Agentic AI and the Dance of Democracy

Traditional “digital war rooms” of political parties—essentially campaign command centres that track voter sentiment, produce content, and coordinate rapid messaging—have existed for over a decade. Artificial Intelligence has now given them a boost, even as Agentic AI raises the stakes while challenging the way democracy functions.

In this week’s edition of Mint Tech Talk:

• AI outperforms human doctors in the emergency room
• AI Tool of the Week: Google Gemini's File Generation
• Freshworks cuts jobs, Cognizant shareholder payouts

While it is difficult to quantify AI’s impact on the recent assembly elections in West Bengal, Tamil Nadu, Kerala, Assam and Puducherry, its presence was unmistakable. From hologram speeches and AI-generated campaign videos to surveillance systems and even a humanoid robot at a polling booth, AI tools were widely deployed to scale outreach and personalise messaging.

The Tamilaga Vettri Kazhagam (TVK), led by actor-turned-politician C. Joseph Vijay, exemplified this shift. The party used holograms, robots and AI-generated content to expand its reach across Tamil Nadu, targeting first-time voters through virtual rallies and hyperlocal messaging. Tactics included AI phone calls—where a politician’s voice is synthetically recreated to interact with voters in real time—personalised WhatsApp videos, and real-time speech translation. Last September, an AI-generated video of former Tamil Nadu Chief Minister and DMK founder C.N. Annadurai went viral. It recreated his voice to endorse Vijay while criticising the current DMK leadership.

Data-Driven Campaigning

For years, political strategists have used behavioural models, chatbots and multi-lingual tools to refine voter outreach. AI now amplifies these capabilities, enabling campaigns to tailor messaging based on detailed demographic, psychographic and behavioural data, and to adapt communication to local dialects and contexts at scale.

India has been experimenting with such tools for some time now. In 2021, the State Election Commission, Bihar, partnered with Staqu Technologies Pvt. Ltd. to deploy video analytics and optical character recognition to monitor CCTV footage during panchayat elections.

In the 2024 Lok Sabha elections, Gen AI was used to improve voter communication. Tools such as Bhashini enabled Prime Minister Narendra Modi to deliver speeches in multiple Indian languages, expanding his reach across non-Hindi-speaking regions. AI-powered robocalls delivered hyper-personalised messages, while chatbots and voice assistants enhanced engagement.

In Rajasthan, BJP volunteers received personalised WhatsApp videos in which a party leader addressed them by name using voice cloning and lip-syncing technologies.

In Kerala, Congress leader Shashi Tharoor even appeared alongside his AI avatar at a literature festival, offering a novel form of voter interaction.

The Risks are Evident

To be sure, the misuse of data and targeted political messaging is not new.

Cambridge Analytica’s exploitation of Facebook data during the 2016 US Elections remains a cautionary tale. AI intensifies these concerns. The 2024 global election cycle marked a turning point, effectively becoming a testing ground for AI-driven campaigning for many countries that had to deal with deepfakes, AI avatars, and synthetic endorsements from fabricated celebrities.

In India, for instance, political parties reportedly spent ~$50 million on AI-generated content, including deepfakes of deceased leaders and manipulated videos of celebs, misleading voters at scale.

In 2024, doctored videos showed actors Ranveer Singh and Aamir Khan criticising PM Modi while endorsing the Congress. Another AI video falsely depicted Rahul Gandhi being sworn in as prime minister while voting was still underway.

Agentic AI Raises the Stakes

Such systems don’t just generate content, but plan, act, adapt and optimise campaigns in near real time. Imagine a system that tracks constituency-level sentiment, identifies undecided voter clusters, generates tailored video or audio messages, deploys them across channels, and reallocates resources based on response rates—all continuously.

Pieces of this already exist in silos—data pipelines, generative models, automation frameworks, and distribution channels—but what’s missing is full integration and autonomy.

Simply put, instead of today’s dashboards, content teams and ad managers, Agentic AI can tie everything together with minimal human intervention. What used to take days of coordination could happen in hours. Over time, such systems could evolve into always-on “campaign copilots”, continuously nudging strategy based on shifting voter mood.

Because Agentic AI systems can operate autonomously across channels, they can coordinate and influence campaigns far more effectively than human teams. Think AI agents posing as real users—sustaining narratives, responding to counter arguments, and keeping misinformation alive longer than traditional troll farms could.

A January 2026 paper titled How malicious AI swarms can threaten democracy warns that AI is making population-scale manipulation increasingly feasible. LLMs and Agentic AI can push political campaigns with unprecedented scale and precision, dramatically expanding the volume of persuasive content without sacrificing credibility. Generative AI can produce falsehoods that often appear more human than human-written text, while techniques designed to improve reasoning—such as chain-of-thought prompting—can be repurposed to make misinformation more convincing.

The paper highlights a more disruptive risk: coordinated “swarms” of malicious AI agents. By combining LLM reasoning with multi-agent architectures, these systems can act autonomously, infiltrate online communities, and manufacture the illusion of consensus. By mimicking human social behaviour and adapting in real time, such AI swarms could become a potent threat to democratic processes.

