AI is no longer arriving in the workplace as a distant promise or a pilot project that lives in a lab. It is showing up as a working presence: chatbots that answer customer questions, copilots that draft emails and code, automated systems that triage tickets, and “agentic” tools that can take actions across business software. That shift is forcing a quieter but more consequential change than many organizations expected—how they lead, how they govern, and what HR is responsible for when part of the workforce is no longer human.
Matt Prebble, an Accenture executive, argues that businesses will be compelled to rethink leadership models as AI capabilities expand. In his view, HR cannot treat AI as purely a technology procurement issue. Instead, HR must help manage AI bots alongside employees, which means building governance, oversight, and role clarity into the everyday operating system of the organization.
The headline version of this story sounds provocative—“HR must manage AI bots”—but the underlying message is practical. When AI systems begin to influence decisions, shape workflows, and interact with people, they become part of the organizational structure. And once something becomes part of the structure, it needs accountability. That is where HR, traditionally the steward of people processes and culture, increasingly becomes the steward of human-and-AI collaboration.
A new kind of workforce problem
For decades, HR has managed the relationship between people and work: hiring, training, performance management, compliance, internal mobility, and the norms that keep teams functioning. Even when technology has been introduced—enterprise software, automation, analytics—HR’s role has usually been indirect. Technology changed tasks; HR adapted policies and training.
AI changes the equation because it can change behavior. A chatbot doesn’t just store information; it responds. A recommendation engine doesn’t just categorize; it nudges. An AI assistant doesn’t just automate a step; it can generate content, interpret context, and propose actions. As these systems become more capable, they start to resemble a functional actor in the organization—one that can communicate, recommend, and sometimes execute.
That resemblance creates a governance challenge. If an employee makes a mistake, there is a chain of responsibility. If an AI system makes a mistake, the chain can become blurry: Is it the model provider? The internal team that configured it? The manager who approved its use? The HR function that trained staff on how to interact with it? Or the policy owner who defined acceptable use?
Prebble’s point is that organizations will have to stop treating those questions as after-the-fact investigations and instead design leadership and HR structures that anticipate them.
Leadership models built for humans are being stress-tested
Leadership models are often built around assumptions that don’t hold when AI is involved. Traditional management relies on direct observation of effort, clear ownership of outcomes, and human judgment as the final layer of decision-making. But AI introduces new dynamics:
First, AI can scale certain activities instantly. A small team can produce outputs at a pace that would previously require a larger headcount. That changes how leaders measure productivity and how HR defines performance expectations.
Second, AI can standardize communication. If an AI drafts messages, generates reports, or summarizes meetings, the “voice” of the organization becomes partially algorithmic. Leaders then face a cultural question: what does it mean to be authentic, accountable, and consistent when parts of the output are machine-generated?
Third, AI can create decision velocity. When AI suggests actions quickly, leaders may feel pressure to accept recommendations without the same level of deliberation they would apply to a human proposal. That can erode governance unless there are explicit checkpoints.
Fourth, AI can shift risk. Some risks are obvious—data leakage, hallucinations, bias. Others are operational—overreliance, reduced critical thinking, and “automation complacency,” where teams stop verifying outputs because the system usually works.
These pressures mean leadership models need redesign. Not necessarily in the sense of replacing managers, but in the sense of redefining how authority is exercised, how accountability is assigned, and how oversight is embedded into workflows.
Where HR fits: governance, oversight, and role clarity
HR’s involvement is often misunderstood as “training people to use AI.” Training matters, but it is only one layer. Prebble’s framing points to a broader responsibility: HR must help ensure that AI deployment includes governance and oversight, and that roles and responsibilities are defined when both humans and AI systems are involved.
That can translate into several concrete HR-adjacent functions.
1) Defining acceptable use as a people policy, not just a technical rule
Organizations frequently publish AI usage guidelines that read like IT documentation. But acceptable use is also a behavioral contract. HR can help translate governance requirements into clear expectations: when employees must verify AI outputs, what kinds of data cannot be entered, how to cite sources, and what constitutes misuse.
The key is making policies usable. If rules are too abstract, employees will either ignore them or interpret them inconsistently. HR can turn governance into language that matches real work: examples, scenarios, and escalation paths.
