Solving 3-way matching of invoices with AI

Find out interesting insights with Jon Naseath , COO, Osmo

Moderated by Emily, Digital Transformation Consultant at Hyperbots

Don’t want to watch a video? Read the interview transcript below.

Emily: Hi, everyone. This is Emily, and I’m a digital transformation consultant at Hyperbots, I’m very pleased to have Jon Naseath on the call with me. Jon is a chief operating officer at Osmo. The topic that we’d be discussing today is why matching invoices with purchase orders and goods receipt notes is tedious, and also how AI solves it. Thank you so much for joining us, Jon. To start off, can you please explain some of the major challenges that organizations face with invoice matching to purchase orders and goods receipts?

Jon Naseath: The fundamental issue is, vendors want to get paid. You’ve got all this purchase order process upfront to get approval for payments, and accounting isn’t going to release the funds until you’ve verified that the services have been done or the goods receipts have been received. The hold-up is usually vendors calling up their contacts within the company and saying, “Where’s my money?” And then you have to verify, “Well, did you get the work or the goods?” Then you can pay them. Accountants love paying people, but they want to make sure the boxes are all checked.

Emily: Can you provide an example of how complex matching requirements can affect the invoice processing workflow?

Jon Naseath: Sure. It’s usually data disconnects. There was a plan when the PO was created, and then the invoice had something slightly different. With goods receipt, it should be straightforward. For example, the invoice lists 100 units of product, and the PO specifies 90, but the goods received say 85. They’re trying to charge you for 100, but you approved 90, and they only sent 85. So what are you going to pay them? It usually takes effort instead of flowing through automatically.

Emily: Understood. What are some of the common format differences between invoices, POs, and GRNs that complicate the matching process?

Jon Naseath: A lot of times, especially in international transactions, there are differences like month-day-year versus day-month-year formats. There are also differences in units of measure whether it’s quantities or services provided. Sometimes the invoice might be for work performed, and you have to verify if they completed the work. Did they do what they were supposed to, or are they just saying that? Also, is the person signing off on the work holding the vendor accountable, or just saying “pay them”?

Emily: Got it. So, Jon, how does data entry error impact the accuracy of invoice matching?

Jon Naseath: If it’s intentional, it’s a fraud, but if it’s an error, it can be small things like entering the amount in euros when you’re expecting US dollars. Data entry errors like this can cause issues with reconciling numbers. For new vendors or publishers, it can be a lot of work to chase down little data points. Meanwhile, vendors are asking, “Where’s my money?” Another example is when a customer uses a DBA (doing business as) name, and they send a slight variation of their name, like Vendor Inc. instead of Vendor LLC. Data quality matters.

Emily: It sounds incredibly overwhelming. So how can AI help in automating the data extraction and normalization process?

Jon Naseath: It’s two-fold. First, avoid the issue in the first place. AI can help by reconciling the data against the PO to catch discrepancies before sending it. This helps vendors get paid faster. On the receiver side, AI can flag errors quickly so they can be resolved before reaching accounts payable. Ideally, it flags the issue and sends it to the business owner of the account to fix it before accounts payable is even involved.

Emily: Got it. What role does AI play in detecting and correcting errors in invoice processing?

Jon Naseath: AI can identify common errors in documents like typos, incorrect item codes, or mismatched numbers. It also looks at historical data trends to detect patterns. If an accounts payable clerk is manually processing hundreds or thousands of invoices, they can easily miss these issues. I remember joining a company where the accounts payable clerk was buried under a mountain of invoices. We automated some of it, but it was still painful. AI can help people in these situations and reduce their workload.

Emily: Can you explain how AI algorithms detect anomalies and discrepancies in invoice matching?

Jon Naseath: AI is very effective at identifying patterns and spotting discrepancies in quantities, prices, or item descriptions. AI does this across hundreds of variables and can instantly flag issues that a human might miss. A typical accounts payable clerk might not be motivated to catch these anomalies, especially if they’re overwhelmed by the volume of work. AI helps mitigate those risks.

Emily: How does AI handle the challenges of matching invoices that reference multiple purchase orders or involve partial deliveries?

Jon Naseath: In accounting, it’s easy to think everything should line up perfectly in a two-way or three-way match, but in reality, you often have invoices referencing multiple POs or partial deliveries. You don’t want to delay payments by asking vendors to reissue invoices. AI can reconcile these discrepancies and help keep everything in order across big POs and multiple transactions.

Emily: To wrap things up, what are the key benefits of integrating AI into the invoice-matching process for an organization?

