<h2>Executive Summary</h2><ul><li><p>A task is only <strong>AI-ready</strong> when the technology is mature, the return is worthwhile, errors are manageable and outcomes can be explained.</p></li><li><p><strong>Entry-level roles</strong> may feel the impact first, weakening the pipeline for future expertise.</p></li><li><p><strong>Reskilling</strong> and <strong>headcount decisions</strong> should reflect shifting work requirements, new roles, changing skill needs and uneven adoption across sectors and markets.</p></li><li><p><strong>AI readiness</strong> is best assessed at the task level, where potential gains can be weighed against reliability, risk, explainability and accountability.</p></li><li><p>As <strong>AI implementation</strong> expands, judgement, domain knowledge and the ability to challenge output will become key workforce capabilities.</p></li><li><p><strong>Scaling AI</strong> requires clear guardrails around approved tools, data use, human oversight and the boundaries of experimentation.</p></li></ul>.<h2><strong>Session 1: Jobs, Tasks and the AI Transition: What Workforce Strategy Gets Wrong</strong></h2><p>Almost daily, headlines speculate about how many jobs AI will eliminate. Yet organisations cannot wait for those predictions to play out: they are already making decisions on headcount, reskilling and organisation design, often without a clear sense of which work is actually ready to shift to AI.</p><p>Sridhar Krishna outlined a framework for assessing AI-readiness based on technological maturity, return on investment, the cost of errors and explainability. He also explored what this means for role redesign and how reskilling investments should be sequenced.</p>.<h2><strong>Think Tasks, Not Jobs</strong></h2><p>AI’s capabilities are ramping up quickly but its adoption is moving much slower. General-purpose technologies usually take time to reshape organisations. Factories once clustered alongside rivers, using water to power their engines. Electricity removed this constraint, allowing them to be located almost anywhere. AI faces a similar lag as companies work through economics, reliability, regulation and evolving processes. This is why near-term job-loss estimates can easily run ahead of on-ground reality. </p><p>Work that might earlier have taken around two hours, such as preparing a presentation, takes roughly half an hour with AI. The tool can pull together selected research and build the deck, while the judgement over content and flow remains human. Many roles are likely to change in much the same way — task by task rather than all at once.</p><p> This makes tasks, rather than jobs, a more useful level for workforce planning. An HR manager or developer may keep the same title even as parts of the role are automated and others become more judgement-heavy. Headcount and reskilling decisions need to follow these shifts.</p>.<h2><strong>What Makes a Task AI-Ready?</strong></h2><p>Not every automatable task is ready for AI. Two tests matter most:</p><ul><li><p><strong>Is it worth automating? </strong>Technology<strong> </strong>must work reliably and deliver a meaningful gain in cost, speed or quality. For example, CV screening has moved beyond keyword matching, but adoption still depends on whether the gains justify the investment.</p></li><li><p><strong>Is it safe to delegate?</strong> The higher the cost of error, the greater the need for human oversight and explainability. For example, AI can flag potentially fraudulent transactions for review, but fully automated fraud decisions require greater confidence in the system’s accuracy and explainability. </p></li></ul><p>Even tasks that clear these tests may take time to move. Regulation, employment concerns, public acceptance and collective bargaining from affected groups can slow adoption, as evidenced by the debate around autonomous vehicles. At a broader level, access to compute, energy and data is increasingly shaped by geopolitics, particularly the US-China technology contest. The result is an uneven transition wherein a particular task may move quickly to AI in one organisation, sector or country but the same task may remain human-led for much longer in another.</p>.<h2><strong>Protecting the Future Talent Pipeline</strong></h2><p>AI is likely to affect entry-level roles first, particularly where junior employees spend much of their time on tasks that AI can already perform well. In software development, for instance, AI can generate code and detect bugs, while architecture and integration still require more experienced judgement.</p><p>Reducing junior hiring may make sense today, but it risks weakening tomorrow’s talent pipeline. Without enough entry-level roles, organisations lose the layer through which people build expertise and organisational knowledge, leaving them to buy that capability later, usually at a higher cost.