Artificial intelligence is increasingly being discussed as a tool for sustainability, emissions reduction, supply-chain oversight, ESG reporting, and operational efficiency. But for many companies, the practical question is not simply whether AI can help. The more important question is where it can create measurable value, what data it needs, and how businesses should manage the risks that come with using it.

In a recent Net Zero Compare conversation, Karol Kaczmarek spoke with Manuj Aggarwal, founder and Chief Innovation Officer of TetraNoodle Technologies. Aggarwal has worked in technology and artificial intelligence for several decades, with experience across enterprise AI systems, automation, data-driven decision-making, and supply-chain applications. The discussion focused on how AI can be used in sustainability work without turning it into another layer of complexity, cost, or unsupported claims.

For sustainability teams, ESG leaders, procurement teams, and business decision-makers, the core message was clear: AI can be useful, but only when it is connected to a defined business problem, reliable data, clear governance, and measurable outcomes. Used well, it can support better decisions and augment human expertise. Used carelessly, it can amplify weak data, unclear priorities, and poorly designed processes.

🎥 Watch the Full Interview: The full Net Zero Compare interview with Manuj Aggarwal explores the practical role of AI in sustainability, reporting, supply chains, and resource efficiency. The conversation is especially useful for readers who want more context on how AI systems are implemented, why many pilots fail, and what companies should consider before applying AI to emissions, ESG, or operational data. Watching the full discussion also provides more nuance around data quality, private AI systems, human oversight, and AI’s own environmental footprint. These are areas where short summaries can miss important details, particularly when companies are deciding whether AI is appropriate for reporting, compliance, procurement, or operational efficiency projects.

AI Is Not a Shortcut Around Poor Processes

Aggarwal began by connecting his early experience on a factory floor with the way he thinks about waste today. In his view, waste is often created not only by inefficient machines or materials, but by gaps between teams, processes, priorities, and decision-making. A manufacturing process may involve multiple steps, each managed by different people or departments. One team may focus only on completing its own task as quickly as possible, while another team manages the next stage of production. If those teams do not coordinate well, businesses can lose time, energy, materials, and money.

This is where AI may help. Aggarwal described AI as a tool that can identify friction points across processes, connect information from different teams, and suggest ways to reduce unnecessary steps. In some cases, AI may make autonomous decisions. In other cases, it may support human teams by showing where processes can be simplified or where resources are being wasted. The useful application is not simply automation for its own sake, but better visibility across workflows that are often fragmented.

However, this does not mean that companies should add AI to every process. A poorly designed workflow will not automatically improve because AI is added on top of it. Before investing in AI, companies need to understand the process, define the objective, and decide what success would actually look like. In sustainability, that means connecting AI projects to measurable outcomes such as lower energy use, reduced material waste, improved reporting accuracy, better supplier oversight, or lower operating costs.

Companies Need to Define the Business Objective First

One of the strongest points from the conversation was that companies should not begin with the technology. They should begin with the business problem. For sustainability teams, this means asking what the organization is trying to improve. The goal may be to reduce energy use, lower emissions, improve compliance, reduce costs, improve supply-chain visibility, or support reporting. Each of these objectives may require different data, workflows, controls, and levels of human oversight.

Aggarwal noted that sustainability departments often have to work across finance, operations, procurement, logistics, and compliance. Each department has its own priorities. Finance may focus on budget limits, operations may focus on practical constraints, procurement may focus on supplier availability, and compliance teams may focus on documentation and risk. Sustainability teams often have to bring these perspectives together before they can make progress on emissions, resource efficiency, or regulatory obligations.

AI can help by reviewing information across departments and identifying options that may not be obvious to one team working alone. For example, an AI system may identify that a specific equipment upgrade appears costly at first, but could create measurable savings over time. It may also help compare competing priorities, such as cost reduction, regulatory compliance, and emissions reduction. The key is that companies must define what they want to achieve before they evaluate whether AI is the right tool.

The Best Use Cases Depend on the Company

AI is often linked to energy optimization, predictive maintenance, waste reduction, supply-chain mapping, transport planning, procurement analysis, and emissions forecasting. These areas can be relevant, but Aggarwal warned against treating AI use cases as one-size-fits-all solutions. A company should not copy another organization’s AI project simply because it produced results elsewhere. Every company has different operations, data quality, supply chains, constraints, budgets, and priorities.

Instead, Aggarwal recommended looking at the company’s own processes and identifying where AI could create the best return. If a business wants to reduce emissions over the next 12 months, it should first look at its major operational areas and determine which processes offer the strongest opportunity for improvement. For one company, the best starting point may be transport route optimization. For another, it may be energy use in manufacturing. For another, it may be better reporting processes or supplier data collection.

