AI decisioning is a feature of marketing software in which an AI model chooses each customer's offer, message, channel and moment, instead of a journey that a marketer drew in advance. It is also called agentic marketing or next-best-action marketing. The model is trained on the company's own first-party data, meaning the data it collected itself, from its own app and website.
Why is it needed? In a drawn journey, every customer in a group receives the same message at the same time. But customers in one group are not alike. One opens push notifications at night. Another reads only email. A third would have bought anyway, so a discount sent to her is wasted money. A marketer cannot draw a separate journey for each of a million customers. A model can make a separate choice for each one.
Here is a typical case. A food delivery app wants more weekday dinner orders from 12 lakh users. The marketer sets the limits. There are four choices of offer: ₹75 off, free delivery, 20% off on biryani, or no offer at all. There are three channels: push, email and WhatsApp. No user may receive more than two marketing messages a week. The model then reads each user's last 90 days of orders and app visits. For every user, it estimates how likely an order is for each combination of offer, channel and time.
For one user, who often orders at about 8 pm but only with a discount, the model sends ₹75 off by push at 7:30 pm. For another user, who orders every Friday whatever happens, the model sends nothing, because a discount would only reduce the app's earnings. The next day, the orders are recorded. The model compares what it predicted with what each user did, and adjusts its estimates. A small share of choices are deliberately varied, so that the model keeps learning about offers it rarely picks.
To check the result, the company keeps 10% of users on the old drawn journey. After four weeks, it finds that the users chosen by the model ordered 6% more, while the app spent less on discounts.
The marketer still decides what the model may do. The marketer writes the offers, sets the budget, and sets rules such as quiet hours and the weekly limit. The model only chooses among them. This is different from journey orchestration, where the marketer decides the order of every step. It is also different from an AI customer service agent, which answers a customer's questions. AI decisioning chooses what to send before the customer has asked anything. It usually runs inside a customer engagement platform, and it may use profiles from a customer data platform.