Finance AI

Custom expense policies demand automated, but flexible, AI enforcement

by Uri Kogan January 14, 2021

Your organization is unique — and so are your expense policies. Expense leaders worldwide spend hours crafting finely honed policies that detail every aspect of what can and cannot be claimed by employees as out of pocket expenses. Yet applying those policies against masses of incoming expense reports is an even bigger task. 

Automation via technology has delivered a lot of value to Expense leaders and their teams but failed to date to help finance directors and spend controllers with enforcement of customer expense policies. Finance-specific artificial intelligence software offers the promise of massive scale, the ability to interrogate data to minute levels of detail, and the potential to remove the pain of expense processing completely. 

But why is enforcing expense policies such a challenge?

Every Organization is Unique

At the highest level, spend rules and policies are simple — employees should only spend what is appropriate when operating on the company’s behalf. 

But as with many things in business, the devil is in the details. The specific way each organization views employee expenses typically include variations to policies based on different geographies or roles, but can equally involve numerous and complex company-specific rules. Some of the most common custom variations are described below.

Project Specific

Many organizations define custom expense rules on a per project basis. For example, in professional services companies limits can be based on what client will reimburse for a particular project.  In universities and non-profits, grant funding may come with specific expense stipulations.

Work from Home

The increase is remote working due to the COVID pandemic has forced a large degree of policy rethink regarding valid expenses while working from home. Allowing employees to claim remote office suppliers and meals is common practice, but so is excluding parking fees on work-from-home days.

Company Approved Suppliers

Increasingly, organizations control expenses by requiring employees to purchase via approved supplier lists for travel, IT equipment and more. Expense auditors should closely scrutinize expense claims that include non-approved suppliers.

In addition, certain companies have lists of banned suppliers — for example a consumer brand may not allow expenses including purchases from a rival brand.

Custom Limits and Thresholds

Every organization has varying levels of leeway in terms of policy application. For example, if an employee cannot produce a receipt for an expense they can submit a missing affidavit form. However, there must be limits — for example, more than 10 missing affidavit reports in a 6 month period could flag a review.

Similarly many businesses impose an overall spend restriction on expense claims — perhaps $5k per report for junior staff, $10k for managers, and $20 for directors.

Travel expenses

Travel-related expenses are a Pandora’s box of complexity, with an unlimited number of variations. As a result many organizations are using corporate portals through which employees should book all travel spend. Any travel booked outside those portals should be flagged.

More intricate, connected claims also need validating. For example, a report that includes gas must also include a claim line for an associated car rental.

Cross report claims

And finally, one of the simplest ways for employees to abuse an expense system is to submit expenses for a group meal with other attendees, and for the other employees listed to do exactly the same in reverse. While simple to execute from the employee side, this is one of the most complicated policy violations to catch.

Controlling the Complex

There’s enormous value in using AI-driven automation to check straightforward expense reports. The best AI systems can handle smudgy receipts, different languages, split multiple receipts from the same image, and more. Most expense reports are fairly straightforward, and AI can help expense teams by eliminating time spent reviewing claims that aren’t problematic. But the examples above show that there are also many unique situations, policies, and requirements across different organizations that are impossible to automate with a best-practice approach, precisely because they are based on your specific practices and needs, rather than the common industry ones.

Leading expense departments have pushed their AI policy enforcement beyond automation of common best practice, tailoring their systems to meet their precise needs. 

Automated Policy Compliance

The step beyond best practice auditing to enabling automation of any custom set of policies within an organization has three key components. Each allows an advanced AI policy automation solution to embed domain and business-specific knowledge into every expense analysis it performs, ensuring that your business has the most accurate, customized, and well-informed policy automation capability possible.

Automated document understanding

Intelligent document understanding goes beyond simply extracting values from receipts. It applies existing knowledge to extract meaning from any type of financial document. So on a receipt it knows that Tom’s Bites is actually not a restaurant but a fishing store, what a completed missing receipt affidavit should look like, and recognizes a travel booking from a non-approved source. This advanced level of understanding is essential to automation.

Integrate business data into Models

In addition to the information extracted from documents, advanced AI automation systems integrate data from other core line-of-business systems such as HR, travel or project level expenses. These systems provide real-time information that feed the AI models and ensure accurate decisions.

Build custom taxonomies into Models

A taxonomy defines the language and terminology for your specific industry and business. With this embedded into expense processing, AI models can detect and account for expenses that may otherwise be treated incorrectly. For example, a spray gun is a perfectly acceptable expense for a car engineer but would likely be flagged as unacceptable by a non-domain enabled AI. 

By intelligently appling all policies against all incoming expense lines and reports, an organization can achieve what is known as 100% coverage. This level of completeness absolutely requires an artificial intelligence platform — it is simply too much work for manual processes, and too detailed and complex for non-AI expense solutions.

The Time To Automate Expense Processing is Now

The use of AI to deliver processing speed and 100% coverage is rapidly becoming table stakes in policy automation. But as we have discussed, very few organizations have simple expense policies — most have unique policies that are heavily-customized to their line of business and company needs. Without accurately and consistently applying those specifics, organizations end up with an expense audit solution that is automated and fast — but generates more issues than it solves.   

Your expense policies are necessarily complex and demand accurate and complete auditing — which cannot be done with vanilla tools. By applying document understanding, integrating other business data sources, and incorporating business-specific taxonomies, AI automation can be customized to the exact needs of an organization. Only with this level of completeness can AI automation deliver an efficient and scalable expense management process for your business. 

Your expense policies are ready for advanced automation and AppZen is able to satisfy that need. Are you ready to take the leap to the world of automated and customized expense management?

Uri Kogan

Vice President of Product Marketing