What would a Martian expert in supply chain planning and technology think when looking back at earth? With all technology available on earth would it continue with a 30+ year old, sequential, cascading, rather sluggish, manual planning process with a lack of focus on decision digitisation, decision quality and decision learning?

Or would it start with a blank sheet and reimagine IBP?

Closing the planning automation gap

After more than hundred years of automating our physical assets in the supply chain, we now have automated warehouses, trucks, productions plants and delivery drones. We can use digital copies of these assets to simulate scenarios to trade off risks and make informed decisions.

Despite all the investments made in advanced planning systems over the last two decades, automation has largely bypassed the planning function. Estimates are that 32% of work can be automated, and 42% augmented. Specifically for supply chain, Gartner estimates are that short term planning decisions will be automated for 65% in 2026. Some go as far as 80%. It will likely depend on industry and function.

However, one McKinsey study found that 7% of CPG companies apply end to end autonomous planning. An Accenture study found that only 3% of businesses applies autonomous supply chain execution. This enormous automation gap in supply chain planning and execution is starting to be addressed by innovators and early adopters.

Digitizing S&OE and IBP Decisions

IBP decisions can be differentiated between machine centric S&OE decisions in the short-term horizon, and human-centric IBP decisions in the longer-term horizon (Regeer, van Hove, 2021).

Below figure shows an overview of demand & supply balancing options in a $50 billion global beverage business. It’s over a decade old, but still relevant. Many supply chain decisions are still the same, and repetitive, even across industries. Repetitive decisions can be digitized. The options to choose from and trade-offs will largely be known, so are the data, plans & insights required.

Decision trees for the most common IBP decisions can be made like this for Finance, Sales, Marketing, Operation & Sourcing, highlighting where there is automation and augmentation potential. Technology can have the planning processes automated and the digitize the decisions trees, data capturing, analysis, and trade-offs that support them.

Closing the traditional IBP gaps with AI

Traditional IBP as defined by its early thought leaders is sequential as can be seen in below figure. It still works, but is not fast and responsive enough in today’s world. Sometimes decision have to be made faster than the IBP cycle, for which I’ve already suggested a solution. Strategy integration remains a challenges and, as I’ve highlighted before, IBP is not decision centric enough. Data collection, planning, reporting and the creation of decks and recommendations, is still largely a manual, and cumbersome, exercise for planners.

Systems of intelligence, which include AI supported decision making, will provide intelligent automation that replaces the human planning process as well as cognitive automation that augments a planner’s decision making with predictions, insights, and recommendations (Regeer, van Hove, 2021). We already have seen examples of S&OE decisions go from a day to merely seconds due to automation (Van Hove, 2023).

This automation will extend further into the planning process, maintenance of planning data and forecasting algorithms, and other planning decisions and actions, including for IBP. More complex and high value IBP decisions that can’t be automated, will see incremental levels of augmentation to support decision makers. Prescriptive recommendations will advice the planner what to do in the IBP horizon.

Not all AI is the same. And you have to thoroughly understand your business processes and your decision making processes to assess where AI can best assist you. Or even better, to rethink your business processes and operating model. As Lorien Pratt, co-inventor of Decision Intelligence says; “Data, technology and AI must take a back seat to diligent understanding of the decision that they support.”

One of the most relevant AI’s for IBP will be the Intelligent Agent (IA). In Artificial Intelligence an intelligent agent (IA) is an agent acting in an intelligent manner; It perceives its environment, takes actions autonomously in order to achieve goals, and may improve its performance with learning or acquiring knowledge (Wikipedia). When applied to decision making, at my employer Aera, we call these IA’s Cognitive Skills.

IA’s can be applied to decision trees, using pre-negotiated decision policies, in order to augment, automate and learn from decisions. On top of a continuous, concurrent, largely automated planning engine, intelligent agents can be used to detect variations to plan and automate and augment decisions where a business deems relevant across all defined decision trees. A decision system of records will be created, which creates a decision library which over time can learn from decisions made.

Imagining a Digitized, Continuous, Decision Centric IBP

When we digitize our most common, repetitive IBP decisions, close the traditional IBP gaps and use the described AI capabilities, we can imagine a new, decision centric IBP model, that is strategy and goal driven, is able to make more frequent decisions, and learns from its decisions. Below figure shows a possible digitized, continuous and decision centric IBP alternative, which I will discuss in further detail.

Set strategic & decision policy setting: this is a leadership meeting, where the strategic priorities by function and trade-offs between functions are updated when required. The meeting can be periodically scheduled, or it can be ad-hoc, at the moment disruption hits or market circumstances change, and priorities needs to be adapted. This solves two problems in traditional IBP. Firstly, IBP can react faster to change than a monthly IBP cycle. Secondly, strategic priorities are integrated by design as a starting point. If no strategic direction or decision policies change, no meeting is required.

Maintain digitized IBP decision library and Intelligent Agents: using updated strategic guidance, the digitized decision trees and the intelligence agents in the decision library will need to be updated. Applying strategy changes directly to decision policies, drives decision centricity. Planning priorities, assumptions and decision policies, decision goals and targets will be updated. As well automation and augmentation thresholds in the IA’s. Boundaries for the planning engine across Source, Make & Deliver, for which it has to provide a feasible plan, will be largely automated, but some might need human maintenance.