Jon Naseath: Integrating AI into invoice matching automates repetitive tasks, reduces manual errors, improves data accuracy, and enhances anomaly detection. It helps you get the job done faster and protects you from costly errors, like overpaying a vendor or missing a payment. AI is like having an extra set of eyes to help you avoid mistakes.

Emily: Got it. Thank you so much, Jon, for talking to us about why matching invoices with purchase orders and goods delivery notes is tedious, and how AI can help. It was great having you today.

Jon Naseath: Great, my pleasure.

Matching Strategies for Open-ended Services like Time and Material

Find out interesting insights with Ayo Fashina, CFO Kobo360

Moderated by Emily ,Digital Transformation Consultant at Hyperbots

Don’t want to watch a video? Read the interview transcript below.

Emily:  Welcome, everyone! My name is Emily, and I’m a digital transformation consultant at Hyperbots. Today, I’m excited to have Ayo, an expert in finance and strategic negotiations, join us for a conversation about managing time and material service invoices. Ayo’s vast experience in financial analysis and transformative project finance is going to give us some valuable insights into this topic. To start, Ayo, could you explain the key challenges that organizations typically face with time and material service invoices?

Ayo: Thank you, Emily. When we talk about time and material services, we’re referring to services where the billing is based on the time spent by consultants and the materials used. This type of service is inherently ambiguous if not tracked properly. The primary challenges include the lack of predefined quantities, variabilities in service delivery, difficulties in verifying time sheets, and ambiguities around what constitutes service completion. Unlike fixed-scope services, where everything is well-defined, T&M services can lead to uncertainties when trying to match invoices with the actual work performed.

Emily:  That makes sense. So, how does an organization typically handle the verification of service delivery for T&M contracts?

Ayo: Ideally, a company should use detailed time sheets, service delivery notes, and performance metrics. Timesheets provide records of the hours worked, while service delivery notes confirm that the work was done. Performance metrics allow us to evaluate whether the service meets the agreed-upon criteria. The clearer these criteria are defined at the start, the easier it is to manage the project and process invoices later on.

Emily: Can you describe the process of matching T&M invoices with purchase orders and contracts?

Ayo: When matching invoices, organizations can use two-way and three-way matching processes. In two-way matching, you compare the invoice with the PO or contract to ensure the rates and terms align. For example, if the rate is $50 an hour for four hours, you confirm that both the rate and time match the agreement. Three-way matching adds another layer, incorporating service delivery notes and time sheets. You verify that the agreed-upon work was done and that materials were delivered as expected. This ensures the invoice amount corresponds to the actual work completed before you proceed with payment.

Emily: AI is increasingly being used in business processes. What role can AI play in matching T&M service invoices?

Ayo: AI can be incredibly valuable in automating data extraction from invoices, tracking time sheets, and analyzing service delivery notes. It can quickly identify discrepancies, like unusual billing rates or hours worked, and predict potential issues based on historical data. This predictive capability is powerful and reduces manual effort while enhancing the accuracy and speed of the matching process.

Emily: What do you think are the most significant benefits of using AI in this context?

Ayo: AI is faster, more accurate, and less prone to errors than manual processes. It can handle large volumes of invoices that would otherwise require entire teams of people. Moreover, AI can detect anomalies and flag issues based on historical patterns, which would take a human much longer to identify. By automating routine tasks, AI not only speeds up the process but also enhances its reliability.

Emily: One challenge that comes to mind is verifying service delivery in the absence of physical goods. How can that be addressed?

Ayo: One way is through service delivery notes that are signed by both parties and both the provider and the recipient of the service. This provides a document trail confirming the service was delivered as agreed. Additionally, using well-defined performance metrics and conducting regular reviews ensures that the hours billed and materials used match what was agreed upon. The person signing off from the company’s side ensures the verification is accurate.”

Emily: Have you seen any recent improvements or changes in the way T&M invoice matching is handled in your industry?

Ayo: Absolutely. We’ve seen significant improvements with more robust documentation requirements and integrated systems that streamline the process. For example, AI now handles a lot of the data extraction and anomaly detection. Traditional OCR technology has limitations, but AI can read even handwritten documents or unclear invoices. These advancements have really enhanced the efficiency and accuracy of matching T&M invoices.

Emily: Do you have any advice for organizations looking to improve their T&M invoice matching processes?

Ayo: My advice is to start with clear and detailed documentation in contracts and purchase orders. The more clearly defined these are from the beginning, the easier it is to process invoices down the line. Implement a rigorous verification process, with digital time sheets and service delivery notes wherever possible. Also, leverage AI to automate tasks and enhance accuracy. Regular reviews and updates of the process are crucial to identify recurring errors and work to mistake-proof them.

Emily: One last question: how do you see the future of T&M invoice matching evolving with advancements in technology?

