<p><em>The AI Reality Check is an IMA India research series examining how artificial intelligence is changing the operational reality of Indian enterprises. It is based on in-depth interviews with CXOs and senior operational leaders from IMA member companies, supplemented by secondary research. </em></p><p><em>This is the <strong>second of five papers.</strong></em></p>.<p><strong>Executive Summary</strong></p><ul><li><p><strong>Data architecture</strong> is often the first and most consistent barrier to AI deployment </p></li><li><p><strong>Deploying AI</strong> on a broken or inconsistent process does not fix it. It simply makes it faster at being broken.</p></li><li><p>If the sequence is reversed and the underlying data is unusable, the architecture needs to be <strong>rebuilt</strong> before AI deployment. </p></li><li><p>Variations in <strong>data standardisation</strong> across two parts of the same organisation can determine where AI adoption succeeds.</p></li><li><p>Organisations should <strong>audit their data estate</strong> and invest when it is fragmented, inconsistent or not structured for AI interrogation.</p></li></ul>.<p>AI is quite plainly here to stay, and the race to adopt it is fully on. While companies have rushed to implement AI across functions, it may be too early to measure the actual productivity and fiscal gains with confidence. IMA India’s AI Reality Check series aims to investigate the rate and outcomes of this dash for AI adoption, along with the successes, challenges and failures that come with it.</p><p>During our interviews and research, the first major hurdle we observed in the AI adoption race was consistent across sectors and company sizes. Companies needed to already have their digital infrastructure in place, with sanitised and structured data ready to be fed to AI systems, before they could meaningfully deploy the technology. But what does this look like in practice? How have different companies dealt with the issue: And what measures must organisations take before AI algorithms can be fully leveraged? </p>.<h2><strong>Getting the Sequence Right</strong></h2><p>One senior executive described their organisation's AI deployment philosophy as a simple sequence: eliminate first, simplify second, then automate. The company found that deploying AI on an unreformed process does not improve it but rather accelerates whatever is already wrong. A workflow built on inconsistent data, or followed differently across teams, does not get better simply because AI is running on top of it. It gets faster at being broken. The eliminate-simplify-automate sequence helps ensure that by the time AI enters a process, that process is actually worth automating. </p><p>The same logic applies to data architecture at the enterprise level. Before AI can be reliably deployed, the data it will work with needs to be in order. It must be consolidated rather than fragmented across systems, captured consistently across functions, structured in formats AI can interrogate, and governed so that the right people know what exists and where. While this seems like an obvious technical requirement, it is one that most businesses have not yet met, and one that consistently emerged as the key binding constraint in the companies we spoke to.</p>.<h2><strong>The Advantage of Getting There Early</strong></h2><p><strong>HT Media</strong> has achieved its current position in the AI cycle by laying the groundwork far before generative AI became a practical enterprise tool. Over the years, it constructed a centralised data lake and a single customer data platform, designed to ingest data from across the organisation, bring it into one place and make it available in a data-privacy-first manner. Thus, by the time gen AI arrived, the company had already solved the foundational problem that remains a roadblock for many others. </p>.<div><blockquote>“AI can amplify what you are already doing. If you are doing things incorrectly, AI can, in a way, make them worse. You may not be able to realise its benefits.”</blockquote><span class="attribution">— Puneet Jain, CEO – Digital, HT Media</span></div>.<p>As HT Media’s CEO, Puneet Jain explained, this preparatory work significantly reduced the barriers to experimentation, as well as the time taken for integration. The upside of AI deployment was also much higher, mainly because the data infrastructure question had already been answered. In a market where most organisations are discovering the data problem while they are dipping their toes into AI deployment, having dealt with it years earlier is a real structural advantage. </p><p>Another interviewee shared a similar story, but one relevant to a very different sector. For a pharmaceutical company, the consequences of bad AI output carries weight beyond just the commercials. It may negatively impact the health of real people, taint operations and potentially even regulatory standing. When the was building out its AI capability for sales analysis, they created a complete architecture of data before deploying AI against it. All data was structured in an organised way first. If they hadn’t, the output would be worthless. Ensuring the ideal data architecture was integral given the high stakes nature of the business and the caution with which their industry regulates itself. </p>.