In June 2026, the ministry of new and renewable energy launched SUNtosh, a WhatsApp based assistant that helps households estimate rooftop solar systems using just two inputs: Sanctioned electricity load and the previous month's electricity bill. Within minutes, users receive recommendations on system size, available subsidies and expected savings, without downloading an app or understanding the technical calculations behind solar design.
The importance of SUNtosh extends beyond rooftop solar. It signals a broader change in how people expect to interact with technology. For decades, software expected users to learn its interface. Today, users expect software to understand their intent.
The same expectation is beginning to reshape industrial infrastructure. India's infrastructure is becoming increasingly connected. As of June 2026, the country had more than 162 GW of installed solar capacity, while the Smart Cities Mission had completed 8,067 projects worth ₹1.64 lakh crore, with Integrated Command and Control Centres using technologies such as artificial intelligence and the Internet of Things. Every solar plant, pumping station, manufacturing facility, commercial building and data centre now generates a continuous stream of operational information through sensors, meters and intelligent controllers.
Yet despite collecting more data than ever before, many organisations still struggle to make faster operational decisions. The reason is surprisingly simple. Infrastructure no longer has a data problem. It has a decision problem.
{{/usCountry}}Yet despite collecting more data than ever before, many organisations still struggle to make faster operational decisions. The reason is surprisingly simple. Infrastructure no longer has a data problem. It has a decision problem.
{{/usCountry}}Most operators are not overwhelmed because information is unavailable. They are overwhelmed because it exists across multiple dashboards, maintenance records, equipment manuals, alarm systems and spreadsheets. A conventional dashboard can display thousands of readings, but it rarely explains why something changed, whether the issue is significant or what action should be taken first.
Consider a portfolio of just 50 solar sites. Reporting only 15 operating parameters every five minutes generates more than 216,000 readings every day, even before weather information, maintenance records, inverter alarms and energy forecasts are considered. The challenge is no longer collecting data. It is converting that data into timely decisions.
This is where generative AI represents a genuine shift rather than another software upgrade.
Instead of searching through multiple applications, an operator could simply ask, Why did output fall at Site 18? Which three assets require immediate attention today? Has this fault occurred before? Rather than presenting isolated data points, AI can combine live telemetry, maintenance history, equipment documentation and operating procedures to provide a probable explanation, supporting evidence and recommended next steps.
The value lies not in generating new information but in making existing information easier to interpret. That capability has significant business implications. A 2024 industrial study estimated that unplanned downtime costs the world's 500 largest companies nearly $1.4 trillion annually, equivalent to around 11% of their revenue. In many situations, the earliest signs of a failure already exist in machine data. The delay occurs because someone still has to recognise the pattern, interpret its importance and decide whether intervention is necessary.
The challenge becomes even greater as infrastructure expands faster than the availability of skilled operational talent. A 2026 Quess Corp study reported a 73% shortage in core IT operations roles within India's data centre sector, while AI operations talent achieved a supply sufficiency score of only 47 out of 100. Experienced engineers are increasingly expected to supervise larger and more geographically distributed infrastructure with little corresponding growth in specialist teams.
Generative AI can help bridge that gap by extending human expertise rather than replacing it. Routine operational questions can be answered immediately, recurring faults can be recognised faster and engineers can focus their attention on exceptions instead of reviewing every asset individually. In distributed monitoring projects, issues that once resulted in precautionary site visits can often be diagnosed remotely before technicians are dispatched, reducing both response times and operational costs.
There is another advantage that receives far less attention. Every infrastructure organisation depends on experienced engineers whose judgement has been developed through years of operating complex assets. Much of that expertise exists only in people rather than documentation. When those individuals retire or move on, organisations risk losing practical knowledge that is difficult to replace. AI creates an opportunity to preserve that institutional knowledge by combining engineering experience with live operational data, making expertise available across the organisation instead of limiting it to a handful of specialists.
None of this suggests that infrastructure should become autonomous. Industrial environments demand a level of reliability that consumer applications never have to achieve. An incorrect restaurant recommendation is inconvenient. An incorrect recommendation at a power plant, manufacturing facility or water treatment plant can damage equipment, interrupt essential services and create genuine safety risks. For that reason, generative AI should function as a decision support layer rather than a decision maker. Existing automation must continue enforcing engineering rules, while human approval remains essential for actions involving safety, compliance and critical operational changes.
Trust will ultimately determine adoption. Operators need to understand not only what the AI recommends but why. Recommendations should be supported by source data, confidence levels and clear reasoning. When information is incomplete or contradictory, uncertainty should be communicated honestly rather than hidden behind confident sounding answers.
Industry adoption is accelerating.
Generative AI will not replace engineers, sensors or control systems. Its greatest contribution will be helping people make better decisions using the information they already have. The interface may become as simple as sending a message, but behind every answer must still sit reliable data, secure infrastructure, engineering discipline and accountable people. The interface can become simpler. The responsibility for operating critical infrastructure safely cannot.
(The views expressed are personal)
This article is authored by Visat Patel, CEO and co-founder, Ubiqedge.