Autonomous Accounts Payable: driving end-to-end workflows with AI.
Take a closer look at the current state of AP and find out where businesses stand regarding the use of Artificial Intelligence (AI).
There’s a gap in AP Automation.
7 ways finance professionals describe their organization’s approach to AP automation
These results indicate many companies are aware of how valuable automation would be for their accounts
payable processes — but are finding it difficult to get the ball rolling on their automation journey.
How Autonomous AP differs from AP Automation.
Basic AP automation requires
configuration to be maintained by humans
Medius's autonomous AP solution maintains itself using AI and Machine Learning, compiling data and managing invoice processing without human oversight. Medius has spent 10+ years building AI into the core of its platform, processing over 6 million invoices per month across 3,000+ customers and $400bn+ in annual spend.
Where basic automation relies on static rules and manual configuration, Medius deploys a multi-stage AI pipeline, covering classification, extraction, verification, and coding, that continuously learns from real-world corrections. The result: a 96.3% touchless processing rate for PO invoices, compared to a 23.4% market average (Ardent Partners 2025), and a ~70% reduction in exception rates. Non-PO invoices reach a 99.5% touchless rate, compared to a 49.2% industry average (Ardent Partners 2025).
Two AI technologies at the forefront of autonomous AP are intelligent document processing (IDP) and OCR. Medius goes further, combining proprietary convolutional neural networks, Markov models for line-item extraction, and confidence-scored ML routing, capabilities that cannot be replicated by bolting a general-purpose LLM onto a legacy workflow. Medius's proprietary models are 947x faster and 25x more cost-effective than LLM-based alternatives, enabling throughput at enterprise scale without margin erosion.
What is preventing your organization from
switching to Automated AP Solutions?
Top 6 reasons companies aren't automating AP
- 60% Lack of internal resources to support projects
- 49% Lack of capital
- 46% Concerns about managing the change
- 44% Competition among other projects internally
- 44% Lack of support from IT for implementation
- 44% Lack of a compelling business case
These only brush the surface of the roadblocks we found in our survey. Download the report to get all the data!
"Without Medius we would have needed to double or triple the headcount, but we’ve been able to maintain the volume, with the team that we had."
Mayra Perez
AP Manager, Chadwell Supply
Implementing Autonomous AP
But what’s the first step to take?
To start a successful AP automation, you must first:
- Perform an objective end-to-end process assessment and benchmarking
- Identify the 3 P’s to support the automation: people, process, and products
- Establish your tactical and strategic approach in your automation journey, set your priorities, measures of success and bring your stakeholders onboard and align your goals with the overall organizational digital transformation plans
Getting beyond basic automation.
Check out the Q&A session SSON conducted with our VP of Capture and AI — Katarina Andersson.
Katarina Andersson:
I think one thing that’s well known in invoice processing is that PDF invoices help you be more automated because you don’t have to deal with manually keying in information which brings a greater risk of errors and takes a lot of time. PDFs provide a well know format that everybody knows how to use and are not perceived as technical. PDFs helps give greater insight and control to your invoices, all while reducing manual labor. PDFs are also easier to archive than a paper invoice because they come in a digital format and don’t require human filing. You can use AI tools to easily scan, interpret, and place PDF invoices in the correct archives for easy access.
Katarina Andersson:
The whole concept of “electronic” invoices has been around for 20+ years. I think it’s important to define what an “e-invoice” actually is. An e-invoice is one that is submitted electronically, i.e. it’s not a paper sent by post. This could be a PDF attachment to an email that is handled automatically by a system or a true point-to-point solution like EDI or an invoice network solution that allows a one to many connection.
The reason PDFs had been utilized more is because they are not as technical as E-invoices, which require technical skills to map fields from supplier invoice to customer invoice record and can still remove a lot of manual work. However, because of the global pandemic, governments have been requesting businesses to file their expected tax receipt through e-invoices which can be regulated and to collect their money in a more controlled manner.
Katarina Andersson:
Businesses are realizing that with the electronic invoice solutions that are available, such as the one we offer at Medius, there are AI capabilities that are making all E-invoices less technical because we don’t have OCR errors, just validation and mapping errors. These solutions make it so E-invoices no longer require c constant oversight and maintenance from IT, and that AP teams can manage these invoices even if they lack technical skills.
Katarina Andersson:
Medius loves all invoices! It’s worth pointing out that our solution is a single UI for invoice capture, processing and payment therefore, giving AP teams end-to-end control of their invoice process in a streamlined fashion, connected to 100+ ERP systems including SAP, Oracle, Microsoft Dynamics, and NetSuite.
Medius uses Machine Learning to learn through manual corrections to invoices, whether from recurring exceptions like PO number format mismatches or supplier identification issues. Medius learns these patterns and resolves them automatically each time, saving both the AP team and the supplier the cost and effort of repeated intervention. The platform has accumulated 2.4 billion+ invoice field data points and over 393 million real-world corrective labels, including critical edge cases such as tax code corrections and cost center assignments, creating a domain-specific learning loop that deepens with every invoice processed. This dataset spans 10+ years of production AP data and is structurally impossible to replicate, even with synthetic data generation.
ML also detects fraud and duplicate payments early in the process, Medius's AI-powered fraud detection reduces payment fraud and error rates by 90–95%.
Machine Learning also helps once invoices are in the system and not matched to PO's. There is often the manual task of invoice routing to the correct person for approval and the coding that needs to be done so it is accounted for properly in finance systems. Once the system has "seen" the user make the same updates again and again it easily learns what to do, Medius's SmartFlow coding reaches 95%+ precision after just two invoices, allowing finance teams to focus on exceptions, not the day-to-day. Medius's Supplier Conversations AI agent goes further still, autonomously ingesting and responding to supplier emails within guardrails, reducing supplier inbox management from ~8 hours per week to ~30 minutes, a ~94% reduction in manual effort.
Katarina Andersson:
Employees don’t want to learn multiple applications that require multiple logins and several training sessions. We hear from clients that what they love about autonomous AP is the single log-in and Machine Learning. The longer employees work with it they also see how helpful it is at capturing the invoice data, and how it removes manual touchpoints they were responsible for throughout the process.
AP teams are not typically made up of technical people. They are numbers people, and they want control of their data and processes. So, if we give them that control, give them a way to prevent fraud, and remove the maintenance E-invoices require, we allow employees to work on the tasks they want to work on. Another benefit for employees is that autonomous AP solutions provide statistics to AP teams that help them understand where they are getting better and shows them what resources they might need for years to come. Medius customers typically go live in 8–12 weeks, without a months-long IT implementation, and have reduced average invoice processing costs from an industry average of ~$9.87 per invoice to as low as $2.81 (Ardent Partners 2025), while cutting invoice cycle time from an industry average of 9–17 days to approximately 6 days.
Want to know more?
Get our free report.
The goal of this report is to illustrate how autonomous AP has become the great differentiator for businesses looking to improve their end-to-end processes.
Autonomous AP does not require constant oversight from IT to implement, it can be an easy plug and play solution that quickly helps AP teams overhaul their processes, backed by 25+ years of AP domain expertise and an AI-native architecture built since 2016, recognized by Gartner, Ardent Partners, and Spend Matters as a leader in AP automation. Medius holds a +64 NPS against a 39-point peer average, a signal that customers experience the performance difference, not just the promise of it.
