Over three days at the ETAutoTech Summit in Bengaluru this June, I listened to OEM engineering leaders, regulators SDV platform builders and AI diagnostics startups present on subjects that appeared, on the agenda, to have very little to do with one another.
Platform commonality. Homologation standards. Edge AI architecture. Design identity in the EV era. ADAS scenario libraries.
By the second afternoon, it was difficult to ignore that all of them were describing the same problem from different seats in the room.
The number that framed everything
One of the speakers opened the Design-to-Manufacture Congress with a figure that set the tone for everything that followed: new product development programmes in the Indian automotive industry are running nine to fifteen months behind their original timelines.
His diagnosis was more uncomfortable than the number itself. These delays are not primarily execution failures that accumulate as a programme runs. They are structural. Many projects, he argued, are effectively delayed before they formally begin, because organisations have not honestly defined how many concurrent programmes their engineering capacity can actually absorb.
The remedy he proposed was disciplined maturity gating: releasing drawings only when 3D data has reached genuine maturity, with structured reviews at each stage rather than one consolidated check near the end.
For anyone who has worked inside a Japanese OEM development system, this is familiar territory. In my years with Honda R&D, design maturity was assessed in explicit increments - 25%, 50%, 75%, 100% — with a review gate at each threshold. Drawings issued only as the design approached full maturity. The discipline was in the mechanism that kept problems small enough to fix.
The same problem - Defined by other stakeholders
What made the summit genuinely interesting was watching four other speakers arrive at the same destination by entirely different routes.
Director of ARAI, presented a framework mapping the summit's themes onto ARAI's national mandate. Under digital engineering and simulation, the industry outcome he identified was faster development with fewer late-stage failures.
Sonatus, presenting on edge AI in modern E/E architecture, described their diagnostic agent's value proposition as accelerating pre-SOP validation by surfacing issues earlier in development — explicitly to reduce costly late-stage fixes.
And on the manufacturing panel, the panlelist described the industry's current limitation plainly: virtual validation today is largely confined to design validation. The goal — making it the single source of truth across the entire development cycle, from design through production — he estimated to be two to three years away.
India's particular version of the problem
The ADAS validation panel gave this general problem a specifically Indian shape, and it is not a comfortable one.
The panellists were candid that scenario coverage remains the central gap. Indian road conditions generate a category of edge case — cattle, stray animals, children entering carriageways — that global scenario libraries were never built to represent. IPG Automotive described building over a hundred India-specific road scenarios for simulation, while acknowledging the work remains very much in progress. NATRAX noted that representative proving grounds are a recent phenomenon in this country, and that whilst expressways offer structured, consistent conditions, city driving introduces variability in both infrastructure and driver behaviour that is far harder to model.
KPIT's framing was the one I keep returning to: in India, the edge case is the normal case. Which means importing a global validation framework and adjusting it at the margins is not a strategy. The learning has to be done for India, from India.
There was also a quieter observation, from ICAT, that ought to concern the industry more than it appears to: much of the ADAS software deployed on Indian roads originates outside India, validated against conditions that do not resemble ours.
Talent is not the constraint
One of the key observations was made by a foreign delegate:
she observed that India has the talent pool for development. What it lacks is platform ownership.
This distinction deserves more attention than it received. The Indian engineering ecosystem has spent two decades proving it can execute work specified elsewhere. Of roughly two thousand global capability centres operating in this country, only around seventy are automotive — and the ambition of those seventy is visibly shifting from supporting global programmes to originating them.
But originating a programme requires something different from executing one. It requires owning the maturity gates. Owning the scenario library. Owning the validation criteria and the authority to say a design is not ready.
That is the gap between having engineers and having engineering ownership. And it is precisely where the late-discovery problem lives, because a team that executes to someone else's specification cannot catch a problem that the specification failed to anticipate.
What this means practically
If the diagnosis across several independent speakers converges, the response probably should too.
Gate maturity explicitly, not implicitly. Structured review thresholds with real authority to hold a release are unglamorous and they work. The industry knew this before it had AI tooling.
Define concurrent capacity honestly. The point that programmes are delayed before they start is the least discussed and most consequential observation of the summit. Overloaded pipelines guarantee late discovery regardless of how good the tooling is.
Build India-specific validation as owned capability, not adapted import. Scenario libraries reflecting Indian conditions are a national engineering asset. They cannot be licensed in.
Treat AI diagnostics as complement, not substitute. The context graphs and agentic tooling demonstrated at the summit are genuinely useful. They accelerate root cause analysis. They do not replace the engineering discipline that prevents the root cause.
The nine-to-fifteen month delay figure is not primarily a tooling problem. It is a discipline and ownership problem.
Swapnil Gawade is Chief Operating Officer of Goken India, an engineering services company with Japanese roots working with automotive OEMs and Tier 1 suppliers across India, Japan, the United States, and South Korea. He spent over two decades with Honda R&D Americas prior to joining Goken.
Explore Our Capabilities
[View All Insights & Case Studies]
© Goken. All rights reserved. Client identity has been kept confidential per engagement agreement.