What is Agentic AI and is it here to stay?

AI has infiltrated many areas of our lives and work, constantly evolving and taking over more and more otherwise tedious, time-consuming tasks. AI and machine learning have become increasingly accessible, supporting small tasks like writing an email as well as more business-critical ones like optimising operations and cutting costs.

Agentic AI is another concept that is slowly entering the vocabulary of professionals and the warehousing industry. Is it just another buzzword, or a lasting paradigm shift? How can it be used efficiently in the market, for what purposes, and what would be a potential timeline for adoption? In this article, we’ll explain what Agentic AI means and what warehouse managers can expect in the coming years.

‍

What is Agentic AI?

Simply put, Agentic AI refers to artificial intelligence systems that can “think” and act on their own, without human intervention. Compared to more traditional software, this technology is defined by “agency” and proactivity:

“Unlike traditional AI models, which operate within predefined constraints and require human intervention, agentic AI exhibits autonomy, goal-driven behaviour and adaptability. The term “agentic” refers to these models’ agency, or, their capacity to act independently and purposefully.” (source)

Agentic AI is built on generative AI models such as ChatGPT but takes them a step further: it doesn’t just suggest or generate, it can actually perform tasks. These agents operate based on a goal you set without needing approval at each step. They adapt in real time, analysing data, identifying patterns, improving from experience, “think”, interact with other tools and make decisions to enhance the overall user experience.

‍

How are agents built and deployed?

Building and deploying Agentic AI within an organisation should start with clearly defining the objectives, expected benefits, success criteria, and relevant metrics. Understanding the potential outcomes and creating a testing environment will pave the way for successfully adopting this technology and improving business processes.

Simple agents can be built in tools like Copilot studio, with no coding experience needed - a practice the industry calls “vibe coding”. More complex use cases may require agents to be built and deployed in a code-based environment, with a more technical team and knowledge of programming languages. This approach adds greater capabilities, giving managers more control over what the agent can do and which systems it can connect with. In advanced tech environments, agents can even collaborate with and manage other agents, functioning as a digital team.

One of the great benefits of launching agents resides in the fact that they are much easier to adapt to changing requirements compared to traditional automations. For example, instead of having to rewrite code or rethink the entire automation, adapting Agentic AI can often be as easy as just adjusting the prompt.

Based on what the agents do, once deployed, they can operate fully autonomously or require different levels of human supervision and approval loops.

‍

‍

Key areas where Agentic AI will show impact

We see three main areas where Agentic AI will show impact first:

Knowledge work orchestration, with agents coordinating tasks across multiple software platforms. Basically, tasks such as research, planning, reporting, executing can be delegated to AI agents to automate the orchestration of multi-step processes. They “handle” processes end-to-end, boosting productivity, reducing time spent on the tasks, and allowing human professionals to focus more on high-value tasks and problem solving.

“Orchestration platforms automate AI workflows, track progress toward task completion, manage resource usage, monitor data flow and memory and handle failure events.” (source)

Customer support and operations, where agents that can resolve issues from start to end without escalating to a human. It is not just a chatbot that answers a pre-defined series of questions, but a tool that can “understand” the problem, look for solutions and implement it. If they cannot solve the issue, they will involve a human, but until then, many repetitive problems will get sorted faster. Working with such a customer-support agent will feel natural, letting people feel they really get help and improving customer service outcomes.

Physical environments, where agents will eventually act as the “brains” that coordinate between systems and robots or machinery. What this means is that AI agents will take information from warehouse management systems, for example, and use it to set tasks for on-the-ground systems or even people.

This emphasises the critical importance of high-volume, qualitative data collected in organisations, as the ground based on which decisions will be made by AI. In a warehouse, an AI agent could pull stock data sets from the digital twin platform, detect if a SKU is running low and trigger a replenishment order, showcasing strong AI capability in real-time decision-making.

