Parveen Sultana starts her day at 7 AM teaching primary school children in East Uluberia, a region in West Bengal. By 6 PM, when most teachers head home, her second shift begins. Sultana pulls out her smartphone and opens the BLO app—a government application designed to digitise India's voter rolls. She'll work until 1 AM, re-entering data that crashed hours earlier, trying to meet a target of 200 voter entries per day.
"There were thousands of problems with the BLO app," Sultana says, exhaustion evident in her voice. "But now that we are at the fag end of the enumeration process, I would rather not talk about them."
Her reluctance speaks volumes about what has become one of India's most troubled attempts at electoral digitisation. The BLO app, developed by Tata Consultancy Services for the Election Commission of India, was supposed to streamline the country's Special Intensive Revision of electoral rolls—a massive undertaking to update voter information across the nation. Instead, it has become the focal point of a labor crisis that families and colleagues say has contributed to at least five BLO deaths.
“Inhuman Workload”
By late November, at least five Booth Level Officers—the grassroots election workers tasked with using the app—had died in circumstances their families and colleagues linked to the crushing workload of the electoral revision. The deaths occurred across multiple states, including West Bengal, Rajasthan, and Kerala, sparking urgent questions about the true cost of India's digitization push.
In Nadia district, West Bengal, one BLO left a suicide note that referred to an "inhuman" workload and expressed fear of punishment for missing online upload targets. The note revealed a particularly cruel irony: the officer lacked computer skills and experienced the app not as a helpful tool but as an existential threat.
After Zakir Hossain, a teacher at a Bengal school died of cardiac arrest in Murshidabad, his family alleged that mounting work pressure and juggling SIR duties and regular teaching work strained him. Relatives also alleged that the school authorities refused to relive him despite repeated requests.
The victims were all Class C government employees—a classification that includes teachers, irrigation workers, anganwadi staff, and other civil servants who were pressed into election duty alongside their regular jobs. They are among the lowest-paid government workers, yet they were handed smartphones and told to digitise democracy.
“Try Again”
The technical problems began on day one and never really stopped.
Debaprasad Majumder, a mathematics teacher at Beleghata High School in Kolkata, describes the daily ritual: finish teaching, sit down to upload voter data at 6 PM, and immediately hit a wall. "The server was always down every evening during peak time," he explains. "As there was a huge rush of people trying to save information on the app during the same given time, the app was invariably crashing."
The result was a grinding schedule that stretched from 6 PM until 1 AM, followed by a return to the classroom the next morning.
Swapan Mondal, a teacher and BLO in West Bengal, breaks down the mathematics of frustration: each voter entry required an average of 20 minutes to upload, assuming the app didn't crash. With a target of 200 entries per day, BLOs needed to dedicate more than 66 hours of work—an impossibility when they already held full-time jobs. "The amount of data per electorate was very heavy and often the app would crash in between work without the information being autosaved," Mondal says.
That lack of autosave, a feature considered standard in consumer apps for over a decade, meant that a crash could wipe out hours of work, forcing BLOs to re-enter everything from scratch.
Media reports described BLOs across states waking at 4 AM to re-upload crashed entries, working through weekends, and losing days of accumulated data when servers went down for hours at a time.
The Bengal Chief Electoral Officer eventually acknowledged the strain, calling an emergency meeting in late November to address "technical disturbances" in the app and its servers.
Updates Were Punishment
If the crashes weren't enough, the app underwent a relentless series of updates, sometimes daily. That added new requirements mid-process. The version BLOs are currently using is 8.88 for Android and 2.36 version for iOS, but getting there meant navigating a minefield of sudden changes.
"Why was the updated version of the BLO app not provided to us in the starting weeks?" Mondal asks. "We are already at the end of the enumeration process and there have been two updates in just three days, including the weekend."
On December 6, a new feature for verifying voters aged 85 and above appeared. On December 8, a duplicate elector verification module was added. Each update sounded reasonable in principle—preventing voter fraud, facilitating home voting for the elderly—but in practice, they meant BLOs had to revisit households they'd already processed.
"These updates are bogus," Sultana says. "We have to repeat the same process of visiting voters who are aged 85-plus again to enter their names on the app, as the new update requires physical verification. Why were these not thought about in the first version?"
The pattern suggested a fundamental problem: critical functionality was being added on the fly rather than designed into the system from the start. National coverage noted that the Election Commission introduced a series of "significant changes" to the app only after appointing a special observer to Bengal
AI Made Things Worse
The duplicate verification feature, presented publicly as a fraud-prevention measure, revealed another problem lurking in the system: artificial intelligence that offered unhelpful suggestions.
Majumder encountered this firsthand when entering data for a voter from his booth. The app flagged that the person had already been mapped to a village in Khejuri, Midnapore—over 100 kilometers away. After investigating, he discovered that another BLO had accidentally assigned his voter to the wrong location due to incorrect suggestion thrown by AI. "I had to call him up and then unmap him from the district," Majumder recalls. "My booth's voter did not care."
The problem wasn't just human error.
"Due to AI throwing in wrong suggestions, oftentimes BLOs are mapping incorrectly. When I upload a form and scan it, AI automatically tries to link the 2025 enumeration form with random people from the 2002 list who have the same names and surnames as parents of the ones listed in the 2025 form," Majumder explains.
The system would propose matches based on similar names, but in a country where many people share names and surnames, these suggestions often led BLOs astray. A husband and wife with the same surname, for instance, might have different EPIC (Electoral Photo Identity Card) numbers, but the AI would conflate them.
