TAFE’s Dilraj Singh Gandhi, EVP Digital & AI, has been building the intelligence layer behind one of India’s largest tractor manufacturers, where AI now shapes farmer decisions long before they reach a showroom, and where legacy data and processes are being reworked for scale. In a discussion with CIONOW, he unpacks what it actually takes to move beyond basic digital transformation.
Manufacturing organizations are moving from basic automation toward Industry 4.0. How can an organization move beyond foundational digital transformation and build a fully intelligent, AI-enabled enterprise?
Organizations must work on four elements together. First, they must focus on people. Employees need to develop the right digital mindset, but mindset alone is not sufficient. Their capabilities must also be upgraded through investment in training, time and resources.
Second, organizations must demonstrate tangible benefits. The benefits may initially be small, but they must be visible. People should be able to see the tangible business result that a digital initiative has brought about. If the outcome we are looking for doesn’t materialise the organisation needs to go and work out what has gone wrong or else it will fail.
Third, digital transformation requires continuity. Organizations need a pipeline of use cases rather than waiting for one use case to be completed before beginning another. Momentum is critical because digital transformation involves inertia, change management and sustained adoption.
Finally, the use case must solve a genuine problem on the ground. Too often, use cases are created at the top for the convenience of a few individuals but do not address operational realities. Organizations must ask whether they are solving an actual problem, a perceived problem or merely a symptom. Digital succeeds when people adopt and use it even when nobody is monitoring them. That happens only when the underlying problem is genuinely solved.
A farmer’s relationship with an original equipment manufacturer now extends beyond the initial purchase across product discovery, equipment usage, maintenance and after-sales support. Where can AI create the greatest value for farmers across this journey?
Digital information has penetrated rural India deeply. Today, a farmer considering an equipment purchase may conduct extensive research through YouTube, Google, social media and other digital channels. In many cases, the farmer may have already made around 70% of the purchase decision before entering the showroom.
AI can aid farmers in making decisions at every step of the business, for example for selecting a tractor; an AI-enabled system could analyze horse power requirements, type of soil, crops growing, applications that need to be satisfied, tools that will be used, and the overall operational environment for the farmer to determine if he needs that particular tractor, a more horsepower enabled tractor or perhaps an implement suitable for the purpose. AI enables better financial access. Collaborating with financial institutions an AI could have wider parameters that are analyzed to calculate credit scores, and suggest available finance options.
Today, assessment of credibility usually is focused upon existing credit scores like CIBIL or Equifax; additional parameters could enable customers with no credit history to obtain loans.
AI could even help in taking the decision if the tractor needs to be purchased or rented. The nature of the challenge will mean that on some occasions (e.g. dictated by the season), an AI enabled app might recommend rental for the season and purchase for next season.
Besides farmer experience, where do you think AI would create the greatest operational impact?
I do not see there being one single ‘area’, but would instead choose whichever one had the greatest inefficiency, the greatest lag in decisions, the highest manual intervention rate and where the data quality was relatively poor.AI can potentially create value across every part of the manufacturing value chain, although the type of AI used may vary.
In the backend, organizations typically have structured and tabular data in ERP systems. Traditional or predictive AI can be useful for forecasting, scenario planning and what-if analysis. At the customer-facing end, generative AI and agentic AI can play a more significant role. Agentic AI could support finance processes, procure-to-pay activities and purchase-order workflows.
On the quality side, agentic AI combined with computer vision could help inspect products and recommend whether they should be accepted or rejected. The final judgment would remain with the relevant production or factory head, but AI could support faster and more informed decisions.
Organizations should invest across all three areas, even if investment levels differ. Otherwise, they risk creating a digital divide within the enterprise. One function may become highly intelligent through AI augmentation while another remains dependent on basic automation.
How difficult is it for CIOs to secure budgets for AI projects, especially when the business outcomes and ROI are not immediately clear?
A CIO should not try to secure an AI budget while working alone. The CIO’s role needs to be redefined. It is no longer sufficient to request a budget, implement a technology and report the outcome. AI is fundamentally about business. The CIO must act as a storyteller, evangelist and seller who helps business leaders understand the opportunity and encourages them to invest. The business should ultimately take the proposal to the board because AI must be embedded in the business and linked to business outcomes. Just sprinkling it over the application like the IT department would do is not an option.
CIO has to get engaged, find business problems (with business heads) along with a quick business relevant pilot and present the first value delivered and use relevant subject matter expert or an example from other enterprises and instill trust in business people.
Legacy organizations often have data scattered across silos, with problems relating to data quality, completeness and accessibility. How should a legacy enterprise begin its AI journey?
Data is important, but it should not become a reason to delay transformation. Legacy organizations do face data problems, including unverified data, incomplete data and the collection of irrelevant or outdated parameters. But those limitations needn’t be showstoppers. It’s easy for organizations to get stuck in a “chicken-and-egg” loop: they must have flawless data to start their AI journey, but they only will obtain high quality data if they begin work on viable AI use cases. Instead, the best approach is to choose a high value-use case and start with existing data, even if fragmented and unverified, possibly synthesizing missing data to develop an initial model.
Simultaneously, the organization can put in place the processes to create new data, clean existing data, vet accuracy, enhance completeness, and put data in transit.
Initial results might be noisy or imperfect and AI could make mistakes, err, or hallucinate. This they will know and must plan to manage. As the organization moves ahead, both data and models should improve. Transformation simply can’t wait. Indeed, it may be only through transformation that perfection becomes possible at all.
Looking ahead, what will define an intelligent enterprise in the agricultural and manufacturing sectors over the next five years?
The intelligent enterprise will be defined by its ability to break traditional linearity and place the customer at the centre of the operating model. At present, a customer complaint often follows a long chain: the customer contacts a call centre, dealer or salesperson; the dealer escalates the issue to sales or service; the service function approaches production; production involves quality; quality may involve R&D, procurement or another department; and the organization eventually responds to the customer.
What a smart enterprise could actually do, is to sever the line of complaint. When a customer reports an issue to someone, the details pertaining to the complaint are forwarded automatically to the person who needs to provide the resolution. Not only that, the customer needs to be informed immediately on what has already been done, whose complaint it is, by what method is being provided as an answer and how that specific complaint has led to enhancement of the process. This requires a connected mesh of information rather than a linear chain of departments.
AI will be one important component because it can process multiple parameters quickly, but other technologies will also need to work together. The intelligent enterprise will therefore be interconnected, customer-centric and capable of converting feedback directly into action.
