Why invoice automation still falls short, and how self-learning AI is changing the equation for IFS customers
If you run IFS, there is a good chance invoice automation has been on your roadmap for years.
Written by Emma Blackmore, CMO at Snowfox
A guest post from Snowfox, written for the Addovation community.
Most organisations have invested heavily in increasing PO coverage, because PO matching is the most established route to automation in IFS, and for purchase-order invoices it works well. OCR is in place, approval workflows are configured, coding rules have been refined and re-refined. On paper, the process looks automated.
In practice, many finance teams still find themselves manually coding a meaningful share of their invoices, particularly non-PO invoices, which remain one of the hardest problems in accounts payable. If that sounds familiar, you are not alone, and it is not a failure of effort or investment. It is a limitation of the underlying approach.
In this post we want to share what we are seeing across the IFS ecosystem, why the traditional automation model has surreptitiously plateaued, and what changes when self-learning AI is added directly into the IFS invoice workflow.
A quick thank you to the team at Addovation for the invitation to share our perspective with their community.
The automation gap inside IFS Finance
Most IFS customers have already taken meaningful steps to digitise their finance processes. The gap we see is almost always in the same place: the step between a scanned invoice and a fully coded, correctly routed document ready for approval.
Purchase order invoices are relatively straightforward, because the PO gives the system enough structure to validate, match and route automatically. Non-PO invoices, on the other hand, are a different problem. There is no PO to match against, the coding depends on organisational context, and the right approver often depends on the nature of the spend, not a fixed rule. This is where finance teams still spend hours every week.
Where traditional automation breaks down
Most invoice automation in place today is built on three components:
- OCR for data extraction
- Rule-based logic for coding
- Workflow tools for routing and approvals
Each plays an important role. But all three depend on predefined structure. OCR can extract the data on an invoice, but it cannot decide how that invoice should be coded.
Rules can apply logic, but only in scenarios you have anticipated and configured. And finance, as anyone who has worked in it knows, rarely sits still:
- New suppliers appear constantly
- Invoice data is incomplete or inconsistent
- Cost structures and GL hierarchies evolve
- Supplier-based coding templates, which are very common in IFS, struggle when the same supplier sends invoices that need to be coded differently
- Multi-line, multi-entity invoices don’t fit neat templates
The result is that only a portion of invoices can be handled automatically. The remainder flows back to a human for review, correction and routing, and every new rule written to catch an edge case becomes another rule to maintain.
We explored this in more depth in a recent piece on why most IFS teams still code invoices manually, if you want to dig further into the underlying mechanics.
The hidden cost of partial automation
This creates a less visible but significant operational drag. In IFS specifically, around 90% of non-PO invoices typically sit outside any automation rule or template, meaning the majority are coded fully manually, with only a small set of recurring invoices picked up by rules. Teams spend their time maintaining those rules, reviewing exceptions, correcting coding inconsistencies and training newer colleagues on an ever growing list of edge cases. Automation exists, but it doesn’t scale. As transaction volumes grow, so does the manual overhead that sits on top of it.
A shift in approach: from rules to learning
Self-learning AI changes the starting point. Instead of being told how to code an invoice through rules and templates, the model learns from your organisation’s historical invoice data, how your team has actually coded invoices in the past, how approvals have actually flowed, and where exceptions have genuinely occurred.
From that, it can predict, invoice by invoice:
- The correct GL account, cost centre and other coding dimensions such as project or business unit
- VAT treatment
- The right approver and approval route
- Free-text field values specific to your coding patterns
And because it keeps learning from every correction and every new invoice, accuracy improves over time rather than eroding as the business changes. There are no templates to build and no rules to maintain. The model adapts automatically as your coding patterns and approval policies evolve.
What this means inside IFS
For IFS customers, the practical question is how this fits with an environment you have already invested in. The short answer: it sits inside it, not alongside it.
Snowfox is designed to work within the existing IFS invoice workflow rather than replace any part of it. IFS remains your system of record. Your approval hierarchy, your chart of accounts, your governance model, all of it stays in place. What changes is that invoices arrive at the approval stage already coded and routed, learned from your own historical data, with accuracy that improves transaction by transaction.
In operational terms, that typically means:
- A materially higher proportion of invoices handled without manual coding
- Fewer touchpoints per invoice, and faster cycle times
- More consistent coding across entities and across new joiners
- Finance capacity redirected from data entry to analysis and control
It is also worth noting that this approach supports organisations at different stages of their IFS journey, including those moving from earlier versions onto IFS Cloud.
Automation capability does not need to be rebuilt every time the underlying platform evolves.
For a closer look at this working inside IFS, including a live walk through, we recently ran a webinar with IFS on automating supplier invoice coding using self-learning AI which is available to watch on demand.
Why the partner matters
Introducing AI into a live finance process is as much about implementation judgement as it is about technology. Workflows need to be understood properly before they can be automated well. Ownership between finance and IT needs to be clear. And integration into IFS needs to be done by people who genuinely know the system.
This is where experienced IFS partners, such as Addovation, play a central role. The combination of deep IFS knowledge on one side and specialist AI on the other is what turns a promising capability into a real operational change. One that lands cleanly, holds up in audit, and keeps improving after go-live.
Join Addovation’s upcoming webinar with Snowfox!
On May 13 at 08:30 CEST, we’re hosting a webinar together with Snowfox, where we’ll explore several of the challenges highlighted in this blog post in more depth.
Secure your spot today—don’t miss out!
From incremental gains to scalable efficiency
When automation moves beyond rules and begins to learn from real data, the character of the benefit changes. It is no longer a fixed percentage saved on a fixed set of invoices. It becomes a capability that scales, one that gets better as your volumes grow, absorbs new suppliers without reconfiguration, and reduces the maintenance burden rather than adding to it.
For IFS customers, the question is no longer whether invoice automation is possible. It is how to move from partial automation, with its rules and its exceptions, to something more scalable, more accurate and genuinely sustainable.
That shift is already well underway across the IFS ecosystem. If it’s something you’re exploring, your Addovation team is a good place to start the conversation, and you can read more about how Snowfox works within IFS on our IFS partnership page.
About the author
Emma Blackmore is the CMO of Snowfox, where she leads marketing and partner storytelling across the Snowfox ecosystem. With a background spanning B2B SaaS and partner-led growth, Emma is focused on how specialist technologies like Snowfox land in the real world, through the partners who implement them and the finance teams who live with them every day. She is a firm believer that the best marketing in this space starts with genuinely understanding the work, and she brings that perspective into everything Snowfox puts out.
