The Value AI Brings to Performance Management
For several years now performance management software vendors have been talking about AI (artificial intelligence) and more specifically machine learning. While of interest to technologists, it hasn’t been obvious why those in Finance, often the primary users of performance management, should care. It has now become obvious. Whether it is due to the vendors improving their implementation of the technology to deliver real value, or simply cleaning up their messaging, Finance is now sitting up and taking notice.
The first AI-based capability gaining traction is predictive analytics. Put simply this capability improves the accuracy of forecasts, a key element of most performance management implementations. By utilizing machine learning (specifically deep learning) the system can generate a forecast based on historical data. Users can then tweak the forecast based on their own read of where the business is headed. Those reluctant to give up that much control of forecast creation can still benefit from predictive analytics. They can generate their own forecast and then have the system assess the probability of their forecast coming to pass. Some may say that predictive capabilities pre-date modern AI technologies. That is true, but what has changed is the accuracy. Earlier versions of predictive analytics used straightforward statistical analysis of the data to produce predicted outcomes. While the results may have been reasonable at a high level, they missed some of the details. For example in a business where the numbers go up and down on a seasonal basis older predictive capabilities might have been able to predict where the business would end up, but might have smoothed over the period to period fluctuations (see illustration below). While purchasers of performance management systems today aren’t saying ‘We need AI’ they are adding predictive analytics to their key requirements list perhaps without realizing that it is often powered by AI.
The next group of AI capabilities are at an earlier stage both in terms of market demand/acceptance and performance management product delivery. They include NLP (natural language processing), AD (anomaly detection), and RPA (robotic process automation).
Natural Language Processing
Natural language processing is gaining acceptance in the consumer market with devices such as the Amazon Echo, but what about in business applications? We are not there yet. Numerous vendors however are moving forward with support for natural language queries. The ability to say to your performance management system ‘Show me the sales numbers for Europe for July’ should be quite compelling. After all, the current alternative is to select options from multiple pull-down menus or in some systems to create a scripted query to retrieve the values. The reluctance seems to be to sitting there and talking to your computer at work. Perhaps the new Surface headphones will make that process easier. What people are missing though is that instead of saying it you can type your request in to a search-like box using business terms as opposed to writing a technical query. In terms of making performance data readily available to a larger group of employees natural-language queries have huge potential. Natural-language generation is another important element of NLP that can provide value. In particular, for the creation of narrative summaries that usually accompany the numbers (‘Sales this month of $ 1,360,000 were down 10% from last year at this time’) much time can be saved. Today creating these narratives is often a manual process with the potential for errors. With AI the system can automatically retrieve the sales data, calculate the variance from a prior period, and determine that sales are ‘down’ and not ‘up’.
Anomaly detection may be less visible than predictive analytics and NLP, but it may be more valuable. This capability should be able to dramatically improve the quality of the data, which is key to a performance management system designed to provide one version of the truth that management can rely on. As the name implies this functionality spots data that is outside the expected normal range. In performance management where a significant amount of transactional source data is bulk-loaded into the system for reporting purposes this capability can both speed the process and improve the quality of the result. For example, if a particular cost center’s data for a certain account has been around 1,000 per month for many months and all of a sudden this month it comes over as 1,000,000 it will be flagged as an anomaly. More than likely someone incorrectly set the scaling factor in the mapping table. If this wasn’t detected and the error got buried in a consolidated roll-up the company could very easily be reporting inaccurate numbers. The same kind of anomaly detection can also work on numbers being keyed in directly. This capability is obviously critical. While some vendors do offer data quality functionality today, this AI feature will make it easier for others to join in while adding an additional layer of checks to existing solutions.
Robotic Process Automation
Robotic process automation seems to be lower on the delivery list for performance management vendors. Yet it would seem to be a big win in terms of making a system more intuitive, easier to use, and therefore ready to be rolled out to larger groups of users. What RPA does is automate the steps in a repetitive multi-step process. So, for example if every month you need to load data, consolidate, run some reports, and notify a list of people that the reports are ready the system can do this all for you. It saves time but it also prevents you from inadvertently missing a step (for example forgetting to re-consolidate after loading the data and therefore producing reports with old data). More simply, a modified version of RPA can help you navigate through the system. If you are a casual user of the system who only enters a budget once a year you may forget the steps the next time the budget rolls around. With this functionality the system can prompt you to proceed to appropriate next steps based on your prior usage pattern. So, after finishing your budget it can suggest you lock and submit your data, print a copy for your records, and notify your manager it is waiting for approval. Again, some systems do a version of this today but with AI it will become more accurate, more adaptable to your particular needs, and more widely available throughout the system.
Every key vendor that we follow in the performance management space has either delivered AI capabilities or has them under development with plans for an initial release in the near future. To see which vendors have delivered AI, and more specifically which of the particular features described above, check out our Vendor Snapshots and Vendor Landscape Matrix.