Most AI investