<h2><strong>Executive Summary</strong></h2><ul><li><p>AI requires <strong>human oversight</strong> at every consequential step.</p></li><li><p>The <strong>CHRO</strong>, rather than the CIO or CTO, <strong>should be the primary sponsor</strong> of <strong>AI transformation</strong> in an organisation.</p></li><li><p><strong>Board and CXO-level immersion in AI</strong> tools through <strong>hands-on sessions</strong> is what shifts default behaviours, instead of just producing isolated pockets of usage.</p></li><li><p>Designating the most <strong>junior employee in each function</strong> as an <strong>AI champion</strong> generates more actionable use cases and places ownership closer to the work.</p></li><li><p><strong>Embedding AI agents formally in the organigram</strong>, with job descriptions and cost allocations, is how AI can move from pilot to operational reality.</p></li><li><p><strong>License access </strong>must<strong> </strong>be<strong> restricted to those actively building with AI</strong>; unrestricted deployment entails <strong>massive costs</strong> that can offset productivity gains. </p></li></ul>.<p>Across sectors, companies are either piloting AI tools, experimenting at the edges or, in some cases, rolling them out across functions. So far, the productivity gains have been patchy, adoption uneven and the workforce has been left largely unchanged. However, based on a year of operational experience at his company, Dr Harpreet Singh Anand, Chief Human Resources Officer of Protean eGov Technologies (the organisation that operates the digital infrastructure underpinning India's PAN, NPS and Aadhaar systems), argues that the gap between AI aspiration and reality is crossable, but only if you ask the right questions first. These include: Who should lead the effort? What should shift culturally? Which problems are worth automating? And where does human judgement remain indispensable? The challenge lies not in getting AI to improve productivity, but in leveraging to entirely redesign how work gets done.</p><h2><strong>Human-Led, AI-Enabled</strong></h2><p>The governing principle behind <strong>Protean's</strong> AI program is that AI is a tool, not a decision-maker. Two recent cases illustrate the perils of forgetting this distinction. When <strong>Deloitte Australia</strong> submitted an AI-generated report to a client without verifying either the underlying data or the analysis, it suffered huge reputational damage. Similarly, when <strong>Google</strong> attempted to automate its hiring process – all the way from sourcing to offer – it ran up against unexpected issues. Specifically, candidates were able to manipulate the system, leading to hiring outcomes that failed basic capability checks. The lesson here is not that AI has <em>no role</em> in recruitment, but that final judgement cannot be outsourced entirely. At least one human check is necessary, particularly in decisions involving capability, fairness, risk or reputation.</p><h2><strong>The CHRO as the Primary Sponsor</strong></h2><p>AI adoption cannot be left to technology teams alone. At Protean, the AI transformation mandate was initially treated as a CIO/CTO-led exercise, but this idea was quickly challenged from within. HR argued that any intervention that changes how people work, what they fear, how they learn and how they are measured must have HR at the centre. The CHRO should become the primary sponsor of the transformation, with the CIO as the execution partner. While technology teams can build the tools, it falls on HR to create the confidence, incentives and behavioural change needed for adoption. Without that, even the best tools can remain underused or misused.</p><h2><strong>Making AI Part of the Culture</strong></h2><p>AI adoption cannot be built through training sessions alone. Rather, it must become part of the organisation’s operating rhythm. At Protean, this was achieved by creating a cross-function AI task force. Importantly, instead of just asking senior leaders or technologists for use cases, each function was asked to nominate its junior-most team member as an ‘AI champion’, and to identify tasks or projects that could be automated. The logic was simple: the people closest to repetitive, frustrating or inefficient work are often best placed to decide where automation can create value. In many organisations, the leadership may simply need to ‘get out of the way’ and allow frontline creativity to surface use cases that would not emerge from a top-down transformation program.</p><p>Recognition was another important lever. Protean’s leadership began publicly acknowledging successful AI deployments. This helped reposition AI, not as a threat but as a source of pride and experimentation. The organisation also made failure acceptable. Employees were encouraged to try, fail quickly and move on. This is critical because early AI use often produces uneven results. Users may struggle with tools, prompts may fail, and productivity may initially dip before it improves. Sustained adoption requires a culture in which experimentation is not punished but encouraged.