For countless businesses, the promise of AI is just out of reach. With 94%¹ of companies eager to embrace AI, many remain stalled at the starting line, grappling with one fundamental question: Where do we begin?
Two AI product management executives, Box’s Yashodha Bhavnani and ServiceNow’s Dorit Zilbershot, unpacked this challenge in a compelling discussion, positing that the true hurdle to adoption is more about the “why” of AI than the “how.” Companies that answer this question — by identifying strategic pain points for which to solve — can succeed in the AI-first era.
Read on for insights they shared during our latest episode of the Box AI podcast.
Key takeaways:
- Identify key use cases: Prioritize AI for repetitive, mundane tasks or complex, unsolved problems to maximize impact
- Understand the agentic difference: Distinguish simple if-then automation from agentic AI, which involves reasoning and nuanced decision-making
- Embrace a hybrid approach: The future of automation lies in combining structured workflows with adaptable problem-solving
- Manage agents at scale: ServiceNow connects agents through an AI orchestrator and its own Control Tower
- Build trust through security: Robust security measures, data privacy, and user controls are paramount to grow AI adoption
Decoding the AI adoption hurdle
In the discussion between Bhavnani and Zilbershot, a consensus emerged: Random experimentation without a defined impact can stall initiatives.
Effective AI adoption, they explained, hinges on finding valuable use cases, cultivating comfort with new technology, and upskilling talent. Bhavnani suggested tackling “the hardest problem you’ve not been able to solve,” using AI to unlock solutions previously out of reach.
“Find the problem that is repetitive, boring, and what nobody wants to do,” she said, citing examples like sifting through thousands of documents for a report.
The “magic,” Zilbershot added, happens when AI, data, and workflow converge in a single platform, eliminating the need to piece together disparate tools.
“When you have a single place where you have your AI capabilities, all your data and data connections, as well as the workflows and the tools,” she said, “that makes it a lot easier for our customers to adopt AI capabilities.”
The right tool for the right problem
To succeed, businesses must also understand a critical distinction between automation and agentic AI: One follows a script, and the other specializes in reasoning.
Sequential workflows, they explained, make excellent candidates for classic automation. On the other hand, tasks that require determining next steps (like negotiating a contract with dynamic clauses or auditing lease documents for specific policy details) are prime use cases for the deeper analysis agents provide.
“That’s where agentics shine,” Zilbershot said.
She recounted a common misconception shared by a customer who requested AI agent notifications for any tickets not reviewed within a 90-day timeframe.
“ServiceNow has been able to do that for the past twenty years. It’s a simple if-then and you get a notification. You don’t need an AI agent for that,” she explained, highlighting how often organizations over-engineer solutions.
A hybrid future: Orchestration and control
The experts spoke highly of a hybrid approach to AI — that is, a strategy that combines deterministic workflows with intelligent agents.
“Agents plus workflow is the actual automation of the future,” Bhavnani said.
Zilbershot agreed, stressing that “it has to be both.” To bridge these two worlds, ServiceNow employs an AI orchestrator, which connects and manages different agents.
With these tools in place, a company can fuel multifaceted workflows like employee onboarding. Steps like setting up direct deposit (universally clear and suited for deterministic workflows) can be automated, while others specific to a role or geolocation benefit from nuanced agent reasoning.
Unlocking value with team-based agents
Zilbershot shared a compelling customer success story where a team of agents helped free up employee time and power 24/7 support. One agent would identify issues and execute solutions, while another verified resolution. Finally, a “paperwork” agent finalized documentation to close each case.
“My favorite, what they call the ‘paperwork AI agent’ can actually close the case and make sure that the work notes are there, all the information is there, and they got the approval from the employee that the issue is working,” Zilbershot said.
Anecdotes like this underscore a shared vision for a reality where agents become powerful augmenters of human potential, breaking down silos, enhancing productivity, and ultimately making work more intuitive and efficient.
“So think about if you’re outsourcing a specific task,” Zilbershot explained. “Can you outsource it to an AI agent that is able to completely do it end to end where you become a supervisor of that agentic workforce?”
Overseeing governance from the Control Tower
Security guardrails of course underpin any discussion about AI, and this conversation proved no exception.
“Security and trust is in our DNA,” Zilbershot said, adding that ServiceNow’s Control Tower provides customers with “a single pane of glass” to govern AI assets, models, systems, and integrations.
This comprehensive process bolters both oversight and cost management. CIOs, for example, can gain a unified view of AI investments mapped to their technology assets and business services for informed decision-making.
Bhavnani explained that, because Box helps ensure data never leaks back into models, customers like ServiceNow can enjoy the flexibility of choosing LLM providers and AI capabilities without worry.
Catch the full episode
Ready to dive deeper into this discussion? Tune in now to gain practical, actionable strategies for integrating AI into your organization or industry.
