Lokdeep Singh, CEO of Unily – Interview Series

Lokdeep Singh has more than 20 years’ experience leading high-growth enterprise software companies. Most recently, as Senior Advisor at Providence Equity Partners, where he supported portfolio companies on AI strategy, operational excellence, and go-to-market acceleration. Prior to that he was CEO of Talkwalker, a global leader in enterprise consumer intelligence, where he led the transformation to achieve sustainable, profitable growth; and at Synchronoss Technologies as General Manager.

Unily is an enterprise employee experience platform designed to help large organizations connect employees, streamline internal communications, and bring workplace tools, knowledge, and services into a unified digital environment. Founded in 2006, the company serves global enterprises with distributed, desk-based, and frontline workforces, combining intranet capabilities, employee communications, search, analytics, integrations, and personalization within a single platform. Unily increasingly positions its technology as AI-native, using governed artificial intelligence to automate content management, surface relevant information, personalize employee experiences, and enable workers to interact with connected enterprise applications through conversational interfaces. The platform integrates with widely used workplace systems such as Microsoft 365 and Teams and is built for organizations with complex technology environments, regulatory requirements, and large multilingual workforces.

You began your career deep in telecom engineering and network infrastructure before moving into CTO, product, general management, and CEO roles, eventually leading Talkwalker through a successful exit and now taking the helm at Unily. How has that progression shaped your view of what it actually takes for employees to trust and adopt a transformative technology like AI?

My career has given me a perspective that sits between technology and the people expected to use it. Early on, I was focused on infrastructure, making sure systems worked reliably, securely, and at scale. As I moved into product, general management, and ultimately CEO roles, I learned that technical capability is only one part of whether a technology succeeds.

Employees will only adopt an emerging technology when they understand how it affects their work, believe it will help rather than hurt them, and have confidence that it will behave predictably.

The biggest mistake leaders can make is treating AI adoption as a software rollout. You can deploy the best model in the world, but if employees don’t trust the information it provides, don’t understand what it is allowed to do, or don’t know where the boundaries are, adoption will remain superficial. Trust has to be designed into the employee experience from the beginning.

AI capabilities are improving extraordinarily quickly, yet trust inside organizations often seems to be moving much more slowly. What is driving this widening gap between what AI can technically do and what employees are actually comfortable allowing it to do?

The trust gap exists because organizations have invested heavily in making AI capable, but not at the same pace in making it understandable, accountable, and grounded in trusted information.

AI can now reason, generate content, and take action, but many organizations still lack an authoritative system of record for knowledge and content. When employees don’t know where AI is getting its answers, whether information is current, or who owns it, trust erodes quickly.

Closing the gap requires more than better models. It requires strong content governance and a trusted knowledge foundation that ensures AI is working from accurate, secure, and up-to-date information. The organizations that succeed will be the ones that make trust part of the AI experience itself, not an afterthought.

Companies often approach enterprise AI as a technology deployment problem centered on models, integrations, security, and infrastructure. Why do you believe the harder challenge is ultimately a people and organizational one?

AI changes the relationship between people and technology. Previous enterprise technologies largely waited for employees to tell them what to do. AI can interpret intent, make recommendations, and increasingly take action.

That makes AI adoption a future-of-work challenge, not just a technology challenge. It changes how work gets done, how decisions are made, and how teams collaborate. Organizations can’t simply deploy AI and expect transformation to follow. Employees need to understand where AI fits into their workflows and feel empowered to use it effectively.

The most successful AI initiatives are driven by grassroots adoption, not top-down mandates. When employees see clear value in their day-to-day work, adoption accelerates naturally.

That’s why the framework for AI ROI is simply around trusting, empowering, and governing. Build trust through transparency and trusted knowledge, empower employees with tools that improve how they work, and establish governance that ensures AI is used responsibly. Technology enables the change, but people, processes, and culture ultimately determine its success.

Organizations need governance, but overly restrictive policies can push employees toward unauthorized tools or discourage experimentation altogether. What does an effective AI governance framework look like if the goal is to increase both trust and adoption rather than simply control usage?

Good governance should feel like guardrails that allow employees to move faster safely.

That means being very clear about what employees can use AI for, what information they can provide to AI systems, and what actions AI is permitted to take. Those rules should be understandable to an employee, not buried in a lengthy policy document.