</p><p>This changes how reskilling should be approached. Rather than training everyone broadly for an ‘AI future’, investment should follow how work is changing. Employees may need to develop higher-value judgement, learn to supervise or validate AI outputs or move into redesigned roles. Workforce planning must therefore account for both today’s headcount requirement and tomorrow’s capability requirement.</p>.<h2><strong>Higher Productivity and Jobs</strong></h2><p>AI may reduce the number of people needed for an existing volume of work, but falling costs can also generate additional demand for products and services. Cheaper, faster diagnosis, for instance, could encourage more people to seek care earlier, while affordable AI tutoring could expand access to personalised tutoring.</p><p>Just as earlier technologies created jobs that could not have been anticipated, new work will emerge, too. However these opportunities may require different skills, arise in different places or appear only once existing work has already disappeared. The immediate challenge is therefore job mismatches, not simply job losses.</p><p>In terms of workforce strategy, this means tracking which tasks will shift first, where human judgement will continue to matter and what new capabilities roles will require. </p>.<h2>Session 2: <strong>From Framework to Practice: An Open House on AI, Tasks and Workforce Planning</strong></h2><p>AI-readiness frameworks are only useful if they hold up during real-world application. For CHROs, this means testing where different tasks sit against the a defined framework, which processes are ready to shift and where reskilling or role redesign should begin.</p><p>Sridhar Krishna, Senior Scholar at <em>The Takshashila Institution</em>, moderated an open-house discussion around the AI readiness curve. Using their own organisational experiences, HR leaders explored how AI readiness, human judgement, workforce capability and governance are starting to shape workforce planning.</p>.<h2><strong>Task Level Workforce Planning</strong></h2><p>A task may seem ready for AI, but this can quickly change once the consequences of handing it over are examined more closely. An AI-generated compensation decision, for instance, may be technically feasible and efficient, but that alone does not make the process suitable for full automation. Organisations must also think through the cost of potential mistakes, the error-detection process and the onus of accountability. Human involvement remains necessary, most of all, when the consequences are significant.</p><p>Before deploying AI on a particular task, there needs to be clarity about expected improvements. Once the outcome is defined, it becomes easier to judge whether AI can improve the current process, or whether the process itself needs to change. A task inventory built around these considerations gives CHROs a clearer basis for action. It can identify work that is ready to shift entirely/mainly to AI, work that can be accelerated with human oversight and areas where the threshold for change remains high. In turn, this helps direct investment, reskilling and eventually workforce redesign.</p>.<h2><strong>The Value of Judgement </strong></h2><p>As more execution shifts to AI, employees are expected to assess, interpret and challenge what the system produces. One healthcare services organisation is reviewing its HR processes through this lens, asking where managers are required to exercise judgement and where they may simply follow a prescribed process.</p><p>Meanwhile, at an engineering company, parts of the drafting work have been automated, but the output is then examined by experts, who can spot things that the system may have missed. To fully enable this, the company trains diploma-qualified employees to work alongside AI and strengthen the judgement required to review output.</p><p>Plainly, AI fluency must sit alongside domain knowledge, curiosity, contextual understanding and the ability to question incorrect results. </p><p>In some organisations, the expectations have widened – from basic AI fluency, the ask is now demonstrated improvements in the scale/quality of output. At the same time, concerns around ‘cognitive surrender’ and excessive experimentation underline the need for people to know when to use AI, when to challenge it and where their own attention is better spent.</p>.<h2><strong>Changing Roles </strong></h2><p>As tasks shift from people to AI, certain existing roles are expanding. In some teams, employees with strong AI fluency are taking responsibility for identifying automation opportunities within their own functions. Elsewhere, organisations are creating bridge roles between engineering and product management or using internal AI champions to spread capability.</p><p>In a few cases, the change goes beyond individual roles. Dedicated transformation teams and multidisciplinary service pods bring technology, product, architecture and security together around a common outcome, changing how work is organised as well as who does it. This has implications for career architecture. Organisations will need to identify which skills remain relevant and which can be developed. They will also need to understand what new capabilities are required for bridge programs, specialist AI roles and more flexible career paths.