Understandably, varying approaches have emerged in the early days of agentic. At the Republic Bank of Chicago, CIO Madhu Reddy emphasizes improved decision-making through agents over quick efficiency wins, thereby turning it into a tool of trust. Over at McKinsey, the focus is on building a 1:1 human and agentic workforce, highlighting a desire to maintain the intern-to-partner pipeline, while creating an all-new outcomes-based business model “where we identify a joint business case with our clients, and we underwrite the outcome by tying our fees to the impact our work delivers for them,” says CEO Bob Sternfels.

Work intensifies (and so do the risks)

Leaders face a choice: hop on to an aggressive automation bandwagon and wait for the promised benefits. Or take a measured approach to ensure the risks that destabilize your people, impact your processes, and ultimately push productivity down (the exact opposite of what should happen) are avoided. Some risks from an over-enthusiastic adoption of AI that our research uncovered include:

Cognitive overload that risks employee burnout: Employees understand what AI can do and are proactively using it to do more. The results, however, highlight a stark reality. While productivity has zoomed, so has employee burnout as the natural boundaries between work and life blur. This intensification, as researcher Xingqi Maggie Ye observed, “can feel positive in the short bursts that make up the day, while the cumulative effect creates strain over time.”
Accountability issues that create systematic risk: A 1979 computer manual from IBM famously cautioned users that “A computer can never be held accountable, therefore a computer must never make a management decision.” But the agent-fueled intensification of work means both the speed and number of decisions being made at the workplace continue to explode and challenge traditional oversight. Without new, human-centric frameworks for accountability, teams could plunge into endless blame games, and trust between employee and employer will break down.
Knowledge silos and the loss of critical thinking: Our research shows that leading firms (14% of our sample) believe all teams should have universal access to insights generated from customer data, with nearly all of them (93%) deploying the systems and policies specifically to scale insights across their firm (like a global data office). Moreover, the lack of mechanisms in place to share insights and best practices means that critical thinking skills fade as faith in what machines confidently say goes unchecked. This over-reliance on machines without human judgment could lead to vulnerabilities creeping into critical processes.

Start with clear cultural principles

There are many theoretical frameworks for organizational design that can be applied⁴. With culture, it's all about people and the work they do. Virtusa Research offers three guiding principles when thinking about the work people do:

Ensure transparency is your new cultural benchmark: Encourage everyone to proactively and inclusively share relevant information. Greater transparency doesn’t just foster trust. It fuels creativity, performance, and profitability–see LinkedIn example. For example, treat agentic workflows as collective IP, where employees are incentivized to build them out for the wider team rather than hoard them. Leaders should set the example by being clear about career growth and clearly delineating where an agent’s role ends, and a human’s begins (see the sidebar: Build an onboarding plan for your agents).
No matter the tech, recognize that only people drive accountability. Underperformance is not a failure of change management but a mismatch between new ways of working and old organizational designs. Human judgement remains the true measure of performance, with analytical assets either supporting human judgement or operating autonomously within defined guardrails. Skepticism of AI decisions must be encouraged to ensure the intuition muscle stays active. Analytical assets either support human judgment or operate autonomously within defined guardrails, shifting responsibility from experts to analysts and algorithms. The result is better and faster decisions, driving impact and growth
Develop trust consistently through empathy, logic, and authenticity: A strong human-agent relationship requires a no-fear environment. Leaders must create this by prioritizing human judgement and providing clear guidance on career growth based on their ability to handle agentic complexity as the tech undoubtedly improves. Our research reveals that building trust among employees means they need to believe three things: that leadership cares about them (empathy), that it's capable of meeting their needs (logic), and that you can be expected to do what you say you’ll do (authenticity). Call it the trust trifecta, but when people lose trust, organizations that are losing trust tend to underperform on one of these three dimensions.⁵

Sidebar: Building an onboarding plan for your agents

Yes, we’re at risk of humanizing the tech, but it’s happened all through history, from Proud Mary powering a steamer on the Mississippi to the names we give our cars. Effective AI agents need a clear onboarding plan that describes what they do, how they work, and how your people should think about their new teammates. Yes, they’re machines, but we recommend:

Giving every agent a job description
We need to stop with vague mandates like "optimize". If you wouldn’t hire a human with a job description that just said "be efficient," don’t do it to an agent. Specify their decision rights and authorities. At Virtusa, we call this the shift from "dumb bots" to "engines of empowerment". If you can’t define what the agent is responsible for, don't be surprised when it hallucinates its own objectives and undermines decision quality.
Design for the "dull and dispiriting
Automation once saved us from the "dirty, dark, and dangerous". Agentic AI is here to save us from the "dull, dispiriting, and deterministic". Clients don't want to burn human capital on repeatable loops or fixing bugs that a technology fabric should have predicted and patched. By designing agents to swallow the drudgery, you give your human talent a reason to actually embrace the tech rather than fear it.
Running performance review (for algorithms)
If you aren't measuring it, you aren't managing it. Agents need a regular evaluation cycle that looks at timeliness and reliability, not just "accuracy". This isn't just for the sake of the code; it’s for the humans. Teammates need to know their algorithmic "colleagues" are being held to the same standards. Without metrics, you can’t tell the difference between a minor glitch and a total system failure.
Giving them a mentor
We’re hearing a lot about "multi-agent orchestration," in which Sales, Dev, and Ops agents communicate with one another. It sounds sophisticated, but someone still needs to be the adult in the room. Every agent needs a human supervisor. Organizations remain legally and ethically accountable for what their AI does. You need a "sentient decision-maker" to ensure that "unverified intelligence" doesn't create a systemic risk for the brand.
Hire agents as interns
No agent should get "full-time status" on day one. They may have the "training" (the data), but they lack the contextual intelligence of your specific culture, strategy, and domain knowledge. Treat them like precocious interns: give them clear guidance, see how they perform in a hybrid environment, and only then make them a permanent part of the workflow. We can leapfrog the competition by focusing on these agentic skills now.
Give the agent a name
This isn’t about being "cute" or "humanizing" the machine. It’s about accountability. When a decision is blamed on "the AI," responsibility evaporates. By giving an agent a name and a clear identity, leveraging platforms like Virtusa Helio makes the role discussable, transparent, and manageable. It moves AI from an "interesting capability" to a core proposition.

Services

Industries

Insights

Careers

About