I was recently invited to a fireside chat to talk about AI (no prizes for that these days). I was luckily the last person to speak in the conference and figured that the day has likely received its share of buzzwords with agentic being at the top of the curve. So, I decided to keep it simple.
The video is 20 mins long and while I have summarized some of it below - it would be great if you can hear the video and provide comments to start a conversation. It would be great to hear how you are using AI, planning to use AI or where you have been successful or not so successful with it.
Some key points discussed in the video:
Marketing Efficiency: By using ready models and third-party tools, I discuss how we have achieved a substantial 30% reduction in marketing cost of acquisition in tested segments. AI also helps us create more creatives faster and generate varied content (text, image, video)
Transforming Processes: In our early career business, an AI-powered resume builder for first-time job seekers from tier 2/3/4 cities we aim to dramatically increased the completion rate from 21% to a higher % by making the process less intimidating. The video covers how we are doing it.
Leveraging Data: AI is used to gain insights from first-party data, such as analysing sales call recordings for advisor training, compliance, and product strategy. We also use signals from our CRM (Salesforce) with partners like Meta to help AI target specific, higher-quality leads, despite potentially higher costs.
Internal Operations: We see significant potential for "agentic" AI to displace legacy automation and streamline internal processes like vendor payments, HR claims, and employee grievances, tackling "thorny problems" that consume significant time.
Customer Experience: Using generative techniques, we are moving from providing static content to creating a dialogue with potential customers at the consideration stage, helping them explore value based on their specific goals. There's also a focus on GenAI UX rather than just chatbots. An interesting plug here is this post:
The $7 Billion Question: Why AI Agents Could Crash the Online Learning Market
In my previous post, I explored how AI might handle endurance tasks and questioned whether AI could genuinely replicate human learning experiences.
AI in Education
Here is how we are exploring AI in the specific use cases of education:
One-on-One Tutoring: Inspired by Bloom's two sigma problem, we are exploring AI to provide scalable one-on-one mentoring, tutoring, and doubt-solving, aiming to replicate the significant learning gains seen in personalised teaching, which was previously too costly or difficult to scale.
Content Creation: AI is extensively used to generate various types of content for educational purposes.
Supporting Early Careers: The AI-powered resume builder is a direct application to help individuals successfully take the first step in their professional journey.
Navigating the AI Hype Cycle
In my talk I have highlighted the practical realities of implementing AI. Key challenges include security and governance, especially as employees independently adopt new AI tools. Balancing the desire for innovation with necessary controls is crucial.
We ensure a human is kept in the loop for production-ready AI outputs to mitigate issues like hallucination. Using secure solutions like NotebookLM by Google which doesn't train on private data and we label initiatives by their short-term versus long-term value to ensure the organisation sees tangible benefits and investments are justified.
Can an AI Agent Run a Marathon?
The allure of AI doing everything is captivating. It writes poems, predicts the weather, codes applications, detects cancer, and even creates art. But can it run a marathon? The question might seem absurd – of course, AI can't physically run. But let's dive deeper into what it means to "run a marathon" in the context of learning and achievement.
Share this post