Ayo: The future is bright with AI leading the way. I expect to see more advanced integration of AI in T&M invoice processes. AI will continue enhancing accuracy and efficiency while getting smarter at identifying acceptable exceptions. Over time, AI will learn from past data, making it even better at tailoring solutions for specific industries. For sectors like construction, where T&M contracts are common, AI will adapt to industry-specific needs and deliver more finely-tuned solutions.

Emily: Thank you so much, Ayo, for sharing your insights today. It’s been a pleasure exploring these complexities with you.

Ayo: Thank you, Emily. It’s always a pleasure.

Detecting Anomalies and Frauds through AI-based Matching

Find out interesting insights with Anna Tiomina CFO & Founder, Blend2Balance.

Moderated by Emily Digital Transformation Consultant at Hyperbots

Don’t want to watch a video? Read the interview transcript below.

Emily: Hey, everyone, this is Emily, and I am a digital transformation consultant at hyperbots today on the call with me. I’m extremely happy to have Anna. Anna Tiomina is the founder of Blend to Balance Llc, with over a decade of experience in senior finance roles. Anna is also the leader of AI innovators in finance and beyond a community dedicated to merging tech innovations with traditional finance.

Emily: Really happy to have you, Anna.

Anna: Thanks for having me.

Emily: Excited to hear your insights on how AI is revolutionizing the detection of fraud and anomalies in vendor payment. So let’s dive right in as a CFO. Anna, how do you see the importance of detecting fraud and anomalies in vendor payments?

Anna: Yeah. So detecting fraud and ensuring compliance is one of the key functions of financial operations. In recent years we see much more fraud in this area. So it is challenging for CFOs to keep up with the technology and to be always ready to react to the new ways. The fraud actors are trying to reach the companies. So it is a really critical and very challenging task nowadays.

Emily: Correct. So, Anna, what are some of the most common types of fraud and anomalies you’ve seen or heard in vendor payments.

Anna: Yeah, so there are simple ones like replacing the bank account number on an invoice with a fraudulent bank account number. It can be as simple as sending the invoice that was never approved for payment, and the service or product was never received. It can also be inflated invoices, either through the pricing or through the quantities of products or services on the invoices or  for example, demanding a payment before the product or service was delivered or received. So these are the most common risks.Listed vendors, and sometimes, if the company has some processes in place, the requester. The payment requester splits the invoice to get to a lower approval level. Sometimes it is a fraud, and sometimes it’s just a mistake or an attempt to speed things up so it can really vary. But each of these events represents a risk.The funds that the company is mistakenly or fraudulent to send cannot be revoked in many cases. So I mean, that’s a huge thing for the company to be able to keep things in order in this space.

Emily: And how can these frauds and anomalies be prevented with traditional methods per se?

Anna: Well, so what companies do they implement internal controls? They implement processes. They implement things like freeway mention. Right? So you have a PO that has to be approved. Then you have an invoice that has to be made to the Po, and then the invoice has to be approved on a different approval. Flow? Some companies don’t have PO’S. They have things like for ice principle, for example, like no payment, gets released until at least 2 people take a look at that. So you need 2 approvers on any payment going out. Also, companies try to make sure that the processes and these rules are followed. So there are things like internal audits. Kind of like processes that make sure that the forest principle is followed. For example, right, or the person that is sending the fund does the verification of the vendor before the funds are being sent.Still, all of this is very time consuming. This is prone to human error. This involves a lot of people in the company, and it doesn’t guarantee that the fraud or mistake doesn’t happen.

Emily: Got it. Got it So, Anna, how does AI enhance the detection of these frauds and anomalies compared to the traditional methods.

Anna: Oh, yeah. So the way I look at that is that AI is another pair of hands or another pair of eyes on your team that doesn’t get biased. That doesn’t get tired. That doesn’t cheat right? So this is another level of control that really helps to correct the bias that your team might have or spot the patterns that can be missed by humans. So AI, for example, can detect subtle differences in the invoices or flag, the duplicated invoices or spot the difference between the original purchase order and the invoice or compare the current, invoice with the historical data, and make sure that the mistakes that happened in the past don’t go forward into the future. So, having an AI complement, your team is a really really helpful tool.

Emily: Got it understood. And can you please help me understand, you know how AI can help prevent duplicate payments and overcharging and vendor invoices?