</p>.<h2><strong>Scaling Needs Guardrails</strong></h2><p>In many cases, organisations will need clearer boundaries around AI use. Employees need to know which tools are approved, what sort of data can be entered and at what stage a task must return to a human being. Some companies are restricting AI’s use to sanctioned platforms; others are building internal environments where employees can experiment within defined controls.</p><p>How much oversight is required depends on the criticality of the task. AI may, for instance, support candidate assessment, but final decisions must be validated by a human. In software and safety-sensitive engineering work, organisations are separating generation from testing and ensuring human review. The aim is to keep accountability clear as AI moves into higher-stakes work.</p><p> In order to confidently scale AI across an organisation, leaders should clarify what the system may access, who owns the output, when a human must intervene and how much freedom employees have to experiment. </p>.<h2><strong>Measure What Changed</strong></h2><p>The success (or even extent) of AI adoption is difficult to measure if the only metric is the number of employees using AI tools. A more useful barometer is what has <em>actually improved</em>. Depending on the task, this might mean less rework, shorter turnaround times, higher throughput, better employee experience or increased output.</p><p>One organisation cut the time taken to resolve complex customer-support cases from several days to under a day. Another reduced a product-development cycle from 18 months to 6 months with a team of 5 agents. In other cases, AI has reduced plant idle time, brought down the extent of back-logs or improved the quality of first-time output.</p><p>These measures feed directly into workforce planning. Some organisations are using productivity or revenue-per-employee targets to calculate the workforce size and skill mix they will need. This gives CHROs richer information to make decisions on reskilling, role redesign and workforce composition. It also helps them better judge whether AI is creating enough value to justify wider deployment.</p><p> AI adoption truly becomes a workforce question once organisations start deciding how work should be redistributed. The task map then connects those decisions to capability, accountability and future role requirements.</p>
<h2>Executive Summary</h2><ul><li><p>A task is only <strong>AI-ready</strong> when the technology is mature, the return is worthwhile, errors are manageable and outcomes can be explained.</p></li><li><p><strong>Entry-level roles</strong> may feel the impact first, weakening the pipeline for future expertise.</p></li><li><p><strong>Reskilling</strong> and <strong>headcount decisions</strong> should reflect shifting work requirements, new roles, changing skill needs and uneven adoption across sectors and markets.</p></li><li><p><strong>AI readiness</strong> is best assessed at the task level, where potential gains can be weighed against reliability, risk, explainability and accountability.</p></li><li><p>As <strong>AI implementation</strong> expands, judgement, domain knowledge and the ability to challenge output will become key workforce capabilities.</p></li><li><p><strong>Scaling AI</strong> requires clear guardrails around approved tools, data use, human oversight and the boundaries of experimentation.</p></li></ul>.<h2><strong>Session 1: Jobs, Tasks and the AI Transition: What Workforce Strategy Gets Wrong</strong></h2><p>Almost daily, headlines speculate about how many jobs AI will eliminate. Yet organisations cannot wait for those predictions to play out: they are already making decisions on headcount, reskilling and organisation design, often without a clear sense of which work is actually ready to shift to AI.</p><p>Sridhar Krishna outlined a framework for assessing AI-readiness based on technological maturity, return on investment, the cost of errors and explainability. He also explored what this means for role redesign and how reskilling investments should be sequenced.</p>.<h2><strong>Think Tasks, Not Jobs</strong></h2><p>AI’s capabilities are ramping up quickly but its adoption is moving much slower. General-purpose technologies usually take time to reshape organisations. Factories once clustered alongside rivers, using water to power their engines. Electricity removed this constraint, allowing them to be located almost anywhere. AI faces a similar lag as companies work through economics, reliability, regulation and evolving processes. This is why near-term job-loss estimates can easily run ahead of on-ground reality. </p><p>Work that might earlier have taken around two hours, such as preparing a presentation, takes roughly half an hour with AI. The tool can pull together selected research and build the deck, while the judgement over content and flow remains human. Many roles are likely to change in much the same way — task by task rather than all at once.