Automated machine learning (AutoML) is software that builds a predictive model from a table of data, without a data scientist choosing and tuning the model by hand. A predictive model is a program that learns from past records to estimate something not yet known, such as which customer will stop buying or which loan will not be repaid. AutoML products are now being rebuilt around language models and AI agents, so that business users can ask for a model in plain English.
A data scientist normally builds such a model in many steps. They clean the data, fill in missing values and turn columns such as dates into numbers the model can use. Then they try several kinds of model, adjust each one's settings and test each one on records it has not seen. AutoML runs all these steps itself. It tries dozens of models and settings, measures each on held-back data and ranks them. It also shows which columns mattered most, so that people can see why the model predicts what it does.
Consider a chain of 120 grocery shops in Indore, Bhopal and Jabalpur with a loyalty programme. Its marketing team wants to know which members will stop shopping in the next three months, so that it can send them offers first. An employee in the marketing team exports a table of 2 lakh members from the company's data warehouse. The table holds each member's visits, spending and last purchase date. The AutoML product tries 40 models in two hours. The best one correctly picks out about 7 of every 10 members who later stop shopping. It also reports that the number of days since the last visit matters most. The team sends a ₹100 voucher to the 15,000 members most likely to leave.
Newer AutoML products add an AI agent. The user types a request, such as 'which shops will run out of cooking oil before Diwali?'. The agent writes the code to prepare the data and build the model, runs it, reads the result and tries again if it fails. The page on AI coding agents explains this loop. The finished model is then served through a model deployment platform.
How is AutoML different from time-series forecasting? Forecasting is one kind of prediction: of numbers over time, such as daily sales. AutoML is a general tool that can build forecasting models and many other kinds. And unlike a large language model (LLM), an AutoML model usually reads a table of numbers, not text.