As AI Moves From Data Insights to Action, Where Do Humans Fit In?
As artificial intelligence systems mature, enterprise leaders have begun to connect data insights to automated workflows, so that agents can act on analytical takeaways on an automated basis.
ServiceNow recently unveiled a host of new capabilities that combine to build the real-time data foundation for autonomous decision intelligence. Among them is a Context Engine that ensures AI agents can access and understand the dynamic business context signals and governance rails needed to work without human supervision.
It’s fed by an Autonomous Data Analytics tool built on technology from Pyramid Analytics, which enables users to query business data in natural language and dig up the contextual insights needed to hone their decision-making automations.
Omri Kohl, co-founder and CEO of Pyramid, which was acquired by ServiceNow earlier this year, says that this marriage of the companies’ technologies has effectively solved the problems associated with data siloes that have stymied enterprise adoption of AI agents until now. Most organizations use multiple software systems, which results in data being fragmented and isolated across platforms, meaning that AI agents cannot access the business context and governance they need to perform work autonomously.
Because the ServiceNow Context Engine provides pivotal context and governance layers, business leaders can use AI for far more than strategic insights. They can now access a ready-made engine for automated execution. “The playbook for strategic business decision-making has been completely rewritten by AI,” Kohl asserts. “More and more organizations are adopting AI-powered, data-driven decision intelligence workflows to help evaluate risks and opportunities.”
But does this mean senior executives can start planning mass layoffs in anticipation of being able to automate the bulk of their business operations? Not necessarily. While agents might be able to handle the work, only humans can be held accountable for the decisions they make.
Context That Bridges the Trust Gap
As Kohl puts it, the main hurdle to enterprise automation was never the lack of data, but the need for agents to understand it. Raw business data, no matter how voluminous or structured it may be, is only one piece of the puzzle.
In addition, you need a sophisticated context layer that’s able to map elements like priorities, people, policies, roles and assets, and it needs to do this in real time. If you don’t have this layer, you cannot trust AI with strategic decisions, Kohl says, because it lacks knowledge of the rules that the business has to abide by.
The integration of Pyramid’s insights engine with ServiceNow’s Context Engine makes it possible for someone to query their organization’s entire data estate using simple language. This results in a kind of “insight-to-action” pipeline that provides agents with the institutional knowledge they need to make strategic choices. The more robust that context layer is, the more autonomous agents can be trusted.
The Role of Humans in the Automation Loop
It’s at this point that business leaders need to ask themselves where humans fit into this constellation. Kohl believes it’s necessary to stop viewing the idea of “humans-in-the-loop” as a binary decision, and instead look at it as a shifting continuum.
Today, there are three distinct paths for business decision-making. On the one hand, there’s full automation, where AI can be trusted to perform low-risk operational tasks at high speed without causing any problems. Alternatively, there are human-led strategies, where AI is limited to doing the research while humans look at what they dig up and make the decisions. That’s a sensible approach in high-stakes scenarios, but it also means moving at the speed of yesterday.
The third path involves augmented decision-making, which is precisely where the synergy between ServiceNow and Pyramid is meant to shine. It’s where AI does the research and suggests which road to take, while a human oversees with full veto powers.
“An experienced human decision-maker can spot when an output doesn’t seem quite right, even if they can’t yet put their finger on what exactly has gone wrong,” Kohl warns. “AI won’t make this leap... it can fail convincingly, creating believable yet dangerous suggestions.”
When Logic Falls Short, Accountability Is Vital
Even with a world-class context engine providing governance, AI agents can still run into problems, especially when it comes to cross-domain logic.
As an example, an AI system might spot what it thinks is a golden opportunity for the company to increase its revenue, but fail to understand that pursuing it would cross an ethical boundary that employees hate. Alternatively, it might not realize that its taking an action that directly contravenes a recent mandate from the CEO.
This is why you still need humans involved with the AI agents. Only real people can be held accountable, after all. Agents are designed to execute commands based on the data presented to them, but they cannot be held responsible for whatever fallout occurs because of that decision.
For Kohl, strategic decision-making requires someone to not only analyze what the numbers say, but also take responsibility for whatever that decision leads to. “AI’s real strength is spotting patterns,” he explains. “However, it’s not good at connecting the dots across domains…. This is why humans retain a monopoly on tasks that involve strategic intuition, analogical thinking and reading the room.”
AI Agents Understand Probabilities, not Principles
One of the fundamental disconnects, according to Kohl, is that AI models are designed to understand logic, rather than make nuanced, ethical choices.
On the other hand, when executives are thinking about strategic problems, it’s a rare case when they’ll only consider the most efficient outcome, for they also have to balance it with cultural aspects and the social consequences of their choices.
“AI can’t take responsibility for enterprise decisions,” Kohl points out. “You need someone to own the outcome and take responsibility for the unexpected fallout that will sometimes crop up. AI can’t be expected to fill that role.”
This is why Kohl believes the role of modern business leaders is undergoing a process of evolution. Instead of being the primary decision-makers, they may soon spend the bulk of their time negotiating the guardrails of the agents that make those decisions. As AI increases its capabilities, humans will directly oversee a lot less manual execution, but we’ll likely become much more involved in higher-level governance.
Accelerating Human Workflows
While Kohl might be considered a pioneer of enterprise decision workflow automation, his vision isn’t a future of empty offices. Instead, he sees a world in which AI helps to improve human intelligence.
“The organizations that succeed aren’t replacing humans with automation,” Kohl says. “They’re adding AI to elevate human workflows.” The insight-to-action engine that ServiceNow and Pyramid have built serves as a copilot that’s unfazed by the volume of modern data.
Here, the human strategist takes on the role of captain, overseeing the copilot and ensuring it doesn’t make any costly mistakes. It’s much like fitting a powerful engine in a car. The vehicle is guaranteed to move faster, but you’ll still need a human at the wheel to follow the right path.
What this means is that the competitive edge won’t belong to the companies with the largest datasets, the most powerful models, or even the fastest processors. Instead, the winners will be those that ensure humans remain accountable for AI’s increasingly potent decisions over time.