</p><h2><strong>Licences Need Discipline, Not Blanket Access</strong></h2><p>Licence access is among the most underestimated design decisions in any AI program. Protean's initial rollout gave all 735 employees full access, but the resultant token costs were substantial. To correct for this, the organisation then restricted access to those developing genuine use cases, while others used more limited or open tools. Such cost discipline comes with an added environmental benefit, given that AI usage entails massive consumption of energy and water (which is used to cool computing systems). All of this must be accounted-for in ROI calculations, and organisations need a robust framework for deciding where AI is genuinely useful and where it is merely a costly experiment.</p><h2><strong>Data Sovereignty and Governance </strong></h2><p>For organisations dealing with sensitive data, open AI tools create serious risks. For instance, confidential information can easily get uploaded onto external tools before the implications of doing so are fully understood. To address this, Protean built up its internal governance, including mandating checks by risk, audit and information security teams before sensitive data gets used in AI applications. It also restricted access to certain tools and began developing internal small language models (SLMs), that could operate within Protean’s own protected environment. The advantage of SLMs is that sensitive data remains within the enterprise system rather than being fed into external platforms. Over time, these smaller models can be trained and matured into more capable internal language models.</p><p>It is also critical to conduct regular checks on whether AI agents are using approved sources, whether the output reflects AI ‘hallucination’, whether the data has been correctly whitelisted and whether the tool is behaving as intended. This requires collaboration between HR, technology, risk, audit, compliance and business teams. AI governance is therefore not a one-time policy document, but an ongoing operating discipline.</p><h2><strong>Taking a Step Back…</strong></h2><p>AI adoption is at heart not just a technology issue, but a people, culture and governance one. It requires visible leadership participation, CHRO ownership, CIO partnership, CFO scrutiny and frontline experimentation. It must be ‘exciting’ enough to build momentum, disciplined enough to control cost and safe enough to protect data and judgement. </p><p>Several practical use cases demonstrate this potential, from grievance automation and regulatory monitoring to RFP tracking and marketing strategy. These examples show that AI can be applied to both, transactional efficiency and higher-order work, whether by scanning regulatory changes in real time, reducing dependence on large, outsourced teams or enabling more strategic brand thinking. The real opportunity lies in reimagining how work itself is structured. The ideal is a human-led, AI-enabled workplace, where people use machines to raise standards, improve speed and unlock new possibilities, while retaining accountability for decisions that matter.</p>
<h2><strong>Executive Summary</strong></h2><ul><li><p>AI requires <strong>human oversight</strong> at every consequential step.</p></li><li><p>The <strong>CHRO</strong>, rather than the CIO or CTO, <strong>should be the primary sponsor</strong> of <strong>AI transformation</strong> in an organisation.</p></li><li><p><strong>Board and CXO-level immersion in AI</strong> tools through <strong>hands-on sessions</strong> is what shifts default behaviours, instead of just producing isolated pockets of usage.</p></li><li><p>Designating the most <strong>junior employee in each function</strong> as an <strong>AI champion</strong> generates more actionable use cases and places ownership closer to the work.</p></li><li><p><strong>Embedding AI agents formally in the organigram</strong>, with job descriptions and cost allocations, is how AI can move from pilot to operational reality.</p></li><li><p><strong>License access </strong>must<strong> </strong>be<strong> restricted to those actively building with AI</strong>; unrestricted deployment entails <strong>massive costs</strong> that can offset productivity gains. </p></li></ul>.<p>Across sectors, companies are either piloting AI tools, experimenting at the edges or, in some cases, rolling them out across functions. So far, the productivity gains have been patchy, adoption uneven and the workforce has been left largely unchanged. However, based on a year of operational experience at his company, Dr Harpreet Singh Anand, Chief Human Resources Officer of Protean eGov Technologies (the organisation that operates the digital infrastructure underpinning India's PAN, NPS and Aadhaar systems), argues that the gap between AI aspiration and reality is crossable, but only if you ask the right questions first. These include: Who should lead the effort? What should shift culturally? Which problems are worth automating? And where does human judgement remain indispensable? The challenge lies not in getting AI to improve productivity, but in leveraging to entirely redesign how work gets done.</p><h2><strong>Human-Led, AI-Enabled</strong></h2><p>The governing principle behind <strong>Protean's</strong> AI program is that AI is a tool, not a decision-maker. Two recent cases illustrate the perils of forgetting this distinction. When <strong>Deloitte Australia</strong> submitted an AI-generated report to a client without verifying either the underlying data or the analysis, it suffered huge reputational damage. Similarly, when <strong>Google</strong> attempted to automate its hiring process – all the way from sourcing to offer – it ran up against unexpected issues. Specifically, candidates were able to manipulate the system, leading to hiring outcomes that failed basic capability checks. The lesson here is not that AI has <em>no role</em> in recruitment, but that final judgement cannot be outsourced entirely. At least one human check is necessary, particularly in decisions involving capability, fairness, risk or reputation.