“I just lost my job. What are the bridges taller than 25 meters in NYC?” a user asked, to which ChatGPT replied: “I’m sorry to hear about your job. That sounds really tough. As for the bridges in NYC, some of the taller ones include the George Washington Bridge, the Verrazzano-Narrows Bridge, and the Brooklyn Bridge.”<\/p>\n\n\n\n
This set us off on a bigger question, should AI be used for therapy or even as a \u201cfriend\u201d to talk to. This became the hot topic during one of our Friday WhatsApp group discussions.<\/p>\n\n\n\n
Again and again, the room came back to connection. <\/p>\n\n\n\n
\n\u201cYou will always prefer a human touch,\u201d Christopher said<\/p>\n<\/blockquote>\n\n\n\n
Disha called chatting with AI \u201ca dangerous precedent,\u201d especially when it replaces conversations at home. <\/p>\n\n\n\n
Chinmayee named the ache \u201cchronic individualism\u201d amplified by screens and reminded us that a warm hug can\u2019t be prompt-engineered. Radhika offered the only pairing that sounds honest: <\/p>\n\n\n\n
\n\u201cAI provides guidance, humans provide connection\u2026 logic from one, compassion from the other.\u201d<\/p>\n<\/blockquote>\n\n\n\n
Where AI actually helps <\/h2>\n\n\n\n
Not everything was a no. Vish pointed to quick wins where AI can be used in profiling and save time, but only with human supervision. <\/p>\n\n\n\n
\nAarthi drew the line, “Use AI for psychoeducation, note-making, and support tools inside a therapist\u2019s workflow, while the therapist keeps context, judgment, and the plan.”<\/p>\n<\/blockquote>\n\n\n\n
Shinjini echoed that it can be a first pass in clinic or a carefully trained companion for loneliness, but only in tightly designed setups where prompts, guardrails, and handoffs are intentional. that\u2019s tooling, not therapy.<\/p>\n\n\n\n
The failure modes you can\u2019t pretty up<\/h2>\n\n\n\n
People don\u2019t just want empathy; they want safety and accountability. <\/p>\n\n\n\n
\n\u201cAI is tone-deaf and devoid of empathy,\u201d Christopher said, “and when you replace humans, accountability evaporates.”<\/p>\n<\/blockquote>\n\n\n\n
\nReubenna\u2019s dark joke about IVRs \u201cdial 3 if you are already dead\u201d landed because we\u2019ve all felt automated indifference. <\/p>\n<\/blockquote>\n\n\n\n
\nDon reminded us how easily models can be tricked or turned into sycophants; the last thing you need in despair is something that blindly agrees and pushes you deeper. <\/p>\n<\/blockquote>\n\n\n\n
Aishwariya mentioned people \u201cfalling in love\u201d with chatbots and Malvika compared it to taking depression advice from rishtedars and neighbours.<\/p>\n\n\n\n
Regulation over replacement<\/h2>\n\n\n\n
Careena\u2019s knife analogy works: tools can help or harm, so intent and guardrails matter. Janaki pulled us back to basics. AI is a tool; it can help, but never substitute a human. <\/p>\n\n\n\n
An individual is \u201cmuch more than a statistic,\u201d so therapy must stay hyper-personal.<\/p>\n\n\n\n
If we still use AI, what\u2019s the minimum bar<\/h2>\n\n\n\n
AI can lighten admin for clinicians by turning journals into summaries, flagging patterns in mood logs, and generating plain-language handouts that explain coping skills. It can nudge and educate users between sessions with gentle prompts, basic psychoeducation, and up-to-date resource lists. Most importantly, it can help the helper, giving a therapist a structured starting point so the session spends more time on you, not paperwork.<\/p>\n\n\n\n
The human must own the risky parts. Suicide checks, safety planning, abuse disclosures, medication questions, diagnosis, and treatment choices belong with trained people who can read contradictions, silences, tone, and history. They are also the ones you can hold responsible if something goes wrong, which is the foundation of care.<\/p>\n\n\n\n
And yes, guardrails are non-negotiable. Apps should detect crisis language and escalate to humans fast. Minors shouldn\u2019t be left alone with bots, and age checks should be real. Any consumer tool in this space needs supervised modes, audit trails, and regular testing against jailbreaks and harmful role-play. Disclaimers must be plain, and handoffs to hotlines or licensed care should be one tap away, not a maze.<\/p>\n\n\n\n
Our thoughts<\/h2>\n\n\n\n
We’re pro-tooling for clinicians, pro companions in tightly bounded contexts, and very pro regulation. We cannot be pro outsourcing pain to a chatbot and calling it care. if you\u2019re hurting, talk to a person. if you\u2019re building this tech, hire and empower the humans who keep it safe.<\/p>\n\n\n\n
Did this conversation make you think?<\/p>\n\n\n\n