<h2><strong>The Cost of Getting the Sequence Wrong</strong></h2><p>A<strong> large agro-chemical enterprise’s</strong> experience illustrates what happens when the sequencing goes the other way. Having evaluated proposals for AI agents in supply chain management, it had a legitimate business use-case, as well as relevant tools at its disposal. However, in 20 years it had accumulated data in whatever format people happened to use. Records were incomplete, systems were disconnected, formats were inconsistent across the organisation. When the company assessed what would be required for AI deployment, it became clear that the underlying data was not fit for purpose. As a result, the AI initiative was paused while the organisation reconsidered its approach.</p><p>Rather than attempting to patch the existing data environment, the company chose to rebuild its enterprise architecture through a comprehensive ERP transformation, using the migration as an opportunity to establish cleaner, more consistent and better-governed data. Legacy data was reviewed, restructured and re-entered where necessary. Although this approach required greater investment and extended implementation timelines, leadership concluded that it would provide a far stronger foundation for future AI deployment. 18-24 months after the migration is completed, the company is planning for full AI deployment in demand forecasting and supply chain management. In the meantime, it is working overtime to get the data right. The cost of this situation, in terms of time, capital and deferred AI deployment, is the direct consequence of data architecture not being treated as a strategic priority from the start. </p>.<h2><strong>The Importance of Data Maturity</strong></h2><p>A <strong>global logistics giant</strong> we approached for this study handles approximately 37 million shipments a year and has two major business lines. The first is North American surface transport, where it operates at scale with a high degree of workflow uniformity built on a proprietary global system. The second is international freight forwarding, where it operates across regions with different regulatory environments, different customer conventions and different operational norms that sometimes vary even by branch.</p><p>AI deployment has moved significantly further in terms of surface transport. The company has deployed 30+ agents that handle workflow stages that previously required human intervention for every shipment. In global forwarding, the work is still in progress. According to the company’s India operations lead, surface transport workflows are standardised at scale, which means AI can be trained on the data, and therefore, the results generalise across the operation. Contrastingly, in global forwarding, a pilot built for one region does not transfer to another. Each region requires its own data collection, standardisation and exception mapping before AI can be reliably deployed.</p><p>The technology and AI capability available to both parts of the business is identical. However, in terms of data, one business has decades of structured, uniform data at scale while the other has operational knowledge that is geographically dispersed and inconsistently captured. For any organisation operating across multiple geographies, business units or legacy systems, this should serve as a practical warning. Data maturity will vary across your organisation, and that variation will directly determine where AI delivers and where it stalls</p>.<h2><strong>What this Means for Your Organisation</strong></h2><p>Most Indian enterprises can benefit from AI adoption, but they must first ask themselves whether their data infrastructure can support it. The honest answer, in many cases, is, ‘Not yet’. Data architecture needs to be more than a technical afterthought and investments in building the foundation for AI must be treated as a leadership priority. AI deployment should be sequenced around data readiness. Companies must honestly audit where data is fragmented, inconsistent or not structured for AI integration and treat the findings as a driver for capital allocation. The companies that do this deliberately will find that when the algorithm is finally ready to run, so are they.</p><p>Leaders should keep the following in mind:</p><ol><li><p><strong>Audit your data estate before committing to an AI budget.</strong> Identify where your data resides, who owns it, how consistently it is captured across functions and whether it is sufficiently structured and governed to support AI in its current form.</p></li><li><p><strong>Treat data architecture as a capital allocation decision, not an IT project.</strong> In organisations that have made meaningful progress with AI, investments in data infrastructure were driven at the leadership level rather than delegated solely to technology teams.</p></li><li><p><strong>Sequence your AI roadmap around data readiness, not the other way around.</strong> If your data is not ready, AI will expose and amplify existing weaknesses rather than solve them. Building a robust data foundation may require significant time and investment, but it is considerably less costly than deploying AI on unreliable data.</p></li><li><p><strong>Expect uneven data maturity across the organisation.</strong> If you operate across multiple geographies, business units or legacy systems, assess where data is already standardised and where gaps remain before scaling AI initiatives.</p></li><li><p><strong>Apply the eliminate-simplify-automate principle to every AI use case.</strong> Before asking whether AI can automate a process, determine whether that process is necessary, consistent and well designed. AI should accelerate good processes, not flawed ones.</p></li></ol>
