Human-in-the-Loop Compensation: Why You Don’t Want Fully Automated Sales Compensation

Should compensation be fully automated? This question pops up every now and then as sales comp systems become more automated and embrace AI for optimization. Determinations like should someone make quota, what incentive changes should we make, and even entire plan building can be forecasted and simulated in real-time. But while the tech exists today for fully automated incentives, is it smart? As with any incentive program decision, automating compensation plans comes with risk. Incentives are tied to financial exposure, seller behavior, and organizational alignment. Make the wrong decision with automation (turn on accelerators without considering margin pressures) and you could be on the hook for massive payouts or left chasing unprofitable revenue growth. AI also still lacks context. Sure it can identify patterns and opportunities but what if your strategy has changed? What about new product go-to-market considerations or competitive intelligence? What if your CEO cares about one metric more than another? Lastly, there are compliance and governance implications of letting a computer set compensation plans. Are they auditable? Are they explainable? What about reconciliation to financial systems? Without visibility into how the computer got to its recommendation, many organizations would be forfeiting control and accountability. The question isn’t if you should use AI, but how much control are you willing to relinquish?

So what is human-in-the-loop compensation? Human-in-the-loop is a term used to describe a process that allows for AI to automate tasks, but ultimately leaves the decision to humans. So in human-in-the-loop comp, the system may automate hunting for behavior changes, predicting commissions and identifying plan performance drivers, but a leader approves all changes. Let’s break that down… 

Human-less loop compensation includes AI tasks such as behavior analysis, pattern identification, predicting commissions, plan performance drivers, and recommending changes based on historical trends. For instance, systems like SAP Commissions and Xactly Incent have the ability to create real-time what-if commissions dashboards that allow companies to see how performers are doing against their targets and how effective the incentive plan is driving that performance. With human-in-the-loop compensation, someone from leadership approves all of these changes. Leaders use AI to quickly analyze what is going on and make decisions, faster.

Decisions Layers: In Control and Setting the Right Level of Automation

So how do you establish human-in-the-loop compensation? First we must define what layers of decisions need to be put in place. For us, there are four layers: 

Layer 1: Insight– What’s happening? This includes the what-if commissions dashboards mentioned above.

Layer 2: Prediction– What could happen? Machine learning can identify potential performance issues and predict future commissions.

Layer 3: Recommendation– What should we do about it? AI can recommend changes to commissions rates, adding in ramp programs, etc.

Layer 4: Decision– Only a human can decide if the change needs to happen. Only leaders can apply the “so what” factor to decide what to do with that recommendation.

The better you understand each of these layers, the easier it is to implement AI and build trust. Trust is a major factor when it comes to AI. You don’t want your CEO making commission adjustments based on insights provided by a tool that your Finance team knows nothing about. Transparency is king with AI. Leaders need to know how the AI came up with the recommendation and what data it used. Having a tool that allows you to simulate potential changes is another way to build trust. If your leadership doesn’t have to fully commit to a decision, they are more likely to test it. Finally, you want to ensure that there is an audit trail behind every decision.

Over time, as trust in the AI developed insights increases, an organization can turn up the amount of automation it wants.

Conclusion 

There will never be a fully autonomous sales compensation system. AI and machine learning can automate tasks, but don’t have the ability to put incentives in context like only a human can. Balancing automation and control is key to enabling your compensation system to make you incentives smarter, faster.