AI in Planning: Why Better Technology Has Not Automatically Delivered Better Performance

AI capability is improving faster than planning performance

Varun Anand’s LinkedIn article, “AI in Planning — why progress hasn’t translated into performance”, raises a useful challenge for the Supply Chain Planning community: technology has advanced from MRP to APS and now AI-enabled decision intelligence, but many organizations still struggle to show a clean link between planning technology and business performance.

The article contrasts planning with areas where AI value is easier to measure. In planning, value is often indirect: better service under volatility, fewer expedites, lower obsolete inventory, lower working capital, faster disruption response or more time for planners to focus on higher-value decisions. That makes Planning Performance harder to attribute, but not less important.

The useful debate: planning is not only a technology problem

The core argument is that AI in planning will not scale simply because the algorithms improve. Varun Anand identifies several structural barriers: decision design, ownership, organizational politics, data maturity, interoperability and trust. These are precisely the areas where many planning transformations remain fragile.

The most important point is decision design. Many organizations still describe planning through process flows, calendars, meetings and handoffs. AI pushes the question toward decision flows: which decisions matter, which trade-offs are being made, which constraints are binding, and who has authority when service, cost, working capital and commercial priorities conflict.

Decision ownership is where AI often gets stuck

The article is especially relevant because it challenges the idea that planning decisions are purely analytical. Inventory, allocation, capacity and service decisions are cross-functional. Finance may prioritize working capital, sales may prioritize customer service, operations may prioritize stability, and logistics may prioritize execution simplicity.

This means that Decision Ownership is not a secondary governance topic. It is the condition for AI adoption. If the organization has not aligned on the decision logic, an AI system may expose the conflict faster, but it will not resolve it automatically.

Data and interoperability still define the ceiling

The article also makes a nuanced point on data. Companies may not need a perfect data estate before starting with AI, but they do need trusted and connected data to scale. Modern decision intelligence systems can infer operational reality from transactional behavior, such as lead time drift, but that does not remove the need for common definitions, semantic consistency and strong Data Governance.

Interoperability is therefore not just system integration. Many companies have technically connected systems that still describe the business differently. If ERP, TMS, WMS, APS and BI tools calculate lead time, availability or constraints differently, AI pilots may look impressive in isolation while enterprise-scale planning value remains difficult to prove.

Trust is the final constraint

For planners and leaders, trust is not created by better user interfaces alone. Teams need to understand why a recommendation was made, how robust it is, where the data came from, what assumptions were used and what happens if the recommendation is wrong.

This is why fully automated “lights-out planning” remains a narrow use case rather than a general operating model. Routine decisions can increasingly be automated, but novel, ambiguous and cross-functional decisions still require human judgment, escalation and contextual interpretation.

The practical implication

The strongest takeaway is that AI planning value depends on the operating model around the technology. Organizations need to define the decisions they want to improve, the value levers they expect to influence, the data required, the ownership model, the integration architecture and the trust mechanisms that allow planners to use outputs consistently.

AI can improve planning capability. But the article’s warning is clear: many companies are still trying to deploy machine-speed planning tools into fragmented, human-speed organizational structures. Until decisions, ownership, data and architecture are addressed together, AI in planning will remain promising but uneven in performance impact.