<p><em>The AI Reality Check is an IMA India research series examining how artificial intelligence is changing the operational reality of Indian enterprises. It is based on in-depth interviews with CXOs and senior operational leaders from IMA member companies, supplemented by secondary research. </em></p><p><em>This is the <strong>second of five papers.</strong></em></p>.<p><strong>Executive Summary</strong></p><ul><li><p><strong>Data architecture</strong> is often the first and most consistent barrier to AI deployment </p></li><li><p><strong>Deploying AI</strong> on a broken or inconsistent process does not fix it. It simply makes it faster at being broken.</p></li><li><p>If the sequence is reversed and the underlying data is unusable, the architecture needs to be <strong>rebuilt</strong> before AI deployment. </p></li><li><p>Variations in <strong>data standardisation</strong> across two parts of the same organisation can determine where AI adoption succeeds.</p></li><li><p>Organisations should <strong>audit their data estate</strong> and invest when it is fragmented, inconsistent or not structured for AI interrogation.</p></li></ul>.<p>AI is quite plainly here to stay, and the race to adopt it is fully on. While companies have rushed to implement AI across functions, it may be too early to measure the actual productivity and fiscal gains with confidence. IMA India’s AI Reality Check series aims to investigate the rate and outcomes of this dash for AI adoption, along with the successes, challenges and failures that come with it.</p><p>During our interviews and research, the first major hurdle we observed in the AI adoption race was consistent across sectors and company sizes. Companies needed to already have their digital infrastructure in place, with sanitised and structured data ready to be fed to AI systems, before they could meaningfully deploy the technology. But what does this look like in practice? How have different companies dealt with the issue: And what measures must organisations take before AI algorithms can be fully leveraged? </p>.<h2><strong>Getting the Sequence Right</strong></h2><p>One senior executive described their organisation's AI deployment philosophy as a simple sequence: eliminate first, simplify second, then automate. The company found that deploying AI on an unreformed process does not improve it but rather accelerates whatever is already wrong. A workflow built on inconsistent data, or followed differently across teams, does not get better simply because AI is running on top of it. It gets faster at being broken. The eliminate-simplify-automate sequence helps ensure that by the time AI enters a process, that process is actually worth automating. </p><p>The same logic applies to data architecture at the enterprise level. Before AI can be reliably deployed, the data it will work with needs to be in order. It must be consolidated rather than fragmented across systems, captured consistently across functions, structured in formats AI can interrogate, and governed so that the right people know what exists and where. While this seems like an obvious technical requirement, it is one that most businesses have not yet met, and one that consistently emerged as the key binding constraint in the companies we spoke to.</p>.<h2><strong>The Advantage of Getting There Early</strong></h2><p><strong>HT Media</strong> has achieved its current position in the AI cycle by laying the groundwork far before generative AI became a practical enterprise tool. Over the years, it constructed a centralised data lake and a single customer data platform, designed to ingest data from across the organisation, bring it into one place and make it available in a data-privacy-first manner. Thus, by the time gen AI arrived, the company had already solved the foundational problem that remains a roadblock for many others. </p>.<div><blockquote>“AI can amplify what you are already doing. If you are doing things incorrectly, AI can, in a way, make them worse. You may not be able to realise its benefits.”</blockquote><span class="attribution">— Puneet Jain, CEO – Digital, HT Media</span></div>.<p>As HT Media’s CEO, Puneet Jain explained, this preparatory work significantly reduced the barriers to experimentation, as well as the time taken for integration. The upside of AI deployment was also much higher, mainly because the data infrastructure question had already been answered. In a market where most organisations are discovering the data problem while they are dipping their toes into AI deployment, having dealt with it years earlier is a real structural advantage. </p><p>Another interviewee shared a similar story, but one relevant to a very different sector. For a pharmaceutical company, the consequences of bad AI output carries weight beyond just the commercials. It may negatively impact the health of real people, taint operations and potentially even regulatory standing. When the was building out its AI capability for sales analysis, they created a complete architecture of data before deploying AI against it. All data was structured in an organised way first. If they hadn’t, the output would be worthless. Ensuring the ideal data architecture was integral given the high stakes nature of the business and the caution with which their industry regulates itself. </p>.