Are requirements different?

AgenticAI can provide opportunities to rethink the way that planning and scheduling processes in industrial businesses are designed, executed, and governed. But it involves more than the installation of a technology. For the Supply Chains group (Procurement, Operations Planning and Logistics), it requires a strategic review, a redesign of workflows, cross-function alignment and a structure of governance. All of which takes perseverance, resources and time, which few industrial businesses have.

Since the 1980s, these requirements have also been identified as necessary for the implementation of ERP systems. And there were failures, mainly due to an inability by senior management to authorise sufficient investment in resources and time. As ERP implementations can still fail, will implementations of AI in operations be different?

A recent survey

Over 450 supply chain professionals from around the world indicated in a recent Reuters Events survey some disconnected thinking about aims and objectives of supply chains and the role of AI. The report noted “The greatest barrier we see is not a lack of tools, but too many of them. Disconnected systems mean duplicated work, siloed data, and incomplete visibility”.

The survey noted that after all the talk about supply chains now being strategic, cost control continues as the #1 driver. The #2 driver is resilience and #3 driver is transparency and visibility through supply chains.

Containing costs appears to be the reason for Industrial businesses attempting profitability gains through implementing digital technology. The survey had nearly three-quarters of respondents citing operational efficiency (but not effectiveness) as a key benefit, two-thirds cite cost reduction and over half identify productivity gains.

However, optimising functions may improve function costs, but will most likely result in higher total costs, because the supply chains of a business is a complex, adaptable and non-linear system. So, aligning source, make and deliver outcomes to improve operating margins is next to impossible. The report notes the ‘genAI paradox’ (by McKinsey) – although many companies say they use (or ‘adopt’) AI, less than one percent of businesses achieve a positive profit impact from its use.

Inventory Analytics is stated in the report as the second most invested technology area and the fastest growing. But who is doing the inventory analytics? The lack of workforce skills is stated in the report as the main barrier to implementation – ‘many organisations underestimate the expertise required to extract value from technology tools’. As a logistics director stated years ago “if you know of people with the required skill set, I will have two right now”. The people challenge is compounded by poor quality data and difficulties with standardising data – challenges that have existed for at least the past 50 years!

The report notes that the way technology investments are made often undermines the intended gains. Although 90 percent of respondent companies expect to add new systems in the next 12–18 months, they are currently working across fragmented tech stacks. The number of logistics technology platforms deployed in tech stacks had a third of businesses with 3-5 platforms and another 20 percent with 6-10 platforms – “…loosely connected tools that create silos and duplicated work”.

For those of us learning about AI, definitions taken from Internet sites are:

• An AI platform is the infrastructure that supports development, deployment and management of AI models. It provides the tools, frameworks and environments to build and scale AI solutions.
• A tech stack is the software components and infrastructure layers as a collection of programming languages, frameworks, libraries and tools to build and run a software application. There are different types of tech stacks, according to a project’s requirements.
• A framework describes how a part of a project works (the front, the back…). A tech stack can include one or more frameworks.
• An AI application is a software program that uses AI technologies to perform specific tasks or solve problems. It is built on top of an AI platform and employs AI models to deliver functionality to end-users.

A challenge with the number of platforms is to get the platforms and applications to communicate. The report stated that about 30 percent of respondent considered their tech stacks had high interoperability, but 70 percent hade moderate or low interoperability. And the risk is magnified in businesses with supply chains that operate across multiple geographies and regulations.

Workforce skill capabilities and lack of interoperability are challenges that are common for shippers (also called Beneficial Cargo Owners (BCOs)) and logistics service providers (LSPs). In addition, respondents at BCOs require investment in data standardisation and analytics to forecast demand and manage inventory. Those in LSPs prioritize applications in operational execution and regulatory compliance.

Linking the control of costs in supply chains to implementing AI, is to optimise existing tasks rather than re-engineering supply chains. A re-design of supply chains and workflows requires mapping the organisation’s supply chains and workflows to identify corporate and user ‘pain points’, which few businesses have done. And, rather than redesign supply chains (or the value chains that also incorporate demand chains), the more common approach is to have AI operating alongside the current planning and scheduling applications, without addressing any of the current ‘pain points’.

There is an alternative approach

Instead of agonising over the justification, selection and implementation of a new technology, an alternative is to optimise the current supply chain planning and scheduling processes by simplifying them. Commence by mapping the supply chains, because an organisation needs to understand them. Using the map, remove data that is ‘redundant, obsolete and trivial’ (ROT) through the elimination approach ‘if it were not for what, could this be eliminated?’

• Map and understand your supply chains
• Simplify the workflows through planning and scheduling
• Identify supply chains using co-efficient of variability management (CoVM)
• Have ‘one plan’ for the business or profit entity
• Use ‘one plan’ to implement S&OP as the heart of the business – the plan incorporates risks
• Move to cross-function alignment through S&OP
• Measure ‘delivery in full, on time, with accuracy’ (DIFOTA)
• Reduce costs but not value. Aim for effectiveness not efficiency

Few companies have used statistical analysis to recognise and understand patterns within the planning master data of their supply chains, such as lead time and cycle time. These can change, especially with geopolitical winds. This can be a starting point for incorporating agentic AI, so that when the above list is completed, rebuild the operating model to be technology based and invest in people, teams and training.