The End of Work as We Know It

“The enterprises that will lead the next decade are those currently building a deep symbiosis between human intelligence and machine scale. By prioritizing high-specificity use cases, aligning data ontology, and courageously redesigning legacy processes with an AI-First mindset, leaders can move beyond simple productivity and achieve the profound operational transformation that Gen AI promises.” – Kunal Tanwar, Director of Artificial Intelligence, Business Intelligence and Data Management, Google.

Reboot recently connected with Tanwar to explore some of the critical moves essential to optimizing AI for today – and for the malleable future of the enterprise.

For starters, it’s time to rethink the use case model for the AI era. He stressed that an effective case design should incorporate as much specificity as possible, and the design should leverage human judgement and machine scale to drive not just an automated workflow but a more intelligent one.

“When I say specific, I mean not getting stuck into our previous old mindset. Because with AI we have machine scale. We can do a lot with this technology at a very fast, large scale, but still leveraging human intelligence because human judgment is critical,” he said. “When we think of a use case, how do we define this human AI interaction model to get the best outcome out there? If we are able to define those use cases from that angle, suddenly we are 50% there in terms of step change.”

AI technology can go across systems and generate insights, work products, and analysis that’s difficult or impossible for humans to do. “But when data is defined differently in one source versus another, AI will get just as confused as a human would, so your outcomes will always be suboptimal,” he cautioned. That’s where data ontology comes in.

Data ontology, a structured framework that defines the concepts, entities, and relationships within a specific domain to give data explicit meaning, is indispensable. “Focusing on data ontology and governance is critical to leverage the power of AI across large and complex data systems, and in turn increases the effectiveness of AI solutions by leaps and bounds,” Tanwar pointed out.

The need to embrace the reality of such an immense technological sea change may neither be comfortable nor convenient. As economist Daniel Susskind once mused, “When systems can use processing power, data storage, and algorithm design to perform tasks in fundamentally different ways to us, a whole new domain of activity that we thought was out of reach turns out to be out of reach.”

Tanwar highlighted a related trap to be avoided: attempting to automate the current state using AI although the current state was designed around limitations that AI doesn’t have. “A step change impact can come if we can rethink the processes and workflows with an AI first mindset,” he said.

“Rather than automating my current limited state, which was based on certain limitations, why don’t I now rethink AI first? Meaning that if I had to redo this process again, keeping this technology in mind, maybe I will redesign this process very differently.” With an AI-first mindset limitations reveal themselves faster and can be dealt with sooner.

The human-in-the-loop machine learning model relies on human interaction to improve its performance. While it bridges the gap between AI capabilities and the need for human-level precision and is used to reduce risk from potential AI mistakes, this approach bottlenecks speed and scale at human level. In Tanwar’s view, reducing human dependency is a must. “For example, “LLM as a judge” framework can help reduce the dependency,” he noted. “The business executive still must be open to accept some risk but the more we make it manageable using such advanced approaches the better outcomes we’ll achieve.”

For AI solutions to be as effective as possible, data governance must move out of the back office into the spotlight. Data is dynamic, and so are the changes to processes and new tooling to accommodate it. “We have to continually stay on top of it,” he warned. “Because we are relying on faster outcomes, if your data is great you’re gonna get great outcomes, so stop treating data governance and data management as a back office function.”

As for AI job loss anxiety, Tanwar hones in on a new paradigm: There is a big opportunity to differentiate yourself as a professional who leverages AI to increase productivity, designs new ways of working and leads the transformation rather than fighting against it.”. “Right now is the best time to learn … and you can actually become a thought leader. Everybody can because it’s so new. Everybody’s running up at the start line, so the opportunity is there for everyone right now to embrace this and actually drive some thought leadership in their respective domains.”

—

The views and opinions expressed herein are those of the interview subject alone and are shared in a personal capacity. They do not reflect the official views or positions of the subject’s employer, organization, or any affiliated entity.