Anna: Yeah. So that’s a great question, because duplicate payments are a little bit hard to catch, because the invoice looks correct, right? And unless you have really really good controls in place. You might miss that. This is the duplicate invoice, and paid twice, either by mistake or as a result of fraud. So AI can compare the invoices and identify these duplicated invoices better than the humans can do. Also, AI is really great at comparing the invoice against the PO. If the company has a PO or the contract, or the sow, making sure that the vendor hasn’t overcharged the company, and that the agreed terms in the original documents are followed. This can be a very time consuming task for people to find the correct document to find the right line on this document, etc. So AI really shines here, and it saves a lot of time and effort for the human team.

Emily: That’s pretty incredible. So, Anna, what role does AI play in ensuring that payments are made only after you know the goods or services are received?

Anna: Well in a classic case, right, You would have a 3 way match process, and the Requester would have to push some button in your software that you’re using to confirm that the goods for services have been received sometimes. This process doesn’t work. Sometimes the requester wants to really push the payment forward to enhance their relationship with the vendor, or like for other reasons. So AI can really track the history of relationship with the vendor and also make sure that I mean, if there was a certain amount of time where the service was expected to be delivered, that these all timelines are followed. So again, this is another level of control, another pair of eyes that can be really helpful.

Emily: Got it and how does AI help in detecting payments made to unlisted or fraudulent vendors per se?

Anna: Well, yeah, that’s an excellent question, because like, usually, the companies have some controls in place to make sure that the payment doesn’t go to an unlisted vendor, and you will have to add each new vendor manually. But what sometimes happens is that you have, for example, an improved vendor. But then the invoice kind of duplicates the name, but that has other payment details. And this is how you send the funds to the wrong account. Number right, so AI can help verify that the original payment details are corresponding to what you received from your vendor when the vendor was listed and approved. Also there are public databases of fraudulent vendors. So AI is great to, you know, to be tasked with monitoring these databases and flagging.You know the fact that you might be paid to someone who is not on your approved vendor database. So again, because there is such a rise of fraud nowadays it is very difficult to keep track of everything that’s going on. So having technology as a compliment, your team is really really helpful and increases the team efficiency, too.

Emily: Got it and just to summarize everything, Anna, you know, looking ahead, how do you see the role of AI evolving in the area of vendor payment management?

Anna: Yeah. So  I think that now we see just the beginning of AI complementing the Ap teams, I think that it’s gonna be used more and more, because this is just so efficient and handy, and also honestly, because the fraud actors are using a lot of AI. It is impossible to really like, offset this effort without having technology on your side too. So it’s like a race of technologies in a way. And I think that at some point it’s gonna be like a must have for the companies to have some level of AI in internal fraud. Detection process.Not immediately. But we are getting there.

Emily: Definitely. Thank you so much, Anna, for joining us today and talking to us about, you know such an important topic. Thank you for the valuable insight.  It’s clear that AI is transforming the way we manage and protect our financial operations, and how, especially in the realm of vendor payments. So thank you once again.

Anna: Thank you for having me.

Evaluating Bot Security in Financial Process Automation

Financial process automation is the use of artificial intelligence (AI) to perform various tasks that would otherwise require human intervention, such as data entry, invoice processing, reconciliation, reporting and more. By automating these tasks, businesses can save time, reduce errors, improve efficiency and enhance customer satisfaction.

However, automation also comes with its own set of challenges and risks, especially when it comes to security. The bots that execute the tasks on behalf of or assuming the role of a human user need to be carefully designed, monitored and controlled. A SaaS-based automation solution, must implement a zero-trust environment, where the bots are also treated just like human users, for the very reason that the bots assume the role of a human user for executing the tasks.

What is zero-trust security?

Zero-trust security is a principle that assumes that no entity, whether internal or external, is trustworthy by default. It requires verifying the identity and permissions of every user and device before granting access to any resource or data. It also requires monitoring and auditing all activities and transactions to detect and prevent any malicious or unauthorized behavior.

Zero-trust security is especially important for financial process automation, as it involves sensitive and confidential data that needs to be protected from cyber attacks, data breaches, fraud and compliance violations. By applying zero-trust security, the bots are provided with just enough permissions to perform their tasks, and that they are not compromised or misused by hackers or rogue employees.

How zero-trust security principles help secure the bots?

Here are a few ways in which zero-trust security principles help secure the bots in financial process automation:

Using strong authentication and authorization mechanisms for the bots. The automation platform must verify the identity and permissions of the bots before allowing them to access any resource or data. The platform must identify a bot executing tasks for a customer organization from other bots executing tasks for different customer organizations. This is very critical in case of Multi-Tenant SaaS based models. 

Implement least-privilege principle for your bots. This means that the bots are granted only the minimum level of access and permissions that they need to perform their tasks, and nothing more. This way, the bots are prevented from accessing data that is beyond the permissible boundaries and also limit the potential damage that a compromised or misused bot can cause.