<h2><strong>The Cost of Getting the Sequence Wrong</strong></h2><p>A<strong> large agro-chemical enterprise’s</strong> experience illustrates what happens when the sequencing goes the other way. Having evaluated proposals for AI agents in supply chain management, it had a legitimate business use-case, as well as relevant tools at its disposal. However, in 20 years it had accumulated data in whatever format people happened to use. Records were incomplete, systems were disconnected, formats were inconsistent across the organisation. When the company assessed what would be required for AI deployment, it became clear that the underlying data was not fit for purpose. As a result, the AI initiative was paused while the organisation reconsidered its approach.</p><p>Rather than attempting to patch the existing data environment, the company chose to rebuild its enterprise architecture through a comprehensive ERP transformation, using the migration as an opportunity to establish cleaner, more consistent and better-governed data. Legacy data was reviewed, restructured and re-entered where necessary. Although this approach required greater investment and extended implementation timelines, leadership concluded that it would provide a far stronger foundation for future AI deployment. 18-24 months after the migration is completed, the company is planning for full AI deployment in demand forecasting and supply chain management. In the meantime, it is working overtime to get the data right. The cost of this situation, in terms of time, capital and deferred AI deployment, is the direct consequence of data architecture not being treated as a strategic priority from the start. </p>.<h2><strong>The Importance of Data Maturity</strong></h2><p>A <strong>global logistics giant</strong> we approached for this study handles approximately 37 million shipments a year and has two major business lines. The first is North American surface transport, where it operates at scale with a high degree of workflow uniformity built on a proprietary global system. The second is international freight forwarding, where it operates across regions with different regulatory environments, different customer conventions and different operational norms that sometimes vary even by branch.</p><p>AI deployment has moved significantly further in terms of surface transport. The company has deployed 30+ agents that handle workflow stages that previously required human intervention for every shipment. In global forwarding, the work is still in progress. According to the company’s India operations lead, surface transport workflows are standardised at scale, which means AI can be trained on the data, and therefore, the results generalise across the operation. Contrastingly, in global forwarding, a pilot built for one region does not transfer to another. Each region requires its own data collection, standardisation and exception mapping before AI can be reliably deployed.</p><p>The technology and AI capability available to both parts of the business is identical. However, in terms of data, one business has decades of structured, uniform data at scale while the other has operational knowledge that is geographically dispersed and inconsistently captured. For any organisation operating across multiple geographies, business units or legacy systems, this should serve as a practical warning. Data maturity will vary across your organisation, and that variation will directly determine where AI delivers and where it stalls</p>.<h2><strong>What this Means for Your Organisation</strong></h2><p>Most Indian enterprises can benefit from AI adoption, but they must first ask themselves whether their data infrastructure can support it. The honest answer, in many cases, is, ‘Not yet’. Data architecture needs to be more than a technical afterthought and investments in building the foundation for AI must be treated as a leadership priority. AI deployment should be sequenced around data readiness. Companies must honestly audit where data is fragmented, inconsistent or not structured for AI integration and treat the findings as a driver for capital allocation. The companies that do this deliberately will find that when the algorithm is finally ready to run, so are they.</p><p>Leaders should keep the following in mind:</p><ol><li><p><strong>Audit your data estate before committing to an AI budget.</strong> Identify where your data resides, who owns it, how consistently it is captured across functions and whether it is sufficiently structured and governed to support AI in its current form.</p></li><li><p><strong>Treat data architecture as a capital allocation decision, not an IT project.</strong> In organisations that have made meaningful progress with AI, investments in data infrastructure were driven at the leadership level rather than delegated solely to technology teams.</p></li><li><p><strong>Sequence your AI roadmap around data readiness, not the other way around.</strong> If your data is not ready, AI will expose and amplify existing weaknesses rather than solve them. Building a robust data foundation may require significant time and investment, but it is considerably less costly than deploying AI on unreliable data.</p></li><li><p><strong>Expect uneven data maturity across the organisation.</strong> If you operate across multiple geographies, business units or legacy systems, assess where data is already standardised and where gaps remain before scaling AI initiatives.</p></li><li><p><strong>Apply the eliminate-simplify-automate principle to every AI use case.</strong> Before asking whether AI can automate a process, determine whether that process is necessary, consistent and well designed. AI should accelerate good processes, not flawed ones.</p></li></ol>
