Originally published in NAVEX Global's Top 10 Ethics & Compliance Trends for 2019 eBook.
Many conferences, white papers and webinars have been dedicated to artificial intelligence (AI). Futurists, technologists, and innovators on the bleeding edge of technology have shared the seismic changes to the workplace the compliance industry should expect. The shift from linear growth to exponential growth and the elimination of administration has been on the tip of every tongue.
At first blush, it was exciting, it was educating. Now, however, the lack of clear steps for implementation has created a bit of anxiety, resulting in a state of analysis paralysis around AI adoption. We all know we need to get up to speed, but we don’t want to run the risk of outpacing ourselves or missing the mark on true ROI.
To calm any unnecessary anxieties around AI, it is important to understand that compliance professionals do not need to become AI experts. We are not technologists, and we shouldn’t be. Compliance professionals are experts on corporate culture, risk mitigation, and change management. It is that expertise that we need to apply to the adoption of AI solutions.
Instead of planning for the entirety of the singularity, reframing the conversation around AI can help compliance professionals stay focused on the things they can control.
Identify the Compliance Problems AI Can Solve
When considering AI solutions, we first must understand the key characteristics of AI and then match those with existing programmatic problems. Displaced efficiencies just create more work. This begins with understanding the tasks AI and similar cognitive computing technologies like machine learning are really good at, including:
- Automating manual data entry
- Filtering for errors or patterns
- Continual monitoring of regularly updated lists
- Predictive analytics
These trailheads point us down the path to several key considerations. First, regulatory changes are always in flux. AI that can keep us informed of real-time updates is key here. Furthermore, many changes required by regulatory updates often require a prescribed modification to a strand of text, policy or training across the organization. Automating regulatory update notifications and the necessary actions in response is administrative overhead AI could effectively streamline.
Hotline reporting data can be latently monitored to create key indicators of when and where certain behaviors are taking place, then trigger the appropriate changes to policy and procedure management and compliance training rollouts.
Similar to the moving target of regulatory compliance, third-party risk requires continuous monitoring of third-party partners as well as sanctions screenings. The ultimate strategy behind engaging new partners will always be up to third-party risk management leaders, but automated screening and due diligence can be effective information catalysts for enhancing or revising strategies as needed.
Pattern detection is also a key perk of AI. For compliance, this can be instrumental for identifying employee behavior trends. Hotline reporting data can be latently monitored to create key indicators of when and where certain behaviors are taking place. This information can then trigger the appropriate changes to policy and procedure management and compliance training rollouts.
Adopt Technology while Remaining an Employee Advocate
Because of the robustness of AI solutions, errors (much like the benefits) can proliferate exponentially and instantly. The most noticeable concern for compliance is currently unconscious bias – that of the engineers who developed the technology, the program administrators integrating the technology, or data analysts processing the patterns and results the technology returns.
Many organizations have employee demographics that are disproportionately something – one characteristic or another dominates over others. Fed into artificial intelligence software, these characteristics become the norm and, when left unchecked, can return false positives for abnormalities.
Compliance professionals have the expertise to understand the human side of the workplace, and therefore need to stay vigilant in selecting the right AI, and scrupulous on how that technology is implemented.
For instance, an AI program might determine that on a given team, in a given year, female employees used more paid time off (PTO) than their male counterparts. Fed into another AI program for hiring, this data teaches the AI to prioritize male applicants. What the software misses is that on that given team, in that given year, two female employees used their PTO to supplement their maternity leave. The next year, this may be mirrored by the team’s paternity leave.
This is not an error with the technology. It is an oversight in the data the technology is being fed. Compliance professionals have the expertise to understand the human side of the workplace, and therefore need to stay vigilant in selecting the right AI, and scrupulous on how that technology is implemented.
Key Steps for Organizations to Take
Create a Framework for Evaluating the ROI of Your Solution
The price tags for AI solutions are at a premium. Before making your purchase, you must be certain the cost savings will ultimately outweigh the investment. Cost is currently high because many artificial intelligence solutions need to be customized for each company. This comes with a pilot, implementation, and improvement phases, each requiring significant time and budget to accomplish. To confirm you are purchasing the right solution, and to have a clear roadmap for measuring success, develop an evaluation framework. In the simplest of frameworks, AI solutions should check three boxes: it’s reliable; it’s real-time; and it creates a single source of truth. Ask your potential vendors how they can deliver on these requirements.
Strive for Digital Harmony
New technology is rarely successful when bolted on to existing tech stacks. Instead, AI solutions should integrate fully into existing compliance and enterprise-wide solutions. This creates ROI that extends further than just the compliance department. You especially do not want your new technology to create challenges or inefficiencies for other teams.
Make Sure “Garbage” Cannot Describe Any of Your Data
“Garbage in, garbage out,” is the pivotal phrase when it comes to AI. For artificial intelligence to truly act intelligently and for machine learning to actually learn, it needs the most accurate raw data. When implementing a new AI solution, do extensive due diligence on the data being processed as well as monitor the results to ensure inaccuracies are in no way driving your business decisions.