‍

Applications in warehousing

For the warehousing industry, the main value will come from what we’ve mentioned above: connecting the “brain” (management systems, analytics, and tools) to the warehouse’s physical assets.

The first application of agents in this environment will probably be in supporting decision processes and orchestrations - running “what if” scenarios, dynamically reprioritising tasks, and closing the loop between process data, optimisation and execution.

In a warehouse, an AI agent - or system of agents, could be used for inventory management, constantly looking and adapting stock, for maintenance - to ensure seamless operations by detecting when hardware needs to be replaced and doing so, but also for financial tracking and reporting, supply chain management, problem solving, or workflow management.

Agentic AI could be used to detect if certain SKU orders are growing and direct a robot to bring more inventory to forward picking areas, to adjust labour assignments or to support the transition to greener operations by optimising the use of resources and helping to improve customer satisfaction.

Agentic AI paves the way to the Adaptive Warehouse that learns, adjusts and improves not under human supervision, but in collaboration with humans.

The expected benefits are obviously related to productivity, operational speed, error and costs reductions - early trials of using AI-driven task orchestration in warehouses have shown an increase of 15-25% in productivity, “before adding any physical automation (...) through better task sequencing, reduced travel time, and elimination of manual coordination errors.” (source)

The difference between this approach and the more traditional one is that people don’t have to implement the changes, but they still maintain oversight and control, depending on the task.

‍

Risks and how to mitigate them

The deployment of AI agents will happen gradually and for narrow, very well defined tasks to prove their value. While in the next 5-7 years we might see more widespread operation use, with supply chain management leaders estimating that by 2028, 33% of enterprise software platforms will include Agentic AI, this depends heavily on the benefits, trust, integration with existing systems and ROI validation especially in complex physical environments such as warehouses.

Safety is critical when operating machinery and large volumes of products, so trusting that AI agents will make the right decision about, say, sending robots and/or people to a certain area of the warehouse is key. Ensuring security measures are in place for data integrity and system safety is also essential.

That trust can be built only over time, through training and transparency. Explainable AI -”a set of processes and methods that allows human users to comprehend and trust the results and output created by machine learning algorithms” methods can be used to show how and why decisions are made, while keeping people in the loop and involving them in Agentic AI projects.

People want to get involved - a McKinsey survey showed that 48% of US employees would use more AI tools daily if they got formal training from the organisation, while 45% said that they would do it if there was a “seamless integration” into their workflows.

Lack of integration with existing tools - and especially with legacy systems, will only create more friction, slowing down the adoption pace. Another obstacle to deploying and using AI agents in the long term is tied to data - a lack of high-quality, real-time data will lead to poor outcomes in the real world.

Improper deployments where ROI is not validated is a high risk. To see results, the entire process has to be well defined, supported by data, with innovation champions working to streamline routine or error-prone tasks.

Last but not least, there are of course, legal and financial risks associated with how agents work, what information they have access to and the business impact of the outcomes, which is why close human supervision for complex tasks remains mandatory.

‍

‍

Conclusion

In the next few years, Agentic AI might handle repetitive tasks and decisions, letting people to focus on strategy, supervision and high-value actions. If these digital co-workers manage to prove their value, we might see a new type of collaboration between humans and agents, where the latter would end up orchestrating the work of other agents and even people too.

For this to happen, the industry needs to gain trust and to foster a culture of innovation. Careful planning, workforce training and a clear perspective of the benefits could lead to the successful adoption of Agentic AI in warehousing and, consequently, to increased productivity, agility and adaptability.

It’s not a temporary hype or buzzword, but a potential transformational technology. It is still maturing and we’re still operating on many predictions, but as the market and customer expectations are evolving faster than ever, the need for the right technology to support them is obvious.

All signs point to the fact that Agentic AI is here to stay, across many industries - warehousing included. They will become, over time, a foundational part of the Adaptive Warehouse and a critical enabler for business processes and customer service innovation in the real world.

Read more about the Adaptive Warehouse here.