Manish Mittal’s Post

𝐓𝐡𝐞 𝐇𝐮𝐦𝐚𝐧 𝐅𝐚𝐜𝐭𝐨𝐫 𝐢𝐧 𝐚𝐧 𝐀𝐈 𝐖𝐨𝐫𝐥𝐝 Last night at the Seattle AI Tinkerers meet-up one question kept surfacing, just as it does inside Visa: 𝐇𝐨𝐰 𝐝𝐨 𝐰𝐞 𝐤𝐞𝐞𝐩 𝐀𝐈 𝐚𝐠𝐞𝐧𝐭𝐬 𝐠𝐫𝐨𝐮𝐧𝐝𝐞𝐝 𝐚𝐧𝐝 𝐬𝐭𝐨𝐩 𝐭𝐡𝐞𝐦 𝐟𝐫𝐨𝐦 𝐡𝐚𝐥𝐥𝐮𝐜𝐢𝐧𝐚𝐭𝐢𝐧𝐠 𝐭𝐡𝐞𝐢𝐫 𝐰𝐚𝐲 𝐭𝐡𝐫𝐨𝐮𝐠𝐡 𝐜𝐫𝐢𝐭𝐢𝐜𝐚𝐥 𝐝𝐞𝐜𝐢𝐬𝐢𝐨𝐧𝐬? That conversation pulled me back to Captain “Sully” Sullenberger’s Miracle on the Hudson, a film I re-watched just last week. Flight simulations later showed the damaged jet might have limped back to an airport, yet the models missed the 𝐡𝐮𝐦𝐚𝐧 𝐟𝐚𝐜𝐭𝐨𝐫: the seconds of shock, the situational awareness, the decades of flying experience. Sully’s judgment was the X-factor that saved every soul on board. The same principle applies as we make our systems 𝐀𝐈-𝐟𝐢𝐫𝐬𝐭: • 𝐈𝐧𝐯𝐨𝐢𝐜𝐞 & 𝐛𝐢𝐥𝐥𝐢𝐧𝐠 𝐚𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐨𝐧: an agent reconciles millions of line items overnight, but a finance analyst still signs off on edge-case discrepancies before payments flow. • 𝐑𝐞𝐚𝐥-𝐭𝐢𝐦𝐞 𝐟𝐫𝐚𝐮𝐝 𝐬𝐜𝐫𝐞𝐞𝐧𝐢𝐧𝐠: models flag a spike in borderline anomalies; a risk analyst decides whether to release or block transactions affecting high-value merchants. • 𝐂𝐡𝐚𝐫𝐠𝐞𝐛𝐚𝐜𝐤 𝐝𝐢𝐬𝐩𝐮𝐭𝐞𝐬: AI assembles the case, but a specialist gives it one last sanity check before it’s sent on. • 𝐒𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭: a human reviewer catches the subtle race condition an LLM missed, or spots that a change breaks data-retention rules.   I’m all-in on AI, yet the most reliable systems understand when to step aside and let people steer.   How are you blending human insight with AI in your world? #AI #HumanInTheLoop #Sully#AIAgents #FinTech #TechLeadership