</p><h2><strong>The CHRO as the Primary Sponsor</strong></h2><p>AI adoption cannot be left to technology teams alone. At Protean, the AI transformation mandate was initially treated as a CIO/CTO-led exercise, but this idea was quickly challenged from within. HR argued that any intervention that changes how people work, what they fear, how they learn and how they are measured must have HR at the centre. The CHRO should become the primary sponsor of the transformation, with the CIO as the execution partner. While technology teams can build the tools, it falls on HR to create the confidence, incentives and behavioural change needed for adoption. Without that, even the best tools can remain underused or misused.</p><h2><strong>Making AI Part of the Culture</strong></h2><p>AI adoption cannot be built through training sessions alone. Rather, it must become part of the organisation’s operating rhythm. At Protean, this was achieved by creating a cross-function AI task force. Importantly, instead of just asking senior leaders or technologists for use cases, each function was asked to nominate its junior-most team member as an ‘AI champion’, and to identify tasks or projects that could be automated. The logic was simple: the people closest to repetitive, frustrating or inefficient work are often best placed to decide where automation can create value. In many organisations, the leadership may simply need to ‘get out of the way’ and allow frontline creativity to surface use cases that would not emerge from a top-down transformation program.</p><p>Recognition was another important lever. Protean’s leadership began publicly acknowledging successful AI deployments. This helped reposition AI, not as a threat but as a source of pride and experimentation. The organisation also made failure acceptable. Employees were encouraged to try, fail quickly and move on. This is critical because early AI use often produces uneven results. Users may struggle with tools, prompts may fail, and productivity may initially dip before it improves. Sustained adoption requires a culture in which experimentation is not punished but encouraged.</p><h2><strong>Licences Need Discipline, Not Blanket Access</strong></h2><p>Licence access is among the most underestimated design decisions in any AI program. Protean's initial rollout gave all 735 employees full access, but the resultant token costs were substantial. To correct for this, the organisation then restricted access to those developing genuine use cases, while others used more limited or open tools. Such cost discipline comes with an added environmental benefit, given that AI usage entails massive consumption of energy and water (which is used to cool computing systems). All of this must be accounted-for in ROI calculations, and organisations need a robust framework for deciding where AI is genuinely useful and where it is merely a costly experiment.</p><h2><strong>Data Sovereignty and Governance </strong></h2><p>For organisations dealing with sensitive data, open AI tools create serious risks. For instance, confidential information can easily get uploaded onto external tools before the implications of doing so are fully understood. To address this, Protean built up its internal governance, including mandating checks by risk, audit and information security teams before sensitive data gets used in AI applications. It also restricted access to certain tools and began developing internal small language models (SLMs), that could operate within Protean’s own protected environment. The advantage of SLMs is that sensitive data remains within the enterprise system rather than being fed into external platforms. Over time, these smaller models can be trained and matured into more capable internal language models.</p><p>It is also critical to conduct regular checks on whether AI agents are using approved sources, whether the output reflects AI ‘hallucination’, whether the data has been correctly whitelisted and whether the tool is behaving as intended. This requires collaboration between HR, technology, risk, audit, compliance and business teams. AI governance is therefore not a one-time policy document, but an ongoing operating discipline.</p><h2><strong>Taking a Step Back…</strong></h2><p>AI adoption is at heart not just a technology issue, but a people, culture and governance one. It requires visible leadership participation, CHRO ownership, CIO partnership, CFO scrutiny and frontline experimentation. It must be ‘exciting’ enough to build momentum, disciplined enough to control cost and safe enough to protect data and judgement. </p><p>Several practical use cases demonstrate this potential, from grievance automation and regulatory monitoring to RFP tracking and marketing strategy. These examples show that AI can be applied to both, transactional efficiency and higher-order work, whether by scanning regulatory changes in real time, reducing dependence on large, outsourced teams or enabling more strategic brand thinking. The real opportunity lies in reimagining how work itself is structured. The ideal is a human-led, AI-enabled workplace, where people use machines to raise standards, improve speed and unlock new possibilities, while retaining accountability for decisions that matter.</p>
