India’s renewable energy sector is beginning to integrate artificial intelligence (AI) tools into project planning, bidding and construction management. However, adoption remains uneven, tools remain fragmented and the path from experimentation to full-scale deployment is still being charted. At Renewable Watch’s second edition of the “AI in Renewables” conference, senior executives from the renewable energy sector shared on-the-ground perspectives on what AI can and cannot yet do during project development, the primary use cases, persistent data gaps and the irreplaceable role of human judgement. Edited excerpts…
Sanjeev Sharma
For a developer working across solar, wind, battery storage and green hydrogen, the economics of AI adoption are not in question. Even the most expensive licensed tools cost very little relative to the value they generate, and the benefits are substantial. The challenge is not whether to use AI, but finding the right tools for a sector where data remains scarce, and the technology ecosystem is still maturing.
At the bidding stage, there is a high use case for AI. Tenders today routinely run to 2,000 pages, covering commercial terms, legal implications, technical specifications and compliance requirements simultaneously. Reading through that volume carefully is itself a time-consuming exercise, and the fear of missing something critical is ever-present. The consequences of a lapse can be severe. Bids can be lost if a single fine print is not carefully read or its implications are not fully assessed. AI tools dramatically reduce this risk by parsing large documents quickly, flagging compliance requirements and preparing bid submission checklists systematically.
Beyond document review, AI plays a critical role in levellised cost of hydrogen modelling. Bidding on a single cost model without understanding whether it is genuinely competitive or merely bankable is a significant risk. AI enables dynamic scenario modelling that balances the winning factor against financial viability, and this analysis must continue all the way through to the final investment decision stage, even after a contract has been won on the basis of the lowest cost.
The tools available today are fragmented across different applications and none can take a developer from start to finish. Five years ago, there were no tools at all, which is some consolation, but a wide gap remains. A fundamental problem is data. AI and machine learning derive their value from analysing large data sets to generate predictive outputs, and the green hydrogen sector simply does not have that foundation yet. Information on electrolyser performance, life cycle costs and the capacity utilisation of hybrid solar-wind systems operating as integrated models is scarce, and what exists is rarely available on open platforms. Sourcing that data is itself a challenge before any meaningful analysis can begin. As a result, even well-trained tools are not currently fully accurate. They still provide a useful case at the bidding stage, through the engineering, procurement and construction (EPC) phase and in risk mitigation, but absolute reliance on these tools without human validation would be a serious error.
On state-level regulatory complexity, AI can help map requirements across states and provide useful design insight. For green hydrogen specifically, the bigger structural challenge is meeting the European renewable fuels of non-biological origin zoning requirement, which mandates that power generation must occur within the same zone as hydrogen production. India operates a unified national grid, and its zoning framework was conceptualised with a different logic altogether. This misalignment has become a serious constraint for Indian exporters targeting European markets. Several discussions are ongoing, but the issue remains unresolved. The problem is further compounded by the fact that land and water availability constraints often make it physically impossible to co-locate renewable generation with hydrogen production facilities, adding another layer of complexity that no regulatory alignment can easily solve.
AI has transformed engineering timelines. For instance, the time taken to make the process flow diagrams has reduced. Time taken during the pre-FEED and FEED design stages has also compressed. Engineering teams have embraced AI tools wholeheartedly, treating them not as a threat to their expertise but as an augmentation that allows deeper engagement with process design rather than being consumed by its mechanics.
Reliability, however, remains a genuine concern. For solar and wind, where data is now reasonably well established, AI tools can generate dependable outputs. For green hydrogen, that data foundation does not yet exist, and no project decision should be driven by AI outputs alone without a thorough human review. The risk of being guided into significant losses by over-reliance on tools whose outputs are based on incomplete data is real, and a disciplined human check at every stage is not optional but essential.
Rajneesh Shrotriya
A major share of cost optimisation work still runs on manual Excel sheets. In EPC contracting, cost optimisation requires going through individual line items one by one with a level of granularity that no single AI tool currently available can replicate comprehensively at project scale.
Sterling and Wilson currently uses several software tools across different functions, generating site layouts, preparing technical specifications, building project schedules, modelling cable and conduit runs, and estimating the best generation output achievable from a given set of coordinates. These tools allow the team to present clients with several scenario options across different cost, timeline and bankability parameters. But integrating all of that into a single optimised recommendation still requires substantial manual effort, and the company continues to look for tools that can bring these functions together meaningfully.
A concern is the risk of AI producing cookie-cutter outputs. Every project is a unique, tailor-made undertaking. Sites differ in soil conditions, corrosion environments, wind loading, terrain and a range of other parameters that materially affect design and cost. If four different vendors are each asked to submit a solution and all four arrive at the same answer because they used the same tool, that is a failure of the process and not a sign of its success. An experienced engineer’s understanding of what a specific client at a specific site actually needs cannot be substituted by a tool, and over-reliance on standardised outputs risks undermining the very quality of thinking that clients are paying for.
In India’s multi-state regulatory environment, project tracking is the area where AI could add the most immediate practical value, but where the industry is currently most exposed. Planning happens meticulously on software, but delays accumulate once construction begins and teams start interacting with vendors, owner’s engineers and state authorities. The root cause is often not poor planning but poor carry-forward of state-specific knowledge. Teams moving from a project in Maharashtra to one in Rajasthan sometimes carry assumptions that no longer apply. The distribution company’s requirements differ, the government guidelines vary and state load despatch centre requirements are not uniform across states. What is needed is a system that raises a real-time red flag when a team is about to apply the wrong state’s regulatory logic to a new project, ensuring that past learnings become an asset rather than a liability.
Where AI has delivered tangible results for Sterling and Wilson Renewable Energy is in construction-phase tracking. Of ongoing projects, some have deployed tools that superimpose real-time site progress onto project drawings, giving project managers an accurate picture of what is happening on the ground without having to call several different people, consolidate daily reports and then issue instructions into a situation that has already moved on. Given that solar projects operate on shorter timelines compared to larger infrastructure projects, this speed of feedback is critical. The tools deliver photographic confirmation and result parameters directly to the project manager’s device within a short time of a scheduled activity being due. That said, converting real-time information into on-ground results remains genuinely difficult in challenging environments. Desert conditions in Rajasthan and the waterlogged terrain around Khavda in Gujarat for some months of the year present realities that no software dashboard can fully address.
Formal cost savings from these tools have not yet been calculated, as the current focus remains on identifying and correcting gaps and saving time. While our own site teams have embraced these tools with enthusiasm, actively reading up on capabilities and suggesting improvements, vendors and subcontractors remain resistant, as faster and more accurate tracking will catch their lapses before they can be concealed.
