It’s early hours at a farm in the Pali block of Gorakhpur district in Uttar Pradesh. Its 47-year-old owner-vendor, Kishenpal Singh*, sits on his regular perch as the harsh midday sun streams through the ears of the freshly harvested maize crop. He feels exhausted and has a recurrent cough, which leaves him short of breath. He feeds his symptoms into an AI app, which suggests he check in at the local hospital and get himself assessed for a viral or bacterial infection. The tool alerts him to a low immunity shield following the CAR T-cell therapy he had for his lymphoma 10 months ago.
“This is like an alert tool which has helped me battle the side effects of my treatment. I can’t do without it as it helps me not ignore any symptoms that are a result of my immune deficiencies post-treatment. I am cancer-free at the moment and the tool helps me keep healthy and fit. My oncologist can track me too,” he says.
A CAR T-cell therapy is a personalised immunotherapy that re-engineers a patient’s own T-cells to target and destroy cancer cells. The process involves collecting a patient’s T-cells or immune cells, modifying them in a lab by inserting a synthetic receptor so that they can recognise and kill specific cancer cells, and then infusing them back into the patient to attack the cancer. This is thought to be highly effective in treating lymphoma, a cancer which involves abnormal growth of white blood cells, which affected the farmer.
Story continues below this ad
In fact, it was after his lymphoma relapsed within 12 months that he consulted Dr Rahul Bhargava, principal director and chief, haematology, haemato-oncology and bone marrow transplant at Fortis, Gurgaon. Dr Bhargava has been using agentic AI as a follow-up tool for his patients in Tier 2 and Tier 3 cities. “This AI tool not only augments the doctor’s capability but bridges the gap between doctor and patient and helps decode and simplify some of the post-therapy instructions for the patient. It has datasets fed from all our patients in our chain hospitals and acts as our eyes when the patient returns to his hometown or city. Most importantly it is hallucination-proof and follows specific US guidelines for CAR T-cell and bone marrow transplant monitoring,” he says.
How it helps in patient communication
It is very difficult for a patient to understand side effects like cytokine release syndrome, which is a massive inflammation of the body as a consequence of T-cell based immunotherapy. “This syndrome usually develops within 3-14 days after the therapy is administered. It often begins with fever and flu-like symptoms but can worsen quickly and cause serious illness. Management includes monitoring and supportive care to control symptoms. Now the AI-guided application, which generates information in Hindi, can tell the patient what to expect on Day 7 if there is fever and discomfort. Fed with all symptoms and data, it can then tell the patient to go to the nearest hospital so that they can be treated with antibiotics at the local centre, for which it can suggest recommendations,” says Dr Bhargava.
What this means is that the patient can get an early discharge instead of having a longer hospital stay after their CAR T-cell therapy or even a bone marrow transplant. This helps them save on costs.
Agentic AI vs Generative AI: What’s next?
“Agentic AI, which is an artificial intelligence system that can help set goals, plan and take actions, has been offering distinct possibilities for how we diagnose, treat and manage patients,” says Dr Bhargava. While so far generative AI has helped produce detailed discharge summaries for bone marrow transplant patients, scan the latest CAR-T cell trial data for relevant insights, or draft instructions in patient-friendly formats, it is limited in haematology. “In cases where bone marrow transplant infections can progress rapidly or patients on CAR T-cell therapy can deteriorate within hours, agentic AI works better. It goes beyond generating content and can perceive, plan, act, and learn in a dynamic environment. Instead of passively answering questions, agentic AI can coordinate tasks across multiple systems: monitoring real-time lab values, tracking toxicity scores, scheduling transfusions, and escalating alerts to clinicians without waiting for a manual prompt,” says Dr Bhargava.
Story continues below this ad
Take CAR T-cell therapy. A generative AI model could summarize trial protocols or help write progress notes. An agentic AI system, however, could continuously integrate cytokine levels, neurotoxicity grading scales, and electronic health record data to detect early signs of infection or inflammation. “It could then trigger standardized intervention pathways: alerting ICU staff, flagging to the pharmacy for tocilizumab preparation, and notifying the treating physician simultaneously. In such scenarios, AI is not just an information tool but a proactive clinical partner,” says Dr Bhargava.
Similarly, in bone marrow transplantation, agentic AI could track donor cell transport, monitor grafts, predict infections and automatically generate antibiotic prophylaxis reminders. By analyzing thousands of transplant data points, it could also identify escalation before symptoms fully manifest.
Why it is a preventive model as well
As more data gets updated in real time and by continuous monitoring of the patient’s tumour profile and progress of therapy, the AI app can even help predict risk factors. “The idea is to prevent cases from falling through the cracks. The app is filling that gap. It helps both doctor and patient modify protocols depending on the precision of the data fed,” says Dr Bhargava.
Lymphoma can relapse because treatment may not eliminate every single cancer cell, leaving behind dormant or microscopic cells that eventually grow again. Genetic mutations can also make some cells resistant to treatment or cause the cancer to become more aggressive over time. Additionally, some types of lymphoma are inherently difficult to cure completely, leading to a recurrence of the disease even after a period of remission. “So AI can create a predictive model based on the patient’s journey,” says Dr Bhargava.
The problem of regulation
Story continues below this ad
Of course, with greater power comes greater responsibility. Agentic AI introduces new ethical and regulatory challenges: Who is accountable if an autonomous system fails to flag risks? How do we ensure transparency? “The shift from reactive to proactive AI requires robust governance frameworks, clear liability structures and rigorous clinical validation in hematology’s most fragile patient groups. Generative AI will continue to excel in knowledge synthesis and documentation, while agentic AI takes charge of orchestration, monitoring and rapid response,” says Dr Bhargava.
As for closing gaps, he explains how all AI applications in medication management today operate with a “human in the loop,” where a licensed clinician must review and sign off on any AI-generated recommendations. “The AI functions as a decision-support tool, not an autonomous prescriber,” says Dr Bhargava.
As for the farmer from Gorakhpur, he has been lymphoma-free for a year and never forgets to record his recovery process. It helps him proof his health from a relapse.