Skip to main content

Q&A: Transforming transactions with AI and machine learning

Artificial Intelligence
(Image credit: Image Credit: Razum / Shutterstock)

What are the core components of most transactions today?

There are multiple layers to transactions today, so it’s easiest to think about them in three ways:

1. The consumer-facing shopping cart
2. Backend processing and aggregation
3. Parties involved and physical location

When it comes to the shopping cart, there are four key pieces that always exist at checkout – items, payment, shipping, and tax. This is what the buyer sees, but on the backend it’s much more complex. 

On the backend, any number of things must happen to make a transaction possible, including payment collection, localization of currency, consumer financing, shipping costs, tax and tariff calculation, and more. Beyond the processing components, where transactional data is housed and aggregated must also be considered. For example, many transactions happen where the buyer makes a direct purchase from the seller, but then that data is aggregated in an ERP, accounting, or other invoicing systems. Any number of systems may be used to make a single transaction possible.

Lastly, it’s important to know who the buyer and seller are, and where the purchase is shipped from and to. These transactional components are essential to correct tax calculations, shipping estimates, and overall payment total. 

While most people think of the transaction solely in terms of the four pieces of the shopping cart, backend systems and transactional details are also critical components of every transaction. 

How has the evolution of e-commerce changed the need for technology at the point of transaction?

If you look at those four primary components of a transaction – product information, payment, shipping, and tax – you can easily see how e-commerce has increased the demand for technology at the point of transaction. 

Today, there are billions of products that exist in the world, and they’re all tracked digitally. Because products themselves serve as digital avatars, businesses need the right information readily available and correct to convert buyers, including things like price, product features, and customer reviews. As a result, all product information – not just characteristics like size, shape, and color – must be digitally available.

Historically, payments used to be a simple cash exchange or credit card transaction. Today, you have multiple payment types – installments, digital currency, money transfers, and more – all of which can happen with a single click. 

In its simplest form, shipping originally happened directly from the store to the end customer. Then the introduction of warehouses added layers to the shipping equation that businesses had to contend with. And now we have even more options, like drop-shipping, which have completely decentralized the shipping process and increased reliance on technology to execute shipments and manage them. 

On the tax front, many online transactions didn’t even include tax until recent legislative changes across the world implemented additional requirements for remote sales. In the U.S., the 2018 Supreme Court decision in South Dakota v. Wayfair, Inc. granted states the ability to impose sales tax obligations on remote sales. Today, all 45 states with sales tax have adopted some form of remote seller law. Earlier this year in the European Union, sweeping VAT reforms introduced new marketplace deemed supplier laws. Under these laws, marketplaces are now responsible for collecting and paying VAT on behalf of sellers for transactions valued below €150. Now, businesses selling online must calculate tax at the time of transaction based on the location of the customer in many jurisdictions. 

Ecommerce has sped up the entire transaction process, created more options for customers to engage in purchases, and spurred new government regulation – all of which means that transactions cannot happen without several systems working together. 

What transactional components can benefit the most from the application AI and machine learning?

There are several transactional components that can benefit from AI – essentially anywhere that there is a large dataset. 

At Avalara, we’ve applied it to product classification and tax code determinations. AI can help classify billions of products and correlate them with the correct tax codes based on product attributes and tax rules. 

There are other tax applications that benefit from AI. In the transaction dataset, AI can take buyer, seller, and shipping information to calculate consumer use taxes instead of B2C or B2B taxes. Another area for AI application is determining nexus – the obligation to collect and remit tax. By analyzing transactional information, AI can determine where obligations exist or where they could potentially exist based on transactional trends.

Applying AI to the management of tax rates and rules also helps increase the accuracy of transactions. With rules and rates changing every day and varying by jurisdiction, the same product can be taxed any number of different ways depending on the location or how it is prepared or provided to the end customer. For example, if you order a whole bagel in New York, it’s tax-exempt; have it sliced and it becomes taxable. AI can help make sense of these nuances and keep the data that powers tax decisions up to date for the most accurate calculation on transactions.

What are the keys to ensuring that transactions are accurate in our real-time world?

To get the most accurate transaction, the processing systems need to know who’s buying and selling, where product is being shipped from and to, what the product information is at a granular level, and the appropriate tax rates and rules. You need to know all this information because the accuracy of a transaction lies heavily in the tax calculation. By bringing product information, tax rules, and nexus requirements together, the tax on each transaction can be calculated in a more accurate way.

Additionally, given the real-time nature of commerce today, automation has become a requirement for accurate and on-time transactions. The need for automation was already happening before the Covid-19 pandemic as more commerce was already taking place online, but now the scale at which online transactions are happening has made automation synonymous with commerce and a requirement for processing payment, tax, and shipping. If you want accurate transactions, you must have automation applied throughout the transactional journey.

Sanjay Parthasarathy, chief product officer, Avalara

Sanjay oversees Avalara’s global product strategy as chief product officer. After an acquisition in February 2019, Sanjay joined Avalara from Indix, an artificial intelligence-based product information platform he founded.