The practical lesson is that AI should be used where it can support a measurable decision. It should not be adopted because it sounds innovative or because competitors are experimenting with it. Sustainability-related AI projects need a business case, a baseline, and a clear reason for why AI is better suited to the task than conventional software, process improvement, or manual review.

Data Quality Determines How Useful AI Can Be

For sustainability and ESG teams, data quality remains one of the biggest barriers to effective AI use. Many companies still work with fragmented systems, missing supplier data, incomplete utility records, inconsistent emissions factors, and estimated Scope 3 information. Aggarwal compared AI to a vehicle and data to the fuel that powers it. If the data is poor, incomplete, or unreliable, the AI system will not produce dependable results.

That does not mean companies need perfect data before AI becomes useful. AI can help identify missing information, flag gaps, and show where the company needs better inputs. For example, if a company wants AI to optimize supply-chain routes, the system may point out that it is missing warehouse addresses, distribution center locations, vehicle schedules, or inventory requirements. This can make AI useful before full optimization begins, because the first role of AI may be to show companies where their data is incomplete and what needs to be fixed.

However, companies must distinguish between operational estimates and defensible reported figures. AI may help fill gaps for internal planning, but sustainability reports, compliance filings, targets, and investment decisions require a higher level of validation. Numbers that appear precise can still be unreliable if the underlying data is incomplete or if assumptions are not reviewed by qualified professionals.

Supply-Chain Tracking Can Improve, But Original Data Still Matters

The conversation also covered supply-chain traceability, including Aggarwal’s previous work involving materials such as copper, silver, and gold. In supply chains, especially those involving high-value materials, companies may face theft, loss, inaccurate reporting, or incomplete records. Aggarwal explained that technology can help monitor quantity, quality, moisture levels, temperature, shipment status, and other variables as goods move through the supply chain.

Blockchain can also help preserve records because entries cannot be easily changed after they are added. This can reduce disputes and improve visibility over what happened at each stage of the process. In a supply chain where goods move across multiple locations, companies can use digital records to understand when materials were shipped, how quantities changed, and whether reported conditions match the available evidence.

However, technology has limits. AI and blockchain can help preserve, monitor, and analyze information, but they cannot automatically prove that the original information was truthful. If supplier data is inaccurate, self-reported, or incomplete, technology can reduce some risks but cannot remove the need for verification, governance, and human accountability. For procurement and sustainability teams, digital traceability can improve oversight, but it is not a substitute for supplier due diligence.

Public AI Tools Create Privacy and Governance Risks

Another major theme was the difference between public AI tools and private AI systems built around an organization’s own data, knowledge, and workflows. For sustainability, ESG, and procurement teams, the risks are significant. Supplier information, employee data, internal energy-use data, draft reporting documents, financial assumptions, and commercially sensitive business information should be handled carefully. Companies need clear rules about what can and cannot be entered into public AI tools.

Aggarwal argued that public AI tools also create a less obvious risk: they may make decision-making more generic. If many companies ask the same public AI systems for strategy, planning, or business recommendations, they may receive similar answers. Over time, this can weaken the company’s own decision-making style, priorities, and institutional knowledge. In areas such as sustainability strategy, procurement, and compliance, this matters because companies need advice and analysis that reflect their own operations, risks, and goals.

Private organizational AI can reduce these risks, but its value does not come only from placing a model in a private environment. Its real value comes from being connected to the organization’s internal knowledge, processes, historical decisions, domain expertise, and operating context. When designed properly, private AI can help teams work with information that reflects how the business actually operates, while keeping sensitive data within controlled environments.

For companies using AI in ESG or sustainability work, governance should not be treated as an afterthought. It should be part of the design from the beginning, especially when AI is used to support decisions involving suppliers, budgets, public disclosures, or regulatory compliance.

Human Oversight Still Matters in Reporting and Compliance

Many companies are already experimenting with AI to collect ESG data, review documents, map regulations, draft disclosures, and flag inconsistencies. These applications can reduce manual work, but they cannot remove responsibility from the organization. AI should be understood as an augmentation tool that supports people, not as a replacement for professional judgment, domain expertise, or accountability.

Aggarwal described a model where different AI agents perform different tasks. One AI agent may specialize in collecting data from relevant systems or documents. Another may validate inputs, identify missing information, or check whether the data appears consistent. A separate agent may analyze the information and prepare reporting outputs, while another may support quality assurance by reviewing the work and flagging possible errors before human review.

This type of multi-agent structure can make reporting and compliance teams more efficient because it mirrors the way many organizations already separate data collection, analysis, review, and approval. AI can prepare materials, surface inconsistencies, and reduce repetitive work. But professionals still need to validate the sources, review the assumptions, check calculations, and approve the final disclosure. The final review should remain with qualified people who understand the organization, the regulatory context, and the consequences of inaccurate reporting.

