Take Humans Out Of The AI Loop, And Put Them At The Helm

We’re making a mistaken assumption when it comes to artificial intelligence. We assume a human needs to be “in the loop” to oversee every process handed over to AI. Instead, just as a manager doesn’t double-check every piece of work from human workers, they should have broad oversight, setting an AI agent’s goals, limits, audit trails, token spend, and accountability structure.

That’s the word from Paula Goldman, chief ethical and humane use officer and EVP of product for Salesforce. The choice of what to hand an agent shapes results, cost and risk, said Goldman, author of Manage the Machine: How to Harness Human-AI Collaboration at Work, providing her take in a recent episode of Michael Krigsman’s CXOTalk.

The question that needs to be asked is “what decisions should technology be able to make and what decisions should humans be accountable for?” she said. Having a human at the helm – versus in the loop – is key to setting an agent’s goal, limits and escalation triggers, then to be able to watch the audit trail for anomalies.

Putting a human “in the loop” doesn’t work in the world of AI agents, she pointed out. “The whole point of an agent is that it’s going to reason through a complex series of steps en route to a goal and then execute them. It is not practical for a person to look through 50,000 data points and verify each one. It sort of defeats the point of having an agent.”

At the same time, it’s essential that humans be able to understand if something’s going wrong, and be able to set the parameters in advance so that anomalies are escalated to human managers. “That’s what we call human at the helm,” said Goldman. “People are still accountable and in control and can fine-tune the tool to get the right results. But you’re also harnessing that power of an agent to get stuff done on your behalf.”

Goldman suggested the following steps when setting up and deploying an agent: