Agile Gurugram 2023 - Keynote I Business Agility to thrive in the AI centric World
- 1. Business Agility to thrive in the AI centric World Vivek Jain Chief Business Officer, Shiksha.com
- 2. Predictive Modelling Statistical technique to predict future behavior Data-mining technology that works by analyzing historical and current data and Generating a model to help predict future outcomes Classification Clustering Outlier Detection Forecast Model
- 3. Predictive Modelling Clustering Agglomerative Hierarchical Clustering Gaussian Mixture Models (GMM) K-Means Classification Logistic Regression Naive Bayes K-Nearest Neighbors (KNN) Decision Tree Support Vector Machines
- 4. Predictive Modelling Domain understanding helps in improve the collection of data and its pre- processing Understanding how to interpret the results and how to use the predictive models is an important pre-requisite to success Build cross functional teams Business Objective Data collection, preparation, and cleaning Feature Engineering Model selection, iterations
- 5. Online Marketing To understand whether a user will buy a product that is advertised online Evaluate the influences for customers (for example, prompts, product options of fellow buyers, and so on) Correlation between online behavior and sales, to allocate an accurate marketing budget Predictive Modelling Banking and Finance Cash flow forecasting – spends and payments to vendors Customer payment predictions – if and when customers will pay Credit risk management – credit scoring models have proven their value, predict if customers will return loans
- 6. Telecommunications Predict customer churn, and attracting new customers through targeted marketing campaigns Predict network congestion and optimizing resource allocation Reduce network maintenance costs, and prevent revenue leakage due to fraud Predictive Modelling Travel Predict demand and offer pricing to fulfil the demand – limited capacity that expires if not utlized Personalized offer to customers and predicting efficacy of marketing campaigns Customer
- 7. Recruitment Identify and rank potential candidates based on their qualifications, skills, and compatibility with the desired job requirements. Skill Gap Analysis: Predictive analytics enables recruiters to analyze the skills and competencies required for a particular job role and identify any gaps within their candidate pool. Candidate Assessment by analyzing past performance, behavioral patterns, and assessment results -> comprehensive understanding of a candidate's abilities, strengths, and weaknesses Retention and Success Prediction- focus on hiring candidates who align with their long- term goals.
- 9. AI has crossed thresholds of acceptability Machines can now decipher what we speak (speech recognition), see (computer vision) and what we understand (natural language understanding). They can also now respond like humans (generative AI).
- 10. Scientists have been attempting Speech Recognition for more than 50 years I am listening and interpreting correctly
- 11. A spoken word recognized from phonemes using acoustic models (pronunciation by different speakers), and understand words in each others context using language model I am listening and interpreting correctly
- 12. Meta AI released Seamless, an AI model that translates speech in real- time while also maintaining similar vocal style. The models deliver vocal style & prosody preservation capabilities as well as near real-time streaming translations (2 sec latency). Speech recognition has evolved from domain focused limited vocabulary (Automated IVR) to continuous speech very large vocabulary (Siri) Technology has evolved from Hidden Markov Models/ Stochastic Language Models to Deep Learning Models
- 13. Object Detection and Tracking
- 14. Object Detection and Tracking for Autonomous Driving Different lighting conditions Moving objects at different speeds Different road conditions, surroundings Cost of an error is very high Who is responsible in case of an accident, Driver or the car manufacturer
- 15. Natural Language Understanding and Generative AI First time I saw a demo of computer conversing in natural language was more than 30 years ago. Technology has evolved from a rule based, decision tree based conversation to a more evolved unstructured conversation. Chatgpt is a milestone in generative AI - open ended contextual conversations at scale of billions. Almost covering the entire gamut of domains, industries and usage contexts.
- 16. Natural Language Understanding (NLU) and Processing (NLP)
- 17. Generative AI – Race against Time to Build the Best GPT Core GPT models can serve multiple applications – Multiple choice questions, Math problems, Coding tasks, Reading Comprehension, Machine Translation and so on. Google released Gemini Ultra and Gemini Pro three days ago and it has beaten GPT-4 in many of the benchmarks.
- 18. Can we wish these innovations away? While the debate over Responsible AI continues…..
- 19. Technology disruptions are opportunities for businesses who are LEADERS and challenges for LAGGARDS
- 20. Customer Service Chatbot Customer service team operates 9am to 6.30pm, 5 days a week. A chatbot operates 24 * 7. Most request comprise of status requests and queries about product/service which required only standard data based responses. Are you working on a solution for loneliness?
- 23. Call Audits using Transcription Score counselors for their ability to handle queries basis the conversation Measure the rate of speech and if conversation indicates violation of expected norms Measure student readiness and understand emotion while talking Complete call audit for a batch of calls Benchmark with respect to manual audits Real time intervention?
- 24. Coding Copilot AI pair programmer that offers autocomplete-style suggestions as you code. Write a natural language comment describing what you want the code to do. Copilot analyzes the context in the file you are editing, as well as related files, and offers suggestions from within your text editor. Copilot is available in multiple languages Author Copilot Author copilot that offers autocomplete-style suggestions as you write. Write a natural language comment describing what you want the content to do. Copilot analyzes the content in the file you are editing, as well as related files, and offers suggestions from within your text editor.
- 25. Only Paranoid Survive! Agile Businesses LEAD…. Create teams to explore new technologies Cross functional teams for collaboration Independent mandate without limitations Must demonstrate successful pilots Tip My experience with new technological solutions – cost effective, + RoI and scalable.
- 26. For AI, Execution is the Strategy Shiksha Assistant Conversations are more engaging Help discover information which students may not be aware of Suggest and guide the journey of course selection Conviction that Assistant is the way forward -> initial delays and failures did not kill the project, instead became a stepping stone for success
- 27. AI Buddy. Will we find an AI Buddy who connects emotionally and cheers you when you are sad after a hard day’s work? Loneliness is a disease and it is decreasing at an exponential rate Are you working on a solution for loneliness? Your Innovative Idea here….. Creativity knows no boundaries..
- 28. 500 million pageviews, 55 lakh students Register on Shiksha Annually Stay the course If you have clarity of thought and resolve of your conviction, you can navigate the uncertain world, icebergs don’t matter
- 29. Shiksha.com next journey -> from a reviews site to counselling for students
- 30. Net Promoter Score +25 I got through Amity Noida, thank you Shiksha for a wonderful experience Sristy, BSc Thankful to Shiksha for providing info on colleges. Got admission in JIMS and it’s very nice. Priyanka Dutta, BCA Shiksha is a very good platform where a student can compare colleges basis genuine reviews for a better decision Haider Jamal, PGDM Journey is more meaningful and fulfilling because students are appreciating our effort