Where we go from here: 7 absolutely critical investments for building the AI platform for modern finance teams

by Anant Kale September 9, 2019

Today we announced that we raised a $50 million series C round of funding, bringing our total investment to more than $100 million. We’re excited about this milestone because of what it says about our market execution and customers’ enthusiasm for AI-powered spend auditing. We also surpassed another important milestone – crossing over the $50B mark for total spend audited in our platform – demonstrating maturity of our offering and finance teams’ increasing comfort with AI as they deploy AppZen in their operations globally.  

For many companies, raising a series C is about expansion within an existing market. Certainly, AI-powered spend auditing offers an awesome opportunity for AppZen, and we are pursuing it with gusto. But we have our eye on a slightly loftier remit: to be the AI platform for modern finance teams. We’re building on what we’ve learned by delivering the leading spend audit solution – from extracting and reading financial documents to constructing a robust semantic model of all-things-finance to being in the approval workflow for billions of spend transactions – to help CFOs and their teams make smart decisions across spend, revenue, cash, treasury, and beyond. In short, we want to be your secret weapon across finance.

At the same time, we’re sober-eyed about what it takes to pull off a vision like this. Without a solid foundation, things can get rickety in a hurry. We also know that the series C often denotes that stage in a startup’s life when managers start making tough calls about whether to head for a fast, convenient exit or build for the long term. Finally, our #1 cultural tenet is “obsess about customers,” guiding literally every decision we make. With these points in mind, we want to share our plans for the series C funds and be very clear about how these initiatives will benefit our customers.

1. Understanding finance

If we’re going to serve the needs across a large swath of finance, we need to speak the language. We will continue to build the world’s most expansive semantic model for enterprise finance so we can understand and organize concepts, terms, transactions, and data; define and enrich them; and learn how they relate to one another. This means honing our competency in natural language processing; parsing and understanding every aspect of complex documents such as contracts, receipts, POs, and quotes; and building robust models for how transactions relate to the general ledger and how those accounts relate to end-of-period financial reporting.

2. Connecting finance with the company (and beyond!)

Financial transactions like expenses, payables, billings, and revenue are closely interrelated with the rest of the activities in a company, from hiring to customer service tickets to sales forecasts, and even to external signals and environmental factors. It seems obvious, yet today’s financial systems seem to be a black box and inward-looking. Financial analysts cobble different data sets and analytics in an attempt to spot anomalies and trends, creating a lot of data “noise” but getting very little for it. Even then they are left with an incomplete picture, as a majority of the data that need to be analyzed are captured in documents like contracts, quotes, and receipts. Our aim is not only to build an AI platform for finance that understands finance semantically from both structured and unstructured data, but also to the rest of the organization (and beyond!) in as extensible a way possible. This means ingesting intelligence from system logs, badge swipes, messages and emails, sensor data, business applications, pricing information, weather signals, financial statistics, merchant databases, regulatory sources, and more, and then making sense of that intelligence in our AI models to identify risks, see opportunities, and make decisions, all in real time.

3. Building in agility

Every CFO is chartered with creating a nimble organization that can respond immediately to rapidly changing business needs. We think the big needle-mover here is to make our AI platform as vendor agnostic – in fact, as vendor resilient – as possible across both applications and software. If we can provide you with a platform that has robust integrations and interfaces built in and we can plug into any enterprise application or connect even with your file shares to extract and understand your documents, that will help you bootstrap regardless of where (or how spread out) your business data are. What does that mean to you? It means you can lean forward into AI-powered process transformation without having to wait for expensive, time-consuming application changes or upgrades.

