Why Most Companies Are Using AI Wrong | Nikhil Singh

Why Most AI Projects Fail Before They Even StartMost AI projects don't fail because the technology doesn't work.They fail because companies start with AI instead of the problem.In this episode, Ravi Harsh sits down with Nikhil to break down what actually makes AI projects succeed inside real businesses.Nikhil explains why companies need to define the business outcome first, how to identify the right AI use cases, and why AI should be treated as one layer of a larger system involving people, processes, software, and automation.They also discuss what happens when AI makes building easier but deciding what to build harder, how companies can calculate the real ROI of AI, and why humans still need to stay in control of important decisions.Timestamps :0:00 - The $5,000 AI vs $500 human cost problem1:43 - Intro: Nikhil — IIT grad, 20+ years in tech, runs an AI company in Sydney with his wife2:44 - Why 95% of generative AI projects fail (MIT report)5:24 - AI is a tool, not "the" solution6:14 - Career journey: dot-com era (2000) vs. the AI era today8:53 - Productivity leap: building multiple designs in parallel with AI10:02 - The "too many choices" problem (Netflix analogy)11:46 - Origin story: how "AI Too Easy" got its name13:58 - Early project: Decipher, the podcast automation SaaS16:16 - Finding the real problem: the AI discovery process17:41 - Case study: the proposal/quotation bottleneck19:28 - Why measurable outcomes matter (not just "enabling AI")20:58 - Deep dive: measuring ROI in time AND cost21:03 - Walkthrough: calculating token cost vs. staff cost24:18 - Addressing "AI is taking my job" fears26:41 - Why layoffs aren't always really about AI27:52 - Why AI Too Easy has 96% client retention28:25 - Case study: insurance company document automation32:11 - Data sovereignty & compliance (Australian data residency)34:47 - Why human-in-the-loop matters for high-stakes decisions34:55 - Managing AI hallucination and mistake risk38:40 - 3 things to do before starting an AI project41:16 - How Nikhil stays updated on AI (newsletters, X, automations)43:17 - Where AI is headed in the next 5 years45:39 - The real threats: overconfidence, blind trust, loss of critical thinking48:37 - Advice for young people entering tech in the AI era51:03 - Personal life: balancing work and home with his wife/co-founder52:38 - Closing question: "What are you still trying to figure out about AI?"54:56 - Final thoughts and sign-offIn this episode• Why most AI projects fail before development even begins• The biggest mistake companies make when adopting AI• How to identify a business problem worth solving with AI• Why AI is only one layer of the solution• How to define measurable outcomes before building• The AI Discovery Process Nikhil uses with businesses• How one company reduced a 30–40 minute process to just a few minutesConnect with NikhilLinkedIn: https://www.linkedin.com/in/nikhilsinghAI Too Easy: https://www.ai2easy.com.auConnect with MeLinkedIn: https://www.linkedin.com/in/raviharsh