The CFO is in a unique position to bring AI to the enterprise.
Customer data and predictive pricing
In many firms, the CFO has responsibility for the reporting layer of data that comes in from customer-facing operations – for example, receivables passed from the sales organization or pricing data from point-of-sale units. In today’s world of digital commerce, this puts the CFO in a powerful position to link predictive analytics with customer behavior.
A controller’s office joins forces with data scientists to run correlations between price behavior and receivables. It then overlays a set of transaction data such as the type of product purchased, payment method, and customer demographics. Just to make it interesting, they add third party data on inputs such as weather and location at the point of purchase.
The result – a projection of the optimal pricing for an age 18 to 24-year-old female customer who is using credit cards to buy gifts for the holidays when the weather is poor.
Example – the holiday war room
A major US retailer combined data from points of sales, inventory, and pricing to create a “war room” for the holiday season. Using AI analytics, the finance team worked with the retail outlets to create dynamic to-the-minute pricing that maximized the balance between revenues and unit sales.
Most of us must have seen applications of this type of predictive analytics, for example, when receiving movie or video recommendations based on content we have viewed in the past.
These same approaches can even help CFOs maintain optimal inventory levels, avoiding inventory write-downs and saving working capital.
Looking beyond the book value
One of the greatest challenges for a CFO can be in assessing the true value of assets. Uncertainties in valuation, whether in acquiring an asset-intensive company or determining the tax basis of current assets, can add or subtract millions from the bottom line.
The most effective method for determining asset values is an assessment of a large number of comparable independent transactions. This is where AI can help.
CFOs and valuation specialists in the real estate business can use AI to assess thousands of housing variables – mortgage rates, quality of schools, number of bedrooms, local employment – to build predictive models of home asset prices. These can be for acquisition or for sales by the enterprise. This not only helps the real estate firm, but also provides value to buyers, sellers and lenders.
Example – instant valuation model
A Dutch based startup uses an Automated Valuation Model (AVM) that leverages AI algorithms to perform property appraisals instantly, unlike the traditional model which can take weeks. CFOs leverage such AVMs to estimate asset values at fair/market prices to determine financial performance of their own company and companies they perform due diligence on.
Predicting deadbeats: forecasting and management of bad debt
According to the US tax authorities1, bad debt accounts for 0.5% of US firms’ revenues. In 2018 that was over US$100 billion in missing money, reducing profit margins by as much as 5%2.
Artificial intelligence puts the CFO in a position to predict which customers will pay, be late in paying, or will not pay at all. A multivariate analysis of B2B customer data such as industry type, credit rating, product purchase and sales person can provide a forecast of the probability that a business will pay its bills – and should be extended credit. Alternatively, identification of likely non-payers helps in customer qualification and credit approvals.
Example – the rogue salesperson
A financial services firm was plagued with higher than-average mortgage defaults. The head of sales and her team tried to find the patterns behind non-payment – including home type, down payment, and credit standing – with no result. Finally, a member of the finance team built a predictive model that included data from salesforce compensation and found the highest correlation was loans approved by certain lending officers.
A model was built that used machine learning to understand the ongoing relationships between certain compensation structures and bad debt, providing a long-term forecast of defaults under different scenarios.
The amount of bad debt impacting US firms in 2018.
Embezzlement and expense fraud
Internal fraud is particularly hard to detect, predict and control. It is episodic – not leaving a clear data trail. It is often executed in small increments that escape detection. Finally, the perpetrator may intentionally distort the data trail to prevent detection.
Expense fraud alone is projected to cost companies US$1.8 billion per year3. Dealing with this risk, which can have significant legal, tax, and financial implications, can distract CFOs from more strategic issues.
Enter artificial intelligence. With AI you can analyze and interpret expense data and detect suspicious expense claims. You can explore spending patterns and employee behaviors in various roles. Also, machine learning technology can identify and predict common behaviors of employees who falsify or exaggerate claims. This makes it possible for a CFO to forecast potential expense fraud before it happens.
Example – setting expense policy
A CFO’s team was able to review an historic set of proven expense abuse cases and captured data on expense type and vendors (for example, certain airlines or hotels). But rather than go after the employees, the finance function decided to provide an aggressive program of education and careful monitoring of areas of abuse. Effectively, AI let the team target the training on the abuses that were most likely to happen – thereby reducing future fraud in the workplace.
The projected annual cost of expense fraud to companies.
Artificial intelligence and the detection of money laundering
Given the seriousness of money-laundering offenses and the possibility of strong regulatory sanctions, many banks have installed alert mechanisms based on known patterns of abuse. Many of these alert programs are mandated by law. The problem is that these systems can generate a flood of alerts and leave the CFO’s team wondering where to put scarce resources.
Artificial intelligence can teach computers to recognize suspicious behavior and to classify alerts as high, medium or lower risk. Applying rules to these alert classifications can facilitate the automatic closing of false alerts, freeing up staff to focus on the small minority that have a strong possibility of an illicit transaction.
Example – ranking alerts
A European bank faced the challenge of a skyrocketing number of alerts and “false positives” in money laundering. The finance team created a predictive model for illicit transactions based on data from law enforcement on proven money laundering transactions.
Using an algorithm that classified alerts on a one-to-ten scale, scores above eight opened an immediate investigation and tens triggered a freeze in assets. The bank’s predictive modeling was cited by regulators as a standard for other banks in the fight against money laundering.
Taking the drudgery out of finance
Perhaps no part of the enterprise has as many repetitive, routine tasks as the finance department. Inputting invoices, tracking receivables and logging payment transactions are high cost, low return, and not of high interest to employees.
Artificial intelligence, combined with robotic process automation, has the power to disrupt the traditional finance back office. While the robotics speeds the transaction, the AI mines the data for insights for the front office: How do we speed collections? Where should we invest? What will be the impact if we decrease our prices?
CFOs are increasingly using AI to address significant changes to accounting regulations. For example, we have seen large companies save significant manpower by using natural language processing (NLP) for the review of lease contracts. Without AI, this would have been a very labor-intensive task.
Example – reducing costs and upskilling employees
A multi-billion-dollar apparel retailer found that its finance function accounted for over 3% of costs. In applying an RPA structure to receivable and payables, it was able to reduce these costs by 15% while harvesting data on financial incentives for the firm. Employees were not let go but were instead “reskilled” to conduct the analysis of the mined data, a remedy that cut costs and reduced cycle time.
The CFO and the power of prediction
Historically, the finance function has focused on documenting the past – recognizing revenues, auditing costs, or monitoring compliance. Artificial intelligence is transformative because it places the CFO in the future with the data-driven power of prediction. Suddenly Finance is empowered to forecast how competitors will react, how customers will respond, and where risks will emerge.
This transformation goes far beyond digital fortune-telling: it transforms the role of the CFO and the finance team, placing it in the strategic heart of the enterprise. The CFO who seizes the opportunities of artificial intelligence and machine learning will not just be transforming the enterprise, but also the scope, responsibilities, and power of the job itself.
An abbreviated version of this article was originally published in CFO Magazine.
Show article references#Hide article references
- US Corporate Income Tax Returns, Internal Revenue Service, 2008
- Margins by Sector, NY Stern School of Business, 2019
- Expense Fraud Projected to Cost $1.8 Billion Annually, National Society of Accounts for Cooperatives, 2018
- US Corporate Income Tax Returns, Internal Revenue Service, 2008
Artificial intelligence is one aspect of the broader Finance 4.0 journey that CFOs are taking. When coupled with other technologies, such as blockchain and intelligent automation, it will empower the finance function to make better business decisions than ever before.