Run Planning Engine: after being fed the latest strategic choices, assumptions, planning goals and tradeoffs, a concurrent planning engine will run to create a new end to end value chain plan, that is commercialized and feasible from operational and sourcing perspective. The planning engine is basically a stripped down APS with the main goal to create a feasible plan, in a planning version, not directly in an execution version. The planning engine runs mostly autonomously and creates a baseline plan in the first iteration.

Decision Intelligence using IA’s: The first planning iteration will activate the intelligent agents (IA’s) that will automatically take decisions around basic gap closing activities. For example, for Marketing, product launches and timings within a category could be adjusted due to reprioritization (higher margin versus higher volume), or NPDs within a category will be reprioritized. For sales, promotional plans and prices can be automatically adjusted in some circumstances. In operations, the IA could swap product between production lines, or factories, or change logistics routings or mode of transport due to ESG tradeoffs. Sourcing could change suppliers, contracts, or interchangeable materials to get a better COGS and ESG score. A second planning iteration could perform some local or functional optimizations within the set boundaries to create a better plan according to the goals that have been set.

Some 100% human centric decisions (for example people & cultural decisions) or unique decisions that are not repetitive, will likely have to run parallel. However, when using a best practice decision framework, these decision flows can still be digitized and feed a digitized IBP decision log, where it can be recorded, audited and learned from.

Digitized IBP decision log: All automated decisions will feed in to the IBP decision log. So will the 100% human centric decisions, for which the decision flow is digitized. Planning and gap closing decisions that can’t be automated or reach a (value) threshold will go through a prescriptive recommendation cycle, where a user or the IBP team has to review and agree, change or disagree with the decision. If a decision can’t be made or goes beyond an IBP threshold, it can be flagged as an executive decision. All decisions in the decision log are transparent and auditable for all IBP stakeholders.

At any time, the decision log shows the total plan, gaps to plan and decisions, which are visible to the IBP manager and the whole leadership team. This continuous visibility of decision made, decisions outstanding, and decision impact, creates the opportunity to apply a continuous Integrated Reconciliation (a traditional IBP term that might needs rethinking) and preparation for on-demand IBP decisions.

IBP decision on demand: The role of the IBP team now becomes to facilitate the decisions to make the newly proposed plan the agreed active plan. The IBP team will check the gaps & outstanding decisions in the IBP decision log to understand unresolved decisions, ones that were escalated to the leadership team and other urgent and impactful gaps to plan. By simply sorting the IBP decision log on value impact, likelihood of success or other criticality criteria, the IBP manager can choose decisions that need to be made with urgency. This could be a few decision out of many that are outstanding.

In the unlikely situation that the new plans meet all the mid and long range targets within certain boundaries (for example projected EBIT is within 2%) the new plan could be copied as the active plan straight away. More likely, the IBP manager will call an executive planning review meeting, only focused on the most urgent planning decisions. After deciding on the few outstanding issues, the proposed plan will become the active plan, against which S&OE is measured. The focus can now shift on solving the next most important gaps, initiatives and decisions for the next iteration of the plan. In the meantime, execution happens against the plan that includes the most urgent and important updates. No need to wait for a monthly cycle, but if you want to, you could.

High impact initiatives in the plan, will have to be maintained in a workflow and linked to a (gap closing, maintain status quo or growth) decision. Initiatives where the IBP team is not confident of its likelihood of execution, a planning review meeting can be called, where the functional owners of the initiative will have to present the state of the initiative to the leadership team. These are quick, ‘Sharktank’ like, meetings, after which a confidence score and impact of the initiative can be updated in the decision log, together with adjusted activities and resourcing to support the initiative.

When the proposed IBP plan is copied to the active plan, the planning engine and intelligent agents go to work in the S&OE horizon. This is where we can envision a similar approach with a digitized S&OE decision log, only the level of automated decision will be much higher than for IBP decisions.

Decision learning: When the S&OE and IBP decision processes and the decisions are treated as data points and stored in a system of record, a decision memory can be created. Using a decision memory, machine learning or other techniques can be applied to learn from decisions. ML can estimate the likelihood of decision success and impact and advice a decision maker accordingly. For frequent S&OE decisions, we already see examples where Decision Intelligence technology provides a user with both a recommendation to change a source of supply, a mode of transport, a stock transfer order, or a safety stock setting, but also the probability of successful impact when accepting this recommendation. Over time, less frequent IBP decisions are likely to follow this example of ‘learning’ from decisions.

 

Conclusion

In this age of digitization, automation and AI, and the enormous progress that technology will continuously be making, we have to reimagine how our operating models can work better. To do this for IBP, thought leaders and planners have to let go of their ERP and APS system anchors and the IBP process dogma’s that have shaped their thinking for the last 20 years. That is not easy, but I just gave an example that it is possible.

I’m looking forward seeing other examples from you. Just take a blank sheet of paper and imagine for a moment being a Martian looking back to earth!

Photo credit: NASA