This is especially important when AI outputs affect regulatory compliance, public reporting, supplier decisions, or capital allocation. Companies may use AI to support the process, but they remain responsible for the final decision. In practical terms, this means AI should be used to assist reporting and compliance teams, not to replace accountability.

AI’s Environmental Footprint Must Be Part of the Business Case

AI systems require data centers, electricity, cooling, hardware, and specialized computing infrastructure. This means companies using AI for sustainability need to consider whether the system creates a net benefit. A company should compare the resources consumed by AI with the benefits it creates downstream, such as reduced fuel use, better transport planning, improved manufacturing efficiency, lower material waste, or more efficient reporting processes.

Aggarwal noted that AI usage costs, such as token-based invoices from AI providers, can serve as one practical proxy for resource consumption. If a company is spending heavily on AI, it should also ask what operational, financial, and environmental returns the system is creating. A sustainability project should not be justified only by the fact that it uses AI. It should be evaluated against the same practical questions as any other investment.

This does not mean every AI project must produce immediate environmental savings. Some projects may first improve data quality, reduce manual work, or strengthen compliance readiness. But companies should still avoid using AI simply because it is available. The business case should include cost, measurable operational value, implementation effort, and the environmental trade-off.

Why AI Pilots Often Fail

Many AI projects succeed as demonstrations but fail to become part of normal business operations. According to Aggarwal, the reasons are usually connected. Companies may lack clean data, fail to define the objective clearly, or struggle to align technical teams with business teams. Employees may not adopt the system, ownership may be unclear, implementation costs may be underestimated, and the company may not have a realistic plan to move from proof of concept to production.

A common problem is that businesses try to use AI only to automate what humans already do, instead of rethinking how work could be done differently. This limits the value of the technology and often leads to projects that look promising in testing but do not change day-to-day operations. In sustainability work, this risk is especially relevant because the work is often cross-functional. A pilot may involve ESG teams, finance, operations, procurement, IT, and legal or compliance teams, each with different priorities.

Successful AI adoption is therefore not only a technology challenge. It is also an organizational transformation challenge. Companies need governance, process redesign, leadership support, employee readiness, and clear ownership. If the organization is not prepared to change how decisions are made, how data is managed, and how teams collaborate, even a technically strong AI system may remain stuck at the pilot stage.

Before approving an AI pilot, business leaders should define the specific objective, the baseline for measurement, the data required, the project owner, the justified budget, and the route into normal operations if the pilot succeeds. They should also decide how progress will be reviewed and what adjustments will be made if the project is not delivering the expected results. Without those decisions, even a technically successful AI pilot may fail to create real business value.

A Practical Starting Point for Mid-Sized Companies

For mid-sized companies facing limited AI expertise, incomplete Scope 3 data, growing reporting requirements, and pressure to reduce both costs and emissions, Aggarwal advised against adopting generic AI templates. The first step is to assess the company’s own operations, priorities, risks, and data. From there, companies can identify two or three practical areas where AI could be tested over a three-to-six-month period.

A realistic first project should be limited in scope, tied to a measurable outcome, and connected to a clear business need. For example, a company might use AI to identify missing supplier data, improve route planning, reduce energy waste in one facility, or support document review for ESG reporting. These projects are narrow enough to manage but useful enough to create learning and measurable results.

The goal should not be to become fully AI-enabled overnight. Aggarwal emphasized that AI adoption is a long-term process. Companies need to think strategically about how AI may change their industry over time, while still beginning with practical projects that match their current resources and capabilities. The most useful first step is not necessarily the most ambitious one. It is the one that helps the company build confidence, improve its data, and create a foundation for better decisions.

Conclusion

The conversation with Manuj Aggarwal showed that AI can play a useful role in sustainability, but only when companies use it with discipline. AI can help connect fragmented information, identify operational waste, support reporting, improve supply-chain visibility, and reduce manual work. It can also help companies understand where their data is incomplete and which processes may offer the strongest return on improvement. At the same time, AI is not a shortcut around poor data, unclear objectives, weak governance, or fragmented internal ownership. Companies need to define the problem, validate the data, protect sensitive information, involve qualified professionals, and measure whether the system creates real value. This is particularly important in sustainability, where decisions may affect emissions reporting, procurement, compliance, investment, and public credibility.

For sustainability professionals and business decision makers, the practical takeaway is straightforward: AI should be treated as a tool for better decisions, not as a replacement for strategy, accountability, or verification. The companies most likely to benefit are those that start with a real operational or reporting problem, build around reliable data, and keep human judgment firmly in the process.

Learn more about Manuj Aggarwal and his work through TetraNoodle Technologies, and AI Merge.