Track and audit various activities of the bots. It is very critical to log and continuously monitor all the actions and transactions that the bots perform, such as what data they access, modify or delete, what systems they interact with, what errors or exceptions they encounter and so on. These logs need to be reviewed regularly using analytics tools to identify anomalies and suspicious patterns that may indicate a security breach or a compliance violation.

Conclusion

Organizations that look to optimize their financial processes through AI-driven SaaS automation solutions should evaluate the solutions paying special attention to the security aspects governing bots, and on how their organization’s data and critical digital assets are secured using security principles such as zero-trust.

Fortifying Financial Data: A CFO’s Guide to Safeguarding in the AI Era

In the rapidly advancing landscape of finance, the integration of Artificial Intelligence (AI) has ushered in unprecedented efficiencies and insights. As Chief Financial Officers (CFOs), your role not only involves steering financial strategy but also safeguarding the invaluable asset that is financial data. In the age of AI, where data is both currency and vulnerability, understanding and implementing robust security measures is paramount. This blog serves as an outline to fortifying financial data against the evolving challenges of the AI era.

The intersection of finance and AI

The marriage of finance and AI has brought about transformative changes, streamlining processes, and enhancing decision-making capabilities. However, the reliance on AI also necessitates a comprehensive approach to data security ensuring privacy of the accounting and financial assets of an enterprise. Here are key strategies for CFOs and their teams to safeguard financial data in the age of AI:

1. Encryption as the first line of defense

One cannot overemphasize the importance of encryption in securing financial data. Implementing end-to-end encryption ensures that sensitive information remains indecipherable both in transit and at rest. Explore advanced encryption methods, such as homomorphic encryption, to enable secure processing without compromising data confidentiality. This directly maps to the regulatory compliances available to vet and test software and SaaS-based offerings in this space.

2. Access controls: Restricting access, mitigating risks

Robust access controls are pivotal in preventing unauthorized access to financial data. Utilize Role-Based Access Control (RBAC) to align data access privileges with job roles. This not only minimizes the risk of internal threats but also ensures that employees access only the data essential for their responsibilities.

3. Continuous monitoring and anomaly detection

Embrace AI-driven continuous monitoring to detect anomalies in real-time. Behavioral analytics, powered by AI algorithms, establish normal user patterns and promptly flag any deviations. Early detection is key to mitigating potential security threats before they escalate. Prefer tools that provide dashboards, alerts, and logging mechanisms to allow deep observability of the functionalities. 

4. Explainable AI (XAI): Trust and transparency

In an era where AI models often operate as black boxes, prioritize solutions and products that offer explainability and transparency towards product capabilities as well as a clear reason and interpretability of any processed output that may be visible. Understanding how AI algorithms reach decisions fosters trust, and accountability, and aligns with regulatory requirements. Ensure that the financial insights derived from AI are not only accurate but also comprehensible.

5. Secure data sharing practices

Tokenization-based approaches emerge as a powerful strategy when sharing financial data externally. By replacing sensitive information with tokens, even if intercepted, the data remains meaningless without the corresponding tokenization key. These strategies include Masking and Anonymization tools, Redaction policies and only sharing the data post-removal of this information. Additionally, deploy secure APIs for data exchange, ensuring the integrity and confidentiality of financial information.

6. Cybersecurity training: Empowering your team

Invest in comprehensive cybersecurity training programs for your finance team. Educate them on AI-specific cybersecurity risks and instill a culture of awareness. A well-informed team is your first line of defense against evolving cyber threats.

7. Incident response planning: Preparedness is key

Develop and regularly update an incident response plan tailored to AI-related security incidents. Ensure that your team is equipped with clear procedures for identifying, containing, eradicating, recovering, and learning from security events. Preparedness is your best defense against unforeseen challenges.

Navigating the future of finance with confidence

As CFOs navigating the dynamic landscape of finance, embracing the power of AI comes with a concurrent responsibility to safeguard the integrity and confidentiality of financial data. By implementing robust encryption, enforcing stringent access controls, leveraging AI for continuous monitoring, and fostering a culture of cybersecurity awareness, you are not only fortifying your organization against evolving threats but also positioning it at the forefront of the AI-driven future.

At Hyprbots, we understand the paramount importance of data security in the financial realm. Our cutting-edge solutions not only harness the power of AI for financial optimization but also prioritize the highest standards of data protection. Together, let’s navigate the future with confidence, ensuring that the transformative potential of AI in finance is realized securely and responsibly.

Securing Finance Data blog Series: This blog is an introductory piece towards blogs around finance data security. We will publish a weekly blog detailing various technical as well as user aspects on this topic.