0:05 — Agenda for the session — The presenter outlines the plan: discussion of AI agents for finance teams, a live demonstration environment, a first look at AI for PO matching, the 2026 roadmap, and time for Q&A. Attendees are asked to post questions in chat, with unanswered ones to be followed up by email. 1:13 — Approach to AI in finance — The rationale for AI in spend management is described: agents shape approvals, accounts payable, and procurement workflows while the finance team stays in control. The stated principle is that AI should assist and suggest but never assume or make decisions on its own. 3:09 — Three agents for finance — The three focus areas are introduced: GL coding, which is already released and in customer use; PO matching, the subject of the sneak peek; and approver selection, which is next on the roadmap. The presenter notes that agents trained on a customer’s own history become more useful the longer that customer has been using the product. 5:52 — How the GL coding agent works — The agent matters most when an ERP default is missing or wrong. By default it evaluates vendor, invoice amount, and description if available, but the fields are configurable and customers have used entity, department, location, and other collected fields to guide its suggestions. 8:11 — Live GL coding demonstration — A Fujitsu invoice is opened in a demo environment connected to Blackbaud Financial Edge NXT, with the ERP defaults deleted so the distribution is empty. Clicking the AI button returns the three most likely account codes from prior invoices for that vendor, with account 015410 — office supplies — selected and saved. 11:02 — GL coding demo wrap-up — The presenter closes the demonstration and invites viewers to schedule a one-on-one session to see the feature in more depth or against their own ERP. 11:35 — Smart PO matching preview — Design mockups for the smart PO matching agent are shown: pulling in an invoice, displaying its lines, flagging matches and mismatches, letting users keep or edit individual items, and keeping POs open when an invoice does not complete them. Builds are underway with development expected around Q1 of the following year, and the feature was prioritized because earlier webinar attendees requested it above everything else. 13:24 — Roadmap overview — A recap of status: GL coding is available now, and PO matching is in development, with the invoice approver suggestion agent expected to follow. 13:56 — Invoice approver suggestion agent — The planned approver agent would recommend the next approvers based on vendor, dollar amount, and how similar invoices were handled in the past. Because it uses an engine similar to GL coding, the presenter expects it to be released relatively quickly, and it is intended to help both when defaults exist and when they do not. 14:52 — Reporting agent and conversational search — The existing advanced search lets front-end users build if-then queries without database administration. The proposed reporting agent would accept a plain-language request — such as all Fujitsu invoices over $10,000 from Q4 2025 — generate the search automatically, and allow it to be saved. Fraud detection and cash forecasting are also named as areas being explored. 16:55 — How to follow up — Attendees are told to reply to the post-event thank-you email to request a demonstration of the GL coding or PO matching agents, or to visit pairsoft.com or email info@pairsoft.com for more information, including ERP integration details.

Home / Resources / Webinars

Demo: AI for GL Coding PLUS AI Product Roadmap with PO Matching

Key takeaways

The GL coding AI agent is generally available in PaperSave today and suggests account codes based on a customer’s own historical invoice data.
By default the GL coding agent looks at three fields — vendor, invoice amount, and description (if present) — but the fields it considers are configurable, and customers have added entity, department, and location.
A smart PO matching agent was shown at the design-mockup stage only; it will match invoice line items to purchase order lines, flag matches and mismatches, and keep POs open when an invoice does not fully consume them.
PO matching was moved to the top of the roadmap because attendees of an earlier webinar asked for it more than any other feature, with development expected around Q1 of the following year.
An invoice approver suggestion agent is planned next, using an engine similar to GL coding to recommend approvers based on vendor, dollar amount, and past handling of similar invoices.
Further roadmap items include a reporting agent that would let users build advanced searches by typing conversational queries, plus fraud detection and cash forecasting.
The stated design principle is that AI agents suggest rather than decide, leaving the finance team in control of the final action.

Overview

This session covered two things: a live demonstration of the GL coding AI agent that is already available in PaperSave, and a first look at the planned smart PO matching agent along with the wider AI roadmap heading into 2026. The presenter framed the company’s approach to AI as assisting rather than deciding — agents should offer suggestions and step aside when appropriate, with the finance team retaining control of approvals, accounts payable, and procurement workflows. The stated goal is faster day-to-day decisions rather than replacing staff.

Three agents were named as the current pillars of work: GL coding (released and in production use), smart PO matching (the sneak peek), and approver selection (next on the roadmap). The presenter noted that agents built on historical customer data become more useful the longer an organization has been using the product, so the work benefits existing customers as well as new ones.

The GL coding demonstration used a live environment connected to Blackbaud Financial Edge NXT, with a Fujitsu invoice loaded through OCR and the ERP defaults deliberately deleted so no distribution was pre-filled. Clicking a small AI button in the upper right of the invoice screen returned the three most likely account codes based on prior invoices for that vendor — in this case account 015410, office supplies. By default the agent evaluates vendor, invoice amount, and description if one is available, but because the fields are configurable, customers have coached it using entity, department, location, and other collected fields. The practical benefit described was avoiding scrolling through a long account list and reducing the number of save actions per invoice.

The smart PO matching agent was shown only as design mockups, not working software. The intended behavior is to pull in an invoice, display its lines against the purchase order, indicate whether each line is a match or a mismatch, allow the user to keep or edit individual items, and leave POs open when the invoice does not close them out. The presenter said builds were underway and development was expected to begin around Q1 of the following year, and attributed the feature’s high priority to feedback from an earlier webinar in which attendees ranked PO matching above everything else. Customers wanting early access were told to contact their account manager.