</p><p> This makes tasks, rather than jobs, a more useful level for workforce planning. An HR manager or developer may keep the same title even as parts of the role are automated and others become more judgement-heavy. Headcount and reskilling decisions need to follow these shifts.</p>.<h2><strong>What Makes a Task AI-Ready?</strong></h2><p>Not every automatable task is ready for AI. Two tests matter most:</p><ul><li><p><strong>Is it worth automating? </strong>Technology<strong> </strong>must work reliably and deliver a meaningful gain in cost, speed or quality. For example, CV screening has moved beyond keyword matching, but adoption still depends on whether the gains justify the investment.</p></li><li><p><strong>Is it safe to delegate?</strong> The higher the cost of error, the greater the need for human oversight and explainability. For example, AI can flag potentially fraudulent transactions for review, but fully automated fraud decisions require greater confidence in the system’s accuracy and explainability. </p></li></ul><p>Even tasks that clear these tests may take time to move. Regulation, employment concerns, public acceptance and collective bargaining from affected groups can slow adoption, as evidenced by the debate around autonomous vehicles. At a broader level, access to compute, energy and data is increasingly shaped by geopolitics, particularly the US-China technology contest. The result is an uneven transition wherein a particular task may move quickly to AI in one organisation, sector or country but the same task may remain human-led for much longer in another.</p>.<h2><strong>Protecting the Future Talent Pipeline</strong></h2><p>AI is likely to affect entry-level roles first, particularly where junior employees spend much of their time on tasks that AI can already perform well. In software development, for instance, AI can generate code and detect bugs, while architecture and integration still require more experienced judgement.</p><p>Reducing junior hiring may make sense today, but it risks weakening tomorrow’s talent pipeline. Without enough entry-level roles, organisations lose the layer through which people build expertise and organisational knowledge, leaving them to buy that capability later, usually at a higher cost.</p><p>This changes how reskilling should be approached. Rather than training everyone broadly for an ‘AI future’, investment should follow how work is changing. Employees may need to develop higher-value judgement, learn to supervise or validate AI outputs or move into redesigned roles. Workforce planning must therefore account for both today’s headcount requirement and tomorrow’s capability requirement.</p>.<h2><strong>Higher Productivity and Jobs</strong></h2><p>AI may reduce the number of people needed for an existing volume of work, but falling costs can also generate additional demand for products and services. Cheaper, faster diagnosis, for instance, could encourage more people to seek care earlier, while affordable AI tutoring could expand access to personalised tutoring.</p><p>Just as earlier technologies created jobs that could not have been anticipated, new work will emerge, too. However these opportunities may require different skills, arise in different places or appear only once existing work has already disappeared. The immediate challenge is therefore job mismatches, not simply job losses.</p><p>In terms of workforce strategy, this means tracking which tasks will shift first, where human judgement will continue to matter and what new capabilities roles will require. </p>.<h2>Session 2: <strong>From Framework to Practice: An Open House on AI, Tasks and Workforce Planning</strong></h2><p>AI-readiness frameworks are only useful if they hold up during real-world application. For CHROs, this means testing where different tasks sit against the a defined framework, which processes are ready to shift and where reskilling or role redesign should begin.</p><p>Sridhar Krishna, Senior Scholar at <em>The Takshashila Institution</em>, moderated an open-house discussion around the AI readiness curve. Using their own organisational experiences, HR leaders explored how AI readiness, human judgement, workforce capability and governance are starting to shape workforce planning.</p>.<h2><strong>Task Level Workforce Planning</strong></h2><p>A task may seem ready for AI, but this can quickly change once the consequences of handing it over are examined more closely. An AI-generated compensation decision, for instance, may be technically feasible and efficient, but that alone does not make the process suitable for full automation. Organisations must also think through the cost of potential mistakes, the error-detection process and the onus of accountability. Human involvement remains necessary, most of all, when the consequences are significant.