4. Creating taxonomies

One way to bring this investment to life is to instantiate our semantic model in a set of customer-centric taxonomies. We’ve begun to create a comprehensive spend taxonomy to help our audiences organize and connect the dots across their spend based on their respective roles and functions. This taxonomy will be the first in our industry that allows users to classify spend based on their own particular vantage points – procurement, accounts payable, accounting, tax, and executive. Beyond providing key enterprise finance personas with a way to sort their spend in a way that’s useful to them, creating a flexible taxonomy will enable enterprises to calculate taxes more accurately and efficiently, correct reporting errors before they are finalized, and even be more predictive in financial planning and analysis. Spend is just the first step! We’re using the taxonomy process as a way to develop a competency in organizing and connecting information. Cutting our teeth on spend classifications will allow us to move faster and be more thoughtful as we organize other areas, including revenue recognition, treasury, foreign exchange, and beyond.

5. Rallying around regulatory compliance

We didn’t expect that regulatory compliance would be such a big driver for our business, but it turns out that AI gives organizations a huge leg up in monitoring all of their spend for regulatory violations. What started out as a perfunctory, “checkbox” requirement has become a major focus area for our engineering and data science teams as we both witnessed a sharp increase in regulatory crackdowns and penalties and realized that the core features in our platform were relevant and useful. Capabilities like our spend taxonomy, anomaly detection, and Star MatchTM methodology of validating spend work together to boost organizations’ chances of detecting regulatory violations many-fold over even the most rigorous manual auditing regimen. While we continue to add regulatory sources to our “lookup list” of politically-exposed persons, foreign officials, and healthcare providers for people and organizations named in expense reports, invoices, and contracts, we are building expertise and enhancing our platform to really rally around regulatory compliance. Six initial areas of focus are anti-bribery and corruption, export controls, sanctions, debarments, anti-money laundering, and healthcare spend transparency.

6. Learning within and across companies

One thing that separates an AI platform from a rules-based tool is the former’s ability to learn and hone its models based on user input, which results in accuracy improvements over time. Over the past year or so we’ve been developing a robust predictive model not unlike that of a meteorologist to predict an organization’s spend quantities and amounts, taking into consideration thousands of inputs including employee growth, mergers, seasonality, external factors, corporate idiosyncrasies, and more. Using it in conjunction with simple anomalies such as invoice duplicates has improved our accuracy by an order of magnitude, allowing our customers to gain far more precision while not having to trade off recall. Put simply, they can detect the most important risky transactions and all of the risky transactions at once. Similar to how we’re approaching taxonomies, honing our anomaly detection capability helps us develop the meta. Indeed, we’re developing a competency around anomaly detection.

7. Being user-obsessed

Harkening to our #1 cultural tenet, “obsess about customers,” we are matching our development of powerful AI features with as much or more attention paid to the user experience for all of our customer personas. Unlike other AI vendors that focus on a particular piece of the value chain, we take a system-wide approach, which helps us identify opportunities to automate process, skip steps, streamline workflows, or simply make the user experience more delightful. One example is how we built Expense Audit, the first product on our platform. Rather than bolt on a post-payment “recovery audit” of sorts, we took a look at the entire expense report workflow and found that not only do managers hate approving expenses, but they’re also notoriously bad at it. So we removed it. We took it out of the process, giving finance the ability to push post-reimbursement reports to managers so they could know how compliant their teams were without having to trade off being compliant and getting people paid back quickly. By addressing the entire workflow, we gave organizations’ managers hours back each month, helped our customers achieve more consistency in their audits, and dramatically reduced the average reimbursement window from more than a week to 1-2 days for nearly all of our customers – to the delight of their employees! This experience helped us develop an important muscle in our product development process: Look at our customers’ problems from a system perspective, and challenge the status quo. We have taken a similar approach in the accounts payables workflows and will continue to push the boundaries on user experience with each new offering.

For us, raising a series C isn’t just about the milestone. It’s about making critical investments in seven key areas — understanding finance, connecting to the business, being agile, building taxonomies, rallying around regulations, learning, and obsessing about users – not only to help enterprises reduce spend, comply with policy, and streamline process, but also to make life just a little more delightful for CFOs and their teams.

Anant Kale

Founder & CEO at AppZen. Anant dreams up the next use of Artificial Intelligence for enterprise automation