Beyond those, the roadmap included an invoice approver suggestion agent built on an engine similar to GL coding — recommending approvers from vendor, dollar amount, and how comparable invoices were handled previously — which was expected to ship relatively quickly for that reason. A reporting agent was also described: instead of assembling if-then statements in the existing advanced search, a user would type a conversational query such as asking for all Fujitsu invoices over $10,000 from Q4 2025, have the system generate the search, and save it. Fraud detection and cash forecasting were mentioned as further areas under exploration without detail. The session closed with instructions to reply to the post-event email, visit the company website, or email the info address to request a one-on-one demonstration or ERP-specific walkthrough.

Chapters

0:05 — Agenda for the session — The presenter outlines the plan: discussion of AI agents for finance teams, a live demonstration environment, a first look at AI for PO matching, the 2026 roadmap, and time for Q&A. Attendees are asked to post questions in chat, with unanswered ones to be followed up by email.
1:13 — Approach to AI in finance — The rationale for AI in spend management is described: agents shape approvals, accounts payable, and procurement workflows while the finance team stays in control. The stated principle is that AI should assist and suggest but never assume or make decisions on its own.
3:09 — Three agents for finance — The three focus areas are introduced: GL coding, which is already released and in customer use; PO matching, the subject of the sneak peek; and approver selection, which is next on the roadmap. The presenter notes that agents trained on a customer’s own history become more useful the longer that customer has been using the product.
5:52 — How the GL coding agent works — The agent matters most when an ERP default is missing or wrong. By default it evaluates vendor, invoice amount, and description if available, but the fields are configurable and customers have used entity, department, location, and other collected fields to guide its suggestions.
8:11 — Live GL coding demonstration — A Fujitsu invoice is opened in a demo environment connected to Blackbaud Financial Edge NXT, with the ERP defaults deleted so the distribution is empty. Clicking the AI button returns the three most likely account codes from prior invoices for that vendor, with account 015410 — office supplies — selected and saved.
11:02 — GL coding demo wrap-up — The presenter closes the demonstration and invites viewers to schedule a one-on-one session to see the feature in more depth or against their own ERP.
11:35 — Smart PO matching preview — Design mockups for the smart PO matching agent are shown: pulling in an invoice, displaying its lines, flagging matches and mismatches, letting users keep or edit individual items, and keeping POs open when an invoice does not complete them. Builds are underway with development expected around Q1 of the following year, and the feature was prioritized because earlier webinar attendees requested it above everything else.
13:24 — Roadmap overview — A recap of status: GL coding is available now, and PO matching is in development, with the invoice approver suggestion agent expected to follow.
13:56 — Invoice approver suggestion agent — The planned approver agent would recommend the next approvers based on vendor, dollar amount, and how similar invoices were handled in the past. Because it uses an engine similar to GL coding, the presenter expects it to be released relatively quickly, and it is intended to help both when defaults exist and when they do not.
14:52 — Reporting agent and conversational search — The existing advanced search lets front-end users build if-then queries without database administration. The proposed reporting agent would accept a plain-language request — such as all Fujitsu invoices over $10,000 from Q4 2025 — generate the search automatically, and allow it to be saved. Fraud detection and cash forecasting are also named as areas being explored.
16:55 — How to follow up — Attendees are told to reply to the post-event thank-you email to request a demonstration of the GL coding or PO matching agents, or to visit pairsoft.com or email info@pairsoft.com for more information, including ERP integration details.

Questions this webinar answers

Is the GL coding AI agent available now, or is it still in development?

The GL coding agent has been released and is generally available in PaperSave. Organizations are already using it in production, and existing customers can take advantage of it immediately.

What information does the GL coding agent use to suggest an account code?

By default it looks at three things: the vendor, the invoice amount, and the invoice description when one is available. It compares these against the historical invoices already coded in the customer’s environment and returns the three most likely account codes.

Can the GL coding agent be tuned to consider fields other than the defaults?

Yes. The fields it evaluates are configurable, and since release customers have used other collected fields — including entity, department, and location — to guide the suggestions. Essentially any field being captured in the system can be used.

What happens if the ERP already supplies a default GL code?

An ERP default is useful most of the time when one exists. The AI agent is most valuable in the two cases where that breaks down: when no default is available at all, and when the default is not the correct code for that particular invoice.

When will the smart PO matching agent be available?

It was presented as design mockups rather than working software. Builds were reported as underway, with development expected to begin around Q1 of the following year. Customers who want early access were advised to email their account manager to be included in the first previews.

What will the smart PO matching agent do?

The design pulls in an invoice, displays its line items against the purchase order lines, and indicates whether each is a match or a mismatch. Users can open individual items to keep, edit, or change them, and purchase orders are intended to remain open when an invoice does not complete them.

Why was PO matching prioritized over other roadmap items?

Feedback from an earlier webinar showed that more attendees wanted to see PO matching than any other capability, so it was moved to the top of the development list.

What other AI features are planned beyond GL coding and PO matching?

An invoice approver suggestion agent is planned next, recommending approvers based on vendor, dollar amount, and how similar invoices were handled previously — it uses an engine similar to GL coding, so it is expected to ship relatively quickly. A reporting agent is also planned that would turn conversational requests, such as asking for all invoices from a given vendor over a dollar threshold in a specific quarter, into saved advanced searches. Fraud detection and cash forecasting were named as additional areas being explored.

Table of Contents

Get A Demo:

Simple solutions. Powerful results. Seamlessly integrated.

Get a PairSoft Demo