How to use and concept of AI chatbot
getty
Every business leader understands how predictive artificial intelligence (AI) and machine learning (ML) can help grow their business, and how they can implement and deploy it with minimal risk and maximum benefit. It's time to. In my experience at large companies and startups, most current initiatives fail to deliver the expected value or at all.
As with any new technology, we recommend careful planning and a detailed implementation process, with controls in place to monitor results and change requirements. Specialized business practices are also emerging, such as bizML, which leading consultant and former Columbia University professor Eric Segal outlines in his new book, The AI Playbook.
This is a summary of Siegal's six critical steps for successful initial adoption of the use of this technology, including artificial intelligence, machine learning, and predictive analytics.
1. Quantify your positive business value proposition. First, document the desired business improvement, such as increased revenue through improved ad response rates. Avoid a technology-first or solutions-first mindset that focuses on technology rather than business outcomes. Use this value to obtain approval to proceed with deployment.
2. Establish predictive goals for machine learning. You need to establish in detail what the deployment predicts and what will be done for each prediction.This is the next intersection business and technologyTranslating business intent into a well-defined technology model requires collaboration between business leaders and technologists.
3. Define specific model evaluation metrics. What we're looking for here is an accuracy measure of how accurately the model predicts, or at least more accurately than it would without guessing or learning. Additional factors include the cost of correct predictions, the cost of false positives or negatives, and the increase in learnability over time.
Four. Prepare the data source for training. Remember, good data as fuel always trumps the best machine learning algorithm. Data must be collected and reorganized into elements relevant to training and deploying a model. To learn, it must include both positive and negative cases, as well as noise and supporting elements.
Five. Generate and train a predictive model. Here we develop the most powerful predictive technology, including a training element. Here, the computer itself is essentially reprogrammed. Evaluate available predictive analytics algorithms, including decision trees and regression analysis, learned from custom-built or purchased AI models.
6. Deploy and evaluate your machine learning model. Implementation means introducing changes to operations. This requires buy-in and collaboration from teams at all levels to turn predictions into action. We recommend setting up control groups to reduce risk, process metrics, and make necessary adjustments to your data and models.
In reality, these steps are just the beginning. Once a model proves valuable, maintaining its continuity requires maintenance, monitoring, and ethical vigilance. As the world around us changes, new technologies tend to lose their edge and become stagnant over time. The economy changes and customer behavior patterns evolve.
They are also sensitive to changes that could negatively impact the learning model for protected classes, show bias or lack of representation against certain groups, or reveal personal attributes that should not be disclosed. Must be.
Your role as a business leader and expert therefore becomes even more important to ensure that the results not only benefit your business, but also your customers and society at large. It's time for all of us to embrace new technology and learn more about how to effectively move forward, rather than attack blindly or ignore new business growth opportunities.
follow me twitter. check out my website.
