Artificial intelligence in food manufacturing is most likely to conjure up images of robotics on the factory floor for packaging products or autonomous quality assurance devices on the assembly line. However, there’s another area where AI is having an enormous impact on efficiency and profits: the back office.
Accounts payable (AP) has become one of the biggest priorities for automation due to its document-heavy processes and high consumption of time and resources, often pulling employees’ attention away from more customer-facing and value-generating tasks. In fact, recent research shows the AP automation market is set to soar from around $6 million in 2024 to $17 million by 2032, underlying its importance in the digital transformation journey.
Mars is one such company that made the decision to prioritize AP automation in order to keep up with ongoing growth. Maxime Vermeir, senior director of AI Strategy at intelligent automation company ABBYY, helped Mars standardize its AP processes across its global offices. Vermeir has a decade of experience in product and tech, and his expertise in AI enables business solutions and transformation initiatives.
FOOD ENGINEERING sat down with Vermeir to get firsthand insight into the challenges, strategies and results with implementing AI in accounts payable.
FOOD ENGINEERING: What were Mars’ most pressing pain points prior to automating accounts payable with AI?
Maxime Vermeir: Their accounts payable department experienced challenges that are common for any organization without AP automation: lots of manual data entry resulted in errors and inconsistencies, ultimately demanding a significant amount of extra work that could otherwise be avoided.
These are obvious targets for AI-powered improvement, but additional layers to their challenge were their rapid growth and global scale. Mars would have had to hire 50-75% more people to keep stride with their invoices and needed to find a way to meet that need through automation instead. Furthermore, their existing AP staff is spread out across many different countries, each holding their own tribal knowledge of best practices and regional variances.
In short, their staff was bogged down with data entry instead of steering the starship to boldly go where no one had gone before.
FE: How exactly did AI fulfill this standardization need? What was Mars’ strategy?
MV: It was important for Mars to keep the business side of their organization in the loop throughout this process. That was the catalyst for success with their strategy, which was ultimately to create a core standard that encapsulated how their AP processes should look at a high level with regional variations taken into account. Mars created two documents, each over two hundred pages, describing the details and nuances of their AP functions across respective regions.
Choosing the right AI solution was like designing a new Iron Man suit: combining components of cutting-edge tech into one robust system that could solve this complex problem. They selected a low-code and cloud-based intelligent document processing (IDP) platform that leveraged natural language processing (NLP) and machine learning, through which they could aggregate invoices from over two thousand different vendors into their ERP system. With NLP enabling semantic analysis to contextualize AP vernacular while machine learning enabled the training of AI models on an infinite array of document formats, Mars could extract valuable data consistently with both speed and accuracy.
Using this IDP approach, Mars socialized AI-enhanced invoice-to-pay processes across 20 global markets in 14 different languages.
By weaving AI into accounts payable, Mars was able to pursue strategy and value with staff that would otherwise be allocated to more monotonous back-office responsibilities.
FE: What were the benefits of this initiative?
MV: By weaving AI into accounts payable, Mars was able to pursue strategy and value with staff that would otherwise be allocated to more monotonous back-office responsibilities. IDP significantly accelerated invoice processing and achieved higher straight-through processing (STP) rates, meaning that a large portion of their documents could be processed without any manual intervention from human employees.
Alleviating this heavy document burden meant that they could engage with judgment-based objectives like disputing transactions and other value-added activities that can’t be performed autonomously. Beyond the obvious benefits to efficiency and revenue, this also meant a reduction of monotonous, slogging tasks for employees. With recent survey data revealing that 92% of employees burn up to eight hours a week scouring documents for information, this isn’t negligible; it could be the difference between employees enjoying their roles and burning out entirely.
FE: Should Mars have done anything differently?
MV: Mars’ implementation was both an anomaly and a master class. We saw an atypically rapid pace in their growth, so they had to be particularly strategic with introducing their automation strategy.
They took a great first step by ensuring their finance department was looped in throughout the implementation process. That’s a must-have for guaranteeing the long-term efficacy of AI—it can’t just be from a technological lens. It has to translate to business value and solve a real challenge.
While it feels like a superficial answer, the only thing that truly comes to mind is starting earlier. Jumping behind the wheel of the DeLorean and accelerating in a strategic, data-driven direction allows more time for interaction and adaptation with the technology, ultimately giving you a head start toward the future of operational excellence. With AI and automation, you can only really know that it works once you’ve had time to interact with it and see how it fits into the full scope of your organization. Without that hands-on experience, it’s difficult to build a strong foundation.
I’d advise organizations who are strongly considering automating AP processes to be careful in their consideration of solutions and implementation partners, and that they make data-driven decisions as to where AI can play the strongest role.
For example, intensive efforts to organize complex and varied processes into formalized documents could benefit immensely from data-driven tools like process intelligence. By gathering data at every step of a workflow, process intelligence yields the most comprehensive visibility into how processes are completed from end to end. This allows for efficient and accurate representations of core workflows, which could drastically expedite initiatives like Mars’ 200-page documents describing AP workflows.
When I recently joined Mars at an SSON AP Automation Digital Summit, 70% of attendees said they were evaluating or learning about using AI in accounts payable, and I believe this to be a strong suggestion that this growing trend isn’t diminishing anytime soon. Missing the AI train could be a recipe for disaster for any food manufacturer.