2) Building training that targets judgment, not button-clicking
AI literacy is not only knowing how to prompt. It is understanding limitations, recognizing uncertainty, and knowing when to stop. HR can design training modules that teach employees how to evaluate AI outputs, how to detect errors, and how to document decisions.
This is especially important because AI systems can sound confident even when they are wrong. Training should therefore emphasize verification habits: cross-checking facts, validating calculations, and using human review for high-impact decisions.
3) Creating accountability frameworks for human-AI workflows
One of the most difficult issues is deciding who owns the outcome when AI contributes to it. HR can help establish accountability frameworks that specify responsibilities at each stage of a workflow.
For example: if an AI drafts a customer response, the employee might be accountable for tone and compliance, while the AI is accountable only for drafting. If an AI recommends candidates for hiring, the human recruiter remains accountable for fairness and final selection. If an AI flags fraud patterns, analysts remain accountable for investigation and resolution.
These distinctions should be explicit. Without them, organizations drift into either over-trusting AI or over-restricting it until it becomes useless.
4) Monitoring and oversight as an ongoing process
Governance is not a one-time policy launch. AI systems evolve, prompts change, models update, and usage patterns shift. HR can partner with legal, compliance, security, and operations to ensure there is continuous oversight—audits, feedback loops, and incident response procedures.
In practice, that means defining what happens when AI outputs cause harm. Who investigates? How are incidents logged? How are affected employees and customers informed? How are improvements made? HR’s role here is to ensure that oversight mechanisms are integrated into organizational routines rather than treated as emergency-only processes.
5) Updating performance management and incentives
If AI changes how work is produced, it also changes how performance should be evaluated. Leaders may need to adjust metrics so that employees are rewarded for quality, judgment, and compliance—not just speed.
HR can help redesign performance frameworks to account for AI-assisted work. That includes clarifying expectations around originality, review responsibilities, and ethical conduct. It also includes ensuring that employees are not penalized for using AI tools appropriately, while still holding them accountable for mistakes.
A unique take: HR as the “translation layer” between governance and daily work
There is a temptation to frame this as a simple division of labor: IT governs the bots; HR trains the humans. But Prebble’s argument implies something more nuanced. HR becomes the translation layer that connects governance principles to human behavior and organizational culture.
Governance documents often fail because they do not map cleanly onto daily decisions. Employees want to know: What should I do in this situation? What is acceptable? What is risky? Who do I ask? HR can provide that mapping.
In other words, HR helps convert abstract risk management into operational norms. That is why the leadership model must change too. Leadership is not only about strategy; it is about how decisions are made and how responsibilities are enforced. When AI is involved, the enforcement mechanism must be redesigned.
What “managing AI bots” could realistically look like
The phrase “manage AI bots” can sound like HR is responsible for model engineering. That is not the intent. Managing bots in an organizational sense means managing their lifecycle and their impact on people and processes.
A realistic approach typically includes:
– Inventory and classification: Knowing which AI tools are used, for what purposes, and with what data access.
– Risk tiering: Categorizing use cases by potential harm (for example, customer-facing content vs. internal summarization vs. decision support).
– Human-in-the-loop design: Determining where human review is mandatory and where it is optional.
– Documentation and traceability: Keeping records of how AI outputs were generated and reviewed, especially for regulated domains.
– Feedback and improvement loops: Capturing user feedback, error reports, and performance metrics to refine prompts, workflows, and policies.
– Incident response: Establishing procedures for when AI causes incorrect, harmful, or non-compliant outputs.
– Change management: Communicating updates to models or workflows so employees understand what has changed and what remains the same.
HR’s contribution is strongest where these steps intersect with people: training, accountability, escalation, and culture.
The cultural shift: from “tool adoption” to “shared responsibility”
Many organizations treat AI adoption as a rollout. They announce tools, encourage usage, and measure productivity gains. But shared responsibility requires a different mindset. It asks employees to see AI not as an invisible helper but as a participant in the workflow whose outputs must be handled responsibly.
That cultural shift can be uncomfortable. It challenges the idea that mistakes are someone else’s problem—either the vendor’s or the model’s. It also challenges the idea that speed is always good. When AI is involved, speed without verification can become a liability.
Leaders will need to model the right behaviors: demonstrating how to validate AI outputs, how to document uncertainty, and