Feature
Research Report
The AI Transformation 100: Concrete Ideas for Fixing How We Work
The "agentic leap" has delivered highly capable AI agents, inspiring several human entrepreneurs to radically reimagine how to automate business processes with an agent-first mindset.
Rather than beginning with the question, "How can AI agents enable humans?" they began with the premise, "How can humans enable AI agents?"
Table of Contents
- Human-in-the-Loop Won't Go Anywhere
- Who Owns Human-in-the-Loop Processes?
- Integrating Humans Into Agentic Workflows
- How to Build Sustainable Human-in-the-Loop Practices
- Operating Models Must Evolve With Human-in-the-Loop
Human-in-the-Loop Won't Go Anywhere
As enterprises look to leverage AI agents to do more knowledge work, the requirements for human-in-the loop (HITL) come into sharper focus.
Tim Law, research director at IDC, AI and automation, explained although teams of AI agents can now accomplish most of the heavy lifting involved in routine workflows and many areas of knowledge work, human workers are still required to execute "last-mile" tasks where humans are required.
“Any enterprise workflow requiring human authorization or physical signature, contractual obligations or regulatory requirements will still require humans, for now,” he said. “Any process that requires human accountability by law or for governance purposes will require HITL.”
That includes many financial workflows, legal proceedings and governmental and regulatory processes. Any high-context workflow requiring highly nuanced, high-risk judgment or interaction in the physical world will still require human-in-the-loop.
“In many cases, the human will be ‘on’ the loop, orchestrating agents rather than directly in the loop,” Law noted.
Related Article: AI Agents Are Already on Your Org Chart. Are You Designing Them Right?
Who Owns Human-in-the-Loop Processes?
All key stakeholders need to be involved in the design, according to Law, but ultimately, much of the judgment will likely fall to risk and compliance. “They understand risk severity, the regulatory compliance landscape and should be intimately involved in design decisions."
This requires business leaders as domain experts to provide insights into the context and nuance in the business processes they own, and IT and operations to ensure timing and execution of the hand-off to HITL.
Law also cautioned against overlooking lower-level employees, pointing to still-existing disparities between how work is supposed to be completed and how it actually gets done. “They often have that insight."
Integrating Humans Into Agentic Workflows
To integrate humans into agent workflows without creating new bottlenecks or hidden operational risk, Julie Bedard, managing director and partner at Boston Consulting Group, said design must begin at the workflow level, not at the isolated task level.
“If you insert human validation into individual steps without redesigning the broader end-to-end workflow, you simply move bottlenecks rather than eliminate them, and you don’t materially create value," she cautioned.
Successful organizations are taking a system-level view, redesigning tasks, roles and team structures together around value streams and end-to-end outcomes. They’re asking: where does human judgment create disproportionate value across the workflow, and how do we embed that judgment at the right points without reintroducing linear approval chains?
“Human-in-the-loop only works when it is intentionally embedded in a redesigned system that is accountable for end-to-end outcomes, not layered on top of the existing one,” said Bedard.
HITL should be event-triggered, added Law, and triggers should be reserved for high-risk, high-impact events only. “Otherwise, HITL can create bottlenecks. Thresholds should be set smartly so the workflow is not subject to constant disruption, and auto-approvals can be implemented based on deterministic rules for routine steps.”
When the workflow is paused for human-in-the-loop actions, the handoff must be seamless, and humans must be able to act quickly. The request for HITL should be packaged, and the decision or action required must be clear. “One-click approval or escalation should be used whenever possible,” Law said.
How to Build Sustainable Human-in-the-Loop Practices
From Bedard’s perspective, human-in-the-loop should not be viewed as a temporary bridge to full autonomy, but rather should be designed as a structural, durable layer in digital operations, even if its intensity or form changes over time.
The right question, she argued, isn’t “Will we eventually remove humans?” but “Where does human judgment create disproportionate value relative to risk, and how do we intentionally design work around that?”
“This is not only a risk question, but also a work design question,” she explained.
As AI reshapes execution, organizations must deliberately design roles around judgment, oversight, exception handling and accountability. “If you don’t design work for people in an AI-enabled system, you either create bottlenecks or erode trust,” said Bedard, adding that the level of human involvement should flex based on:
- The organization’s risk tolerance
- The maturity and reliability of the AI system
- The materiality of potential errors
“But given the evolving and non-linear nature of AI systems, and even more so in in complex, regulated or high-trust environments, we don’t see a future where human oversight disappears entirely,” Berdard said. “The form of human involvement will evolve. Execution may automate.”
However, judgment, accountability and exception handling remain durable features of enterprise work, and must be intentionally designed into future systems.
Related Article: How Human Employees and AI Agents Can Collaborate Safely and Efficiently
Operating Models Must Evolve With Human-in-the-Loop
As agents take on more execution, humans increasingly become accountable for judgment, oversight and system-level performance.
Webinar
Jun
3
From AI Investment to CX Results: What Enterprise Leaders Need to Know
Move beyond experiments. See how top enterprises scale AI for CX results.
Register
Webinar
Jun
4
Content Strategy Leaders Live: Managing Risk, Compliance & AI in Financial Services
Learn how financial services leaders are modernizing content systems without disrupting trust, compliance or experience.
Register
Webinar
Jun
16
From Legacy to Launch-Ready: How Gainbridge Made Its Website a Marketing-Led Growth Engine
Join in to learn how a D2C annuity brand gave marketing full website ownership — without slowing down or risking compliance.
Register
Webinar
On demand
Content Strategy Leaders Live: Managing Scale, Safety & AI in Manufacturing
How manufacturing leaders update content systems to handle product complexity and scale while integrating AI safely.
Watch Now
Webinar
On demand
AI With Intent: How CX Leaders Decide Where Automation Belongs (And Where It Doesn't)
Senior CX leaders from REI & Hyatt reveal their framework for using AI to boost retention.
Watch Now
Webinar
On demand
From Content Sprawl to Competitive Advantage: A Knowledge Operations Roadmap
How automation, AI and a centralized content platform turn knowledge fragmentation into a scalable competitive advantage.
Watch Now
As tasks inside roles evolve at higher velocity, jobs evolve with them, said Bedard, and operating models must evolve as well — a shift that has three structural implications:
- Operating models move toward orchestration and value flow: Roles become defined less by execution and more by judgment, prioritization and end-to-end accountability
- Vendor ecosystems shift toward integrated accountability: Organizations will evaluate partners based on how well AI systems integrate into end-to-end workflows and support human oversight aligned to clear outcome ownership.
- Accountability becomes sharper, not diffused: As transparency increases, the question “Who owns this outcome?” becomes more central. Enterprises will need clearer ownership models for AI-enabled results, especially in complex or regulated environments.
“We’re likely to see fewer roles defined purely by execution and more defined by judgment, systems thinking and outcomes-based, because when the tenets of work change, structure must follow,” said Bedard.