Agentic, Generative And Predictive: How The AI Orchestra Works In Harmony
[Note: This article was originally published on Forbes on March 10th, 2025.]
Over the last few years, you’ve likely been hearing the term "generative AI" constantly, bringing pure AI back into the fore as well—even though AI has been a part of retailers’ businesses for over two decades.
A new AI term has moved onto the scene: agentic AI. With different names surrounding AI and solidifying themselves in the retail vernacular (without clarity as to how they differ and work together), I felt compelled to identify what each one actually does with regard to using data to maximize efficiencies and provide better service to customers.
To start, think first in terms of the key types of abilities (or capabilities) that are needed to help solve business problems:
- Optimization (Optimizer): This uses advanced analytics and algorithms to find the best possible solution among a very large number of potential options. This is the oldest of the capabilities, having been in use for decades.
- Prediction (Predictive): This uses machine learning algorithms to examine historical data, patterns and trends to forecast future outcomes. Predictive AI surged in adoption over the last two decades with the additional availability of data, computing power and breakthroughs in techniques.
- Generation (Generative): Generative AI uses machine learning algorithms to "create" new content based on patterns it learned from analyzing large amounts of data. The most common type, large language models (LLMs), can understand text input and create relevant responses by selecting words that best fit the context of the conversation.
Traditionally, solving a problem with AI involves identifying a primary need or challenge, documenting a detailed sequence of steps, decision points, and activities, and then writing software code to automate that workflow.
However, recent advancements in generative AI and LLMs have introduced reasoning capabilities, creating new opportunities to rethink how software solutions can address larger, more complex problems. This emerging approach is known as agentic AI.
Agentic AI makes use of this new reasoning ability of generative AI to allow a solution that, instead of being programmed for a specific task, can evaluate, plan and execute tasks with the tools they have at their disposal to achieve their goal. The value of agentic AI is best utilized with larger, more complex tasks that require a sequence of different smaller tasks and the possibility to execute those tasks in various orders and scenarios.
Even though the number of terms continues to increase, they are not interchangeable -- in fact, they should be used together, like an orchestra, to create more complex solutions that work in perfect harmony. Below are a few examples of how this can take place today.
Shopping
The modern digital shopping experience begins when customers interact with retail platforms through natural conversation in search bars, expressing their needs -- from complete outfits to weekly groceries.
Generative AI can process these conversations, understanding explicit requests as well as implicit preferences, dietary restrictions, brand affinities and budget constraints. Predictive AI can analyze historical purchases and similar customer segments to recommend products, substitutes and complementary items.
For budget constraints, optimizers can balance preferred products within their constraints. While predictive AI focuses on specific outcomes like "propensity to buy," optimizers rearrange options aligned to constraints.
Agentic AI can enhance this further by improving household inventory management and replenishment through autonomously tracking usage patterns and market conditions for authorized consumables like laundry detergent, coffee pods and pet food.
By analyzing real-time factors, such as current prices across retailers, upcoming promotions, delivery windows and consumption rates, it determines the optimal time to reorder. The AI can automatically place orders at the best possible moments, ensuring supplies never run out while maximizing savings. For example, it may purchase in bulk during price drops or expedite orders when it anticipates higher usage due to seasonal trends or upcoming events.
Warehouse Organization
In a modern warehouse operation, the warehouse manager can communicate specific needs to an AI system in natural language, outlining goals for storage efficiency, facility constraints and specific requirements like temperature-controlled zones. Generative AI can then translate these conversational inputs into structured parameters that other systems can process.
In this scenario, predictive AI leverages historical data to forecast inventory patterns and demand fluctuations, accounting for seasonal trends, supply chain variables and external factors like promotions, stock velocities and regional weather conditions that could impact inventory flow.
The optimization engine can then determine the most efficient warehouse configuration by calculating optimal product placement, designing streamlined picking routes and defining storage zones that maximize space utilization while maintaining operational efficiency. It balances multiple competing priorities, such as keeping high-demand items accessible, ensuring proper weight distribution across storage levels and preserving logical product groupings.
Agentic AI acts as an autonomous warehouse manager, continuously monitoring operations and adapting strategies in real time. When it detects inventory imbalances or efficiency bottlenecks, it proactively generates reorganization plans with ROI projections and step-by-step implementation strategies, enabling staff to act swiftly and effectively.
Conclusion
As you start to evaluate the opportunities of AI in your organization, understanding the different capabilities and their applicability in your reimagined scenario can help accelerate implementation and satisfy expectations. Too often, teams use the “wrong” type of AI for a particular aspect of the solution.
In my next article, we’ll further explore key strategies focused on how to make each of those examples come to life. In the meantime, let’s briefly look at important questions you can ask to effectively assess the ways different types of AI can best contribute to your solution:
- Do you need AI to optimize a process or system?
- Is there a need to predict future trends or outcomes?
- Does your solution require translating between natural language and structured data—or generating new content?
- Is the process based on a fixed set of rules, or does it need to adapt dynamically and make autonomous decisions?
A clear understanding of these roles is the first step toward orchestrating an effective AI strategy. By leveraging each AI capability - predictive AI’s foresight, generative AI’s creativity, optimization AI’s precision and agentic AI’s autonomy - you can create a harmonious blend of innovation that enhances customer experiences and drives business growth.