Governance also needs to be dynamic. The risk associated with AI generating a first draft is very different from AI approving a payment, changing an employee record, or communicating with a customer. The level of human oversight should correspond to the potential consequence.

Ultimately, the goal shouldn’t be to minimize AI usage. It should be to maximize responsible usage.

Unily research found that 52% of employees reported having no known AI policy at their organization, while some employees acknowledged entering sensitive information into unsanctioned AI tools. To what extent is the rise of “shadow AI” actually a failure of communication and governance rather than an employee behavior problem?

We should be careful about framing shadow AI as a bad-employee problem. A significant reason for shadow AI use is a symptom of unmet demand. Employees are finding ways to solve problems that their organizations haven’t yet given them approved tools to solve, which is a natural human behavior.

Rather than expecting them to ignore technology that is changing how work gets done, the answer is to give employees a safe, sanctioned path to experiment, including clear boundaries and education.

Transparency and explainability are frequently presented as essential to trustworthy AI, but employees do not necessarily need to understand how a large language model works internally. What information should an AI system provide so that an employee can make an informed decision about whether to trust its answer or recommendation?

An enterprise AI system should make it clear where information came from, when that information was last updated, what sources or enterprise systems were used, and why a particular recommendation was made.

It should also be honest about uncertainty. One of the most important trust signals is an AI system saying, in effect, “I don’t have enough information to answer that confidently.”

Trust doesn’t come from AI pretending to be certain. It comes from giving employees enough context to exercise their own judgment.

Unily Glass is designed to move enterprise AI beyond answering questions and toward executing actions across systems such as HR, IT, and customer relationship management platforms. As AI moves from providing information to actually taking action, how should organizations decide which decisions agents can make autonomously and which should continue to require human approval?

The key principle should be risk proportionality. If an AI agent is scheduling a meeting or finding information, the downside of an incorrect action is relatively limited. You can give it much more autonomy. However, if an agent is approving a financial transaction, changing an employee’s compensation, terminating access, or making a consequential customer decision, the threshold should be dramatically higher.

Organizations should establish explicit levels of autonomy based on factors such as financial impact, regulatory exposure, reversibility, and impact on people. The more consequential or irreversible the action, the more important human approval becomes.

The objective is to keep humans meaningfully in control where it matters.

Enterprises are increasingly deploying multiple copilots, specialized agents, large language models, and AI-enabled applications across the technology stack. How can organizations prevent this from creating another layer of fragmentation, and do you expect a centralized trust or orchestration layer to become essential?

We are at risk of recreating the same fragmentation that enterprise software has created over the past several decades, except now every application can have its own AI interface and its own agent.

Employees shouldn’t have to understand which model sits behind which application or remember where each organization’s AI lives. They should be able to interact with AI in the context of their work and have the right information and capabilities brought together for them.

That’s where an orchestration and trust layer becomes increasingly important. It can provide a consistent way to manage identity, permissions, context, data access, and actions across different AI systems.

There likely won’t be one model or one AI platform that owns everything, but organizations will increasingly need a common layer that makes all of these systems work together responsibly.

Many organizations can measure how often employees use an AI tool, but usage alone does not prove that the investment is creating value. Which metrics should leaders be tracking to determine whether AI is genuinely improving productivity, decision-making, employee experience, or business outcomes?

Usage is an interesting starting point, but it is not the outcome. Leaders should ask whether AI is reducing the time it takes to complete important workflows, eliminating repetitive work, improving the quality or speed of decisions, and reducing friction for employees.

I’d look at metrics such as time saved per workflow, task completion rates, error rates, employee satisfaction, time-to-resolution, and ultimately business metrics tied to the specific function deploying AI.

The strongest organizations will connect AI metrics directly to business outcomes rather than celebrating activity for its own sake.

As enterprise AI evolves from assistants that generate content into agents that can execute increasingly complex workflows, what do you believe business leaders are still underestimating about the relationship between AI capability, employee trust, and successful adoption?

Many leaders are still underestimating how much trust determines the ceiling for AI adoption.

Companies often talk about AI capability as though more capability automatically means more value. It doesn’t. An incredibly capable system that employees don’t trust can create less value than a more limited system that they understand and confidently use.

This requires companies to think about technology, governance, employee experience, and organizational change as one system. Companies must figure out how to make AI trusted enough to actually change the way people work.

Thank you for the greast interview, readers who wish to learn more should visit Unily.