Divij Joshi told Decode that errors in the app could arise from multiple sources. "There are many issues that arise in the live and scaled usage of these systems which lab testing does not indicate. There is generally a much deeper need to test and consider the effects of these experimental systems before rolling them out widely, especially when they concern such important rights as the right to vote." Joshi is a lawyer and researches the political economy and governance of emerging technologies at ODI, a global think tank.
Joshi stressed on how it is important to push for more transparency on such apps' functioning.
Bengal media documented widespread complaints of mismatches between existing electoral rolls and what appeared in the app, with BLOs flagging wrong booth assignments and unfamiliar names appearing in their lists during training sessions.
The irony was sharp: an app built to reduce errors was generating new ones through automated suggestions that workers under extreme time pressure were prone to accept.
Then there were the photographs. In some cases, uploading a single photo consumed 20 minutes, assuming it worked at all.
The app imposed strict requirements: ears at the same level, recent photos only, white backgrounds mandatory. An update introduced midway through the process forced BLOs to capture live images through the app rather than uploading existing photos. "That was worse," Mondal says.
The frustration of having an entry rejected at the final step, after filling all details correctly, became a defining experience. In a later version, the app required photo approval before any other data could be entered, forcing workers to clear the highest hurdle first.
The Urban-Rural Divide
In Kolkata and other cities, BLOs faced their own challenges: high-rise residents who didn't know their neighbors, couples in bitter divorces, families locked in property disputes. But in rural Bengal, the problems ran deeper.
"There's no electricity in my village at most times of the day," says a BLO from Birbhum district who requested anonymity. "I am worried about how I will charge my phone and upload data on the application. Field work is far easier in villages, as families know each other. We are getting stuck in the data entry part."
District offices tried to compensate by setting up WiFi zones and charging points, but these ad-hoc solutions barely addressed the underlying infrastructure gaps. Majumder describes rural colleagues who struggled with basic smartphone literacy: "Getting an OTP every time you make an entry can be very traumatizing for someone who does not even order food from a smartphone or uses the device for sharing funny videos."
The app's requirement for constant connectivity and digital fluency exposed an uncomfortable truth: India's push toward digital governance was racing ahead of the population's actual technological capacity—including many of the government workers tasked with implementing it.
Managed by Spreadsheet and WhatsApp
Training for the app consisted of a few hours in sessions covering three or more constituencies, after which BLOs were expected to manage largely on their own through WhatsApp groups and video conferences.
The WhatsApp groups became command centers where Election Commission representatives posted as sole admins. BLOs were expected to provide hourly updates on entries completed, receiving Excel spreadsheets in return that tracked their progress—and highlighted their failures in red.
The sheet came in the WhatsApp groups everyday in an interval of two hours. It had columns listing ‘pending digitisation, pending targets and digitisation improvement every two hours.’ A copy of one such spreadsheet is with Decode.
"At the beginning, just after Diwali, we started getting Excel sheets almost every hour," Majumder recalls. "The sheets had parts highlighted in red which showed incomplete entries and errors. We were supposed to send submission reports on an hourly basis. It was like a target of 200 entries per day."
The red highlights created waves of panic. BLOs who met with Bengal's Chief Electoral Officer in November complained not just about the app's technical failures but about this surveillance system, saying they feared losing their jobs over failures rooted in software and server issues beyond their control.
Then came the deaths. "But after the unfortunate suicide reports, the sheets stopped coming," Majumder says.
Photo of spreadsheets sent to BLOs on an hourly basis to track targets
A Feature That Worked, and Mattered
Not every update made things worse. On November 20, the app gained an edit function that allowed BLOs to review and correct entries they had made. Previously, data was locked once submitted, meaning errors became permanent.
The feature had an unexpected benefit: it reduced political pressure. BLOs had reported incidents where Booth Level Agents (BLA) from ruling parties attempted to force incorrect entries. Without an edit option, BLOs faced potential legal consequences for data they'd been pressured to enter.
According to Mondal, there were instances where Booth Level Agents snatched enumeration forms and tried to fill them in rural and semi towns of Bengal. "There were cases of gross violation. OTPs were forcefully taken from BLOs and forms were being uploaded in case of shifted or untraceable voters by BLAs themselves. BLOs were not able to complain much against them. Now they are relieved that no such forceful entries can be made," Mondal says.
The Election Commission formally announced the edit feature after widespread protests over its initial removal, essentially admitting that a basic function had been left out of the original design.
The BLO app crisis reveals what happens when digital transformation is imposed from above without adequate infrastructure, training, or consideration for the people who must use the technology. But perhaps most troubling is what the crisis reveals about labor conditions in India's vast government workforce. BLOs are not technology professionals; they're teachers, irrigation workers, anganwadi staff. They were handed a complex digital tool, given minimal training, assigned impossible targets, and monitored through automated spreadsheets that highlighted their failures in red.
When the system failed, as it did repeatedly, the blame fell on them.
As of December 11, the enumeration phase of the Special Intensive Revision is drawing to a close in Bengal. BLOs like Majumder and Mondal are relieved that their daily interaction with the app is ending, for now.
Mondal is part of an organisation called Votekormi And BLO Aikya Mancha (Election staff and Booth Level Officer United Forum) that submitted a formal request to West Bengal's Chief Election Commissioner for an extension of the enumeration deadline.
The voter data they've struggled to enter will form the foundation of India's electoral rolls.
For workers like Sultana, who spent a month juggling teaching and data entry until 1 AM, the relief of the deadline is tempered by uncertainty about what comes next. The electoral revision cycle will come around again. The app will still be there, and it may continue to be glitchy.
This story has been edited by Adrija Bose.
