Why use Agentic AI?

The hype through 2024 was about GenAI, but sales appear to have fallen short of expectations, so in 2025 the hype has moved to Agentic AI. The level of hype is not a signal to buy, even though some software suppliers have added AI features within their applications.

The previous blogpost identified Agentic AI as “a development that can be of interest for supply chain professionals”. To use AI in business is not an upgrade from an older system to a newer version. The development will be a journey that involves many people, some of whom will not be willing participants.

Before commencing the journey and to ensure a level of acceptance, an agreement is required from all managers who ‘need to know’ about why the organisation should consider using AI in its supply chains. Is the goal to make the current Operations Planning processes faster; to re-define the planning and scheduling processes; to improve efficiency in warehouse operations, or the effectiveness of Procurement, or something else?

Plan the journey

The outline implementation plan, including resources required and budget should have a five-year horizon. Following approval, selecting a competent software supplier will be critical and they will commence the journey with an introduction to affected staff about the new way of working. Next, a pilot project of a task or ‘stand-alone’ process that relates to supply chains is selected.

This implementation enables supply chain managers, professionals and all affected staff to understand and gain trust in the technology and supplier. It also builds confidence to address cost, time and other challenges and to demonstrate system and people capability and the value for users. A pilot requires its own project plan with expenditure approvals identified at each milestone.

The pilot can be selected from tasks in supply chains which cause capacity constraints. It must be structured, rules-based and repetitive, with a relatively substantial volume of transactions. Alternatively, a ‘stand-alone’ process is selected, but it must not require integration with other applications. Examples of processes are: supply chains risk analysis; supplier risk analysis; reviewing supply chain regulations by country or inventory optimisation calculations. The previous blogpost discussed a ‘stand-alone’ pilot which automates the Bill of Lading process.

At this stage, success should be measured by the elimination or simplification of steps in the task or process. Do not replicate the current method, but do it faster. When a pilot project is accepted as successful, it can be incrementally expanded. The previous blogpost stated that the developers planned to build on the successful Bill of Lading trial by implementing more tasks within the container shipping movement process.

The next stage is ‘workflows that travel across functions’. Here, teams co-ordinate and (hopefully) receive intelligent responses as agents communicate across systems without the intervention of people. This stage allows Agentic AI to initiate and escalate co-ordinated actions, based on signals (or triggers) viewed as critical to the business, such as a supplier disruption or a change in international trade policy. Importantly, a governance framework is implemented that enables adherence to policies and regulations, tracing escalations and performance tracking for accountability of the agent actions.

In addition to a governance framework, the system requires standardized data and processes, with a data governance framework, so that systems and people understand the data element structure in the same way. As a Supply Chains group move to a more predictive style of work, it must have responsibility for the selection and structuring of data that is collected from internal and external sources.

The final stage of an Agentic AI development is portrayed in articles (but yet to be seen) as a network of agentic systems that “operate with shared context, awareness of policies and regulations and autonomous decision-making capabilities”, that learn as each incident occurs in a supply chains network.

Also noted in articles is the requirement to include a desired level of ‘orchestration’ for bi-directional redesign of supply chains between the supply markets and sales channel markets. The redesign is deployed through the Sales & Operations Planning (S&OP) process, across the capabilities of source, make and deliver and over the strategic and tactical time horizons.

People and technology work together

Agentic AI is expected to change the role of people from routine execution (doing) to guidance and oversight (thinking), so just installing technology will not ensure success. As learnt since the early introduction of IT applications, and stated in the previous blogpost “the big challenge for Agentic AI in supply chains will not be technical, but people”.

To gain acceptance, supply chain managers and professionals will need to “learn and train with staff to develop the governance protocols around agent autonomy”. And importantly, as the system interacts across supply chain functions, “so the organisation structure changes from a collection of departments to a Supply Chains group platform”.

People from the Supply Chains group must be involved early in the implementation, to build across-function interaction, co-operation and trust among all employees who will work with, but not on, a different type of system. Build into the implementation plan allowances that provide:

• Information (no hype) concerning the project implementation: ‘why’, ‘goal’ ‘timeline’ and ‘how’. Explain how the initial pilot and eventual system are intended to work
• Training programs, presented at a realistic learning pace, that help to improve the skills which people can use in their current and future job
• Training to improve the management of data, including the elimination of biases and inaccuracies
• Training programs and tools that assist people who show an aptitude for using the technology to engage and help others in the group

Encouraging multi-functional teams to collaborate across all aspects of supply chains will enable a business to more easily change their organisation structure from individual business units to a platform based operating model that better reflects the flows of supply chains.

A different organisation structure and the operational requirements of Agentic AI will require a rethink of leadership in the business. As the system is designed to evaluate future risks, so must managers. Therefore, the effectiveness of teams becomes at least equal to being efficient. This requires managers with the capability to work among multi-discipline teams, as they improve effectiveness for more resilient and adaptable supply chains.