That said, even as agentic systems become smarter, they will require rich, high-quality data to be effective. Second, there are platform controls, and social media firms are getting better at detecting coordinated inauthentic behaviour. Third, there are legal guardrails, and authorities like the Election Commission of India are already pushing for disclosure and rapid takedowns.

The ECI maintains that all stakeholders must ensure the responsible and ethical use of digital platforms, in line with the Information Technology Act, 2000, the IT Rules, 2021, and the Model Code of Conduct. Political parties and candidates are also required to clearly label any AI-generated or modified material and disclose its source to maintain transparency and voter trust. During the recently concluded state elections, the ECI took action against over 11,000 social media posts during the campaign—ordering takedowns, issuing clarifications, and filing FIRs—with a significant share reportedly linked to Tamil Nadu and West Bengal.

Finally, if every party deploys similar advanced tools, their effects may partly cancel out. At least, that’s what one can hope for as guardrails try to keep pace with AI advancements.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlock today is based on a capability: Gemini File Generation.

What problem does it solve? Here is a painpoint most professionals quietly endure: You have a great session with AI, you’ve structured a report, drafted a proposal, built a budget breakdown and then spend 30 minutes doing the one thing AI was supposed to eliminate—copy, paste, reformat, fix spacing, redo headers. The AI gave you the content. The file was still your problem.

This is not a niche complaint. Any professional who works on documents faces this last-mile friction every week. Google Gemini’s new file generation closes that gap entirely. With a single prompt, Gemini delivers a formatted, downloadable file directly in chat, ready to send or save to drive.

How to access: https://gemini.google.com/app

Gemini File Generation can help you:

• Stay in flow: Turn any AI-drafted content into a shareable, formatted file without leaving the chat.
• Go client-ready instantly: Produce docs across formats in one prompt.
• Kill the loop: Eliminate the copy-paste-reformat cycle that kills momentum after every AI session.

Example: A strategy consultant needs to submit a structured risk assessment to a client before an evening deadline. She prompts Gemini to draft it—five risk areas, likelihood ratings, mitigation recommendations. Here’s how it plays out:

• Prompt the content: Draft the full assessment—sections, ratings, recommendations—all within one conversation.
• Request the format: Add “Export this as a .docx” and Gemini generates a ready-to-download Word file instantly.
• Specify structure: Instruct it to “use a table for risk ratings” or “add an executive summary at the top”—it applies layout before generating the file.
• Add branding cues: Ask it to “use bold section headers, add a title page, keep the tone formal”—the file arrives client-ready without manual clean-up.
• Switch formats on the fly: If the client wants a PDF instead, simply re-prompt. No redoing any work.

What makes Gemini File Generation special?

• Format breadth that matches real workflows: Word for approvals, Excel for data, LaTeX for technical and academic documents.
• The prompt IS the workflow: The conversation and the deliverable happen in one place—no app switching, no reformatting

Availability: All Gemini users for free.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS & BYTES

AI outperforms human doctors

A new Harvard study has found that advanced AI models can outperform doctors in diagnosing patients, including in high-pressure emergency room scenarios.

The researchers pitted OpenAI Inc.’s advanced o1 language model directly against hundreds of physicians across multiple diagnostic touchpoints, finding that the AI consistently outperformed doctors in both diagnosing conditions and planning clinical management. The new research, published in the journal Science, gave 76 clinical cases from the emergency room of Beth Israel Deaconess Medical Center to OpenAI’s o1 model and two expert attending physicians.

On the flip side, Pennsylvania has sued an AI company, saying its chatbots illegally hold themselves out as licensed doctors.

39%

According to data from Podcast Index, over a nine-day period, more than a third of new podcasts were created by AI. Many of these podcasts are hosted on free platforms that let anyone easily join their ad programs. There’s no real barrier, so even AI-generated shows can start earning money.

IBM CEO calls for an AI business model

IBM CEO Arvind Krishna believes maximising returns on AI investment requires a fundamental restructuring of business workflows, rather than just adopting new technology. The use of AI within a company typically evolves from the individual contributor to small teams, cross-functional groups and ultimately the entire firm.

As a company progresses from one stage to another, so do the potential returns on AI investment. That’s less about technology alone than it is about updating age-old processes and social dynamics, according to Krishna. Read more.

A digital wardrobe, courtesy Google Photos

Google is bringing AI to the Google Photos app, allowing users to try out clothes right from their gallery app. In a blogpost announcing the new feature, Google said that the ‘Wardrobe’ feature will “soon catalog the clothes you’re wearing in photos and create a digital closet that puts your wardrobe at your fingertips”.

The company noted that its new AI-powered feature will automatically catalog the clothes you wear in your pictures to build a fully searchable digital closet. Read more.

🔗 ALSO READS

• India is the biggest market for ChatGPT Images 2.0
• AI, ambition and the Gen Z money revolution
• American labs say Chinese AI tigers are copycats
• Freshworks cuts 500 jobs as AI writes half its code
• Cognizant cuts shareholder payout to fund AI deals

Tech Talk is a weekly newsletter by Leslie D'Monte on everything happening in the world of technology and AI. Want this newsletter delivered to your inbox? Subscribe here.