</p><p>Before deploying AI on a particular task, there needs to be clarity about expected improvements. Once the outcome is defined, it becomes easier to judge whether AI can improve the current process, or whether the process itself needs to change. A task inventory built around these considerations gives CHROs a clearer basis for action. It can identify work that is ready to shift entirely/mainly to AI, work that can be accelerated with human oversight and areas where the threshold for change remains high. In turn, this helps direct investment, reskilling and eventually workforce redesign.</p>.<h2><strong>The Value of Judgement </strong></h2><p>As more execution shifts to AI, employees are expected to assess, interpret and challenge what the system produces. One healthcare services organisation is reviewing its HR processes through this lens, asking where managers are required to exercise judgement and where they may simply follow a prescribed process.</p><p>Meanwhile, at an engineering company, parts of the drafting work have been automated, but the output is then examined by experts, who can spot things that the system may have missed. To fully enable this, the company trains diploma-qualified employees to work alongside AI and strengthen the judgement required to review output.</p><p>Plainly, AI fluency must sit alongside domain knowledge, curiosity, contextual understanding and the ability to question incorrect results. </p><p>In some organisations, the expectations have widened – from basic AI fluency, the ask is now demonstrated improvements in the scale/quality of output. At the same time, concerns around ‘cognitive surrender’ and excessive experimentation underline the need for people to know when to use AI, when to challenge it and where their own attention is better spent.</p>.<h2><strong>Changing Roles </strong></h2><p>As tasks shift from people to AI, certain existing roles are expanding. In some teams, employees with strong AI fluency are taking responsibility for identifying automation opportunities within their own functions. Elsewhere, organisations are creating bridge roles between engineering and product management or using internal AI champions to spread capability.</p><p>In a few cases, the change goes beyond individual roles. Dedicated transformation teams and multidisciplinary service pods bring technology, product, architecture and security together around a common outcome, changing how work is organised as well as who does it. This has implications for career architecture. Organisations will need to identify which skills remain relevant and which can be developed. They will also need to understand what new capabilities are required for bridge programs, specialist AI roles and more flexible career paths.</p>.<h2><strong>Scaling Needs Guardrails</strong></h2><p>In many cases, organisations will need clearer boundaries around AI use. Employees need to know which tools are approved, what sort of data can be entered and at what stage a task must return to a human being. Some companies are restricting AI’s use to sanctioned platforms; others are building internal environments where employees can experiment within defined controls.</p><p>How much oversight is required depends on the criticality of the task. AI may, for instance, support candidate assessment, but final decisions must be validated by a human. In software and safety-sensitive engineering work, organisations are separating generation from testing and ensuring human review. The aim is to keep accountability clear as AI moves into higher-stakes work.</p><p> In order to confidently scale AI across an organisation, leaders should clarify what the system may access, who owns the output, when a human must intervene and how much freedom employees have to experiment. </p>.<h2><strong>Measure What Changed</strong></h2><p>The success (or even extent) of AI adoption is difficult to measure if the only metric is the number of employees using AI tools. A more useful barometer is what has <em>actually improved</em>. Depending on the task, this might mean less rework, shorter turnaround times, higher throughput, better employee experience or increased output.</p><p>One organisation cut the time taken to resolve complex customer-support cases from several days to under a day. Another reduced a product-development cycle from 18 months to 6 months with a team of 5 agents. In other cases, AI has reduced plant idle time, brought down the extent of back-logs or improved the quality of first-time output.</p><p>These measures feed directly into workforce planning. Some organisations are using productivity or revenue-per-employee targets to calculate the workforce size and skill mix they will need. This gives CHROs richer information to make decisions on reskilling, role redesign and workforce composition. It also helps them better judge whether AI is creating enough value to justify wider deployment.</p><p> AI adoption truly becomes a workforce question once organisations start deciding how work should be redistributed. The task map then connects those decisions to capability, accountability and future role requirements.</p>
