In India, public procurement has lengthy confronted challenges akin to corruption, lack of transparency, vendor collusion, worth manipulation, and bureaucratic inefficiencies. Guaranteeing honest and environment friendly procurement throughout central, state, and native ranges is complicated. On this context, the Authorities e-Market (GeM) mannequin turns into particularly necessary.
Pankaj Dikshit, CTO at Authorities e Market (GeM), informed AIM on the sidelines of DES 2025 that GeM started utilizing AI for fraud and collusion detection properly earlier than the arrival of Generative AI.
The Typical Manner
“On the time (earlier than GenAI), we relied on typical AI, primarily machine studying and deep studying. Our strategy started with analysing knowledge from our inner storehouses, specializing in the data, metrics, and metadata surrounding purchaser and vendor behaviour,” he stated.
This included metrics like location (inferred by way of IP addresses), system utilization and bidding patterns. Additionally they checked if customers have been utilizing the identical machines, software program or language, and whether or not their bids featured similar product specification and pricing.
“These metrics are fed into our ML and deep studying fashions to establish potential pink flags. The insights generated from these typical AI fashions assist floor patterns that will point out non-transparent transactions,” he stated.
The Authorities e-Market (GeM) portal was launched in 2016 for on-line purchases of products and companies by all central authorities ministries and departments.
He recalled how, reasonably than taking direct motion, GeM used this data to assist and information patrons. “If a purchase order seems suspicious or not completely clear, we flag it for evaluate,” Dikshit stated.
This strategy, he famous, is supposed to be collaborative and academic. “Typically, discrepancies could also be inadvertent, and as soon as made conscious, patrons grow to be extra cautious. Over time, this led to improved decision-making and a gradual discount in such incidents.”
Reduce to the Current
GeM, the federal government’s digital procurement platform, has revived how public funds are spent and is now sprinting towards changing into the most important authorities procurement system on the planet, second solely to South Korea.
Dikshit stated GeM has democratised that total expertise, including that “at present, a vendor in Assam can bid for a contract in Tamil Nadu or Delhi, supplied they’ve an web connection and the related credentials to entry the web site.”
He remarked that inclusivity isn’t only a completely happy accident, noting that about 40% of GeM’s enterprise is pushed by MSMEs and SMEs—entities usually ignored or pushed out by bigger gamers within the tender ecosystem.
“It’s one thing the Prime Minister himself has emphasised—the necessity to empower small and medium enterprises,” Dikshit added. “Within the monetary 12 months 2024–2025 alone, GeM facilitated transactions value round $65 billion, a large leap from the $48 billion the 12 months earlier than and $24 billion the 12 months earlier than that,” he shared.
Referring to educational research, he stated that each transaction by way of the GeM portal saves the federal government about 8–10%. “That interprets to roughly $15 billion saved to this point. All of that’s taxpayers’ cash—our cash—being preserved and redirected for different nationwide wants,” stated Dikshit.
The AI sport
Behind this scale and financial savings lies a strong technological spine, one that’s now being revamped with an AI-first imaginative and prescient. “Whereas the 1.0 model continues, we’re constructing a brand new platform with AI at its core,” he revealed.
GeM has already carried out an AI chatbot to enhance its price effectivity and compatibility. This multilingual assistant handles primary queries, converts voice to textual content, and permits customers to create assist tickets or test their statuses.
At present, the one post-login options are ticket creation and standing test. With the post-login function enabled, he stated, “We might additionally like to supply extra data akin to my orders, incidents, and the place my model or OEM device is pending.”
“The subsequent step is to scale back incoming human-handled calls,” he added, noting the workforce has accomplished some POCs on conversational digital interfaces and graph technology. “We’re additionally utilizing generative AI for code evaluation and technology,” he revealed, explaining that Generative AI helps us perceive code performance.
He famous that this considerably advantages the event groups and accelerates the time-to-market. Moreover, Dikshit talked about that it’s being utilised for safe and progressive code technology.
The workforce has additionally established a Retrieval-Augmented Technology (RAG) system and a data base, directing the LLM to reply solely to permitted content material.
“We’ve added enter and output guardrails. Enter guardrails forestall non-contextual or mischievous queries, and Output guardrails make sure the generated solutions keep inside acceptable and correct bounds,” Dikshit famous.
The put up Inside GeM’s Drive for Clear, AI-First Public Procurement appeared first on Analytics India Journal.
