta title=”Wondering where GenAI would deliver most in your finance team?” url_new_window=”on” button_text=”Book a Discovery Session” _builder_version=”4.27.9″ _module_preset=”default” link_option_url=”juvo.be\/en\/get-in-touch” link_option_url_new_window=”on” global_colors_info=”{}”]<\/p>\n

In a Discovery Session we look at your processes together and identify your first use case.<\/p>\n

[\/et_pb_cta][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” background_color=”RGBA(255,255,255,0)” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row make_equal=”on” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_text admin_label=”Tekst” _builder_version=”4.27.9″ _module_preset=”default” background_color=”#F2F2F2″ min_height=”100%” custom_padding=”25px|25px|25px|25px|true|true” hover_enabled=”0″ border_radii=”on|6px|6px|6px|6px” global_colors_info=”{}” sticky_enabled=”0″]<\/p>\n

Frequently Asked Questions about GenAI in Finance<\/h2>\n

 <\/p>\n

What is the difference between GenAI and classical automation in finance?<\/h3>\n

 <\/p>\n

Classic automation follows fixed rules and works well for the standard case. Generative AI understands language and can therefore also deal with deviations: an unknown invoice format, a free payment reference, or a complaint in an email. In a good solution, both work together. <\/p>\n

 <\/p>\n

Is GenAI in finance safe and controllable?<\/h3>\n

 <\/p>\n

That depends on the arrangement. When the agent only makes suggestions, each suggestion receives a reference citation, and a human decides, the process is more verifiable than a manual booking with a loose note. The language model does not calculate or execute anything. <\/p>\n

 <\/p>\n

Where do you start with AI in finance?<\/h3>\n

 <\/p>\n

In a task with a high volume, many exceptions, and a measurable outcome, which operates outside the ERP system. The creditors’ mailbox, invoice coding, and the reconciliation of payments are often good candidates. <\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=”1″ _builder_version=”4.20.4″ _module_preset=”default” background_color=”RGBA(255,255,255,0)” custom_padding=”0px||0px||true|false” locked=”off” global_colors_info=”{}”][et_pb_row make_equal=”on” _builder_version=”4.17.6″ _module_preset=”default” width=”100%” global_colors_info=”{}”][et_pb_column type=”4_4″ _builder_version=”4.17.6″ _module_preset=”default” global_colors_info=”{}”][et_pb_testimonial author=”Bart Nulens” job_title=”Head of AI” company_name=”JUVO” url=”https:\/\/www.linkedin.com\/in\/bartnulens\/” url_new_window=”on” portrait_url=”https:\/\/juvo.be\/wp-content\/uploads\/bart-nulens-juvo.jpeg” _builder_version=”4.27.9″ _module_preset=”default” link_option_url=”https:\/\/www.linkedin.com\/in\/bartnulens\/” link_option_url_new_window=”on” global_colors_info=”{}”]<\/p>\n

Bart has been helping companies improve their finance processes for almost twenty years. In recent years he has done so mainly with AI and process automation, from invoice processing to the close. At JUVO he leads the AI offering for finance, customer operations and logistics. <\/p>\n

[\/et_pb_testimonial][\/et_pb_column][\/et_pb_row][\/et_pb_section]<\/p>\n","protected":false},"excerpt":{"rendered":"

sounds to many CFOs like a chatbot that takes over bookkeeping. That image is both too large and too small. Too large, because generative AI does not make bookings, does not execute payments, and does not make decisions that fall under your controller’s responsibilities. Too small, because it does much more than answer questions: it […]<\/p>\n","protected":false},"author":15,"featured_media":79528,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_seopress_titles_title":"GenAI in finance: what it does today, and what it doesn't","_seopress_titles_desc":"GenAI in finance is not a chatbot that takes over your accounting. What it does today across five finance domains, three misconceptions, and how to choose a first use case.","_seopress_robots_index":"","_seopress_robots_follow":"","_seopress_robots_imageindex":"","_seopress_robots_snippet":"","_seopress_robots_primary_cat":"","_seopress_robots_breadcrumbs":"","_seopress_robots_freeze_modified_date":"","_seopress_robots_custom_modified_date":"","_seopress_robots_canonical":"","_seopress_social_fb_title":"","_seopress_social_fb_desc":"","_seopress_social_fb_img":"","_seopress_social_fb_img_attachment_id":0,"_seopress_social_fb_img_width":0,"_seopress_social_fb_img_height":0,"_seopress_social_twitter_title":"","_seopress_social_twitter_desc":"","_seopress_social_twitter_img":"","_seopress_social_twitter_img_attachment_id":0,"_seopress_social_twitter_img_width":0,"_seopress_social_twitter_img_height":0,"_seopress_redirections_value":"","_seopress_redirections_enabled":"","_seopress_redirections_enabled_regex":"","_seopress_redirections_logged_status":"","_seopress_redirections_param":"","_seopress_redirections_type":0,"_seopress_analysis_target_kw":"","_seopress_news_disabled":"","_seopress_video_disabled":"","_seopress_video":[],"_seopress_pro_schemas_manual":[],"_seopress_pro_rich_snippets_disable_all":"","_seopress_pro_rich_snippets_disable":[],"_seopress_pro_schemas":[],"_et_pb_use_builder":"on","_et_pb_old_content":"","_et_gb_content_width":"","content-type":"","footnotes":""},"categories":[90],"tags":[],"class_list":["post-79194","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-in-finance"],"acf":[],"_links":{"self":[{"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/posts\/79194","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/users\/15"}],"replies":[{"embeddable":true,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/comments?post=79194"}],"version-history":[{"count":24,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/posts\/79194\/revisions"}],"predecessor-version":[{"id":79551,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/posts\/79194\/revisions\/79551"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/media\/79528"}],"wp:attachment":[{"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/media?parent=79194"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/categories?post=79194"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/juvo.be\/en\/wp-json\/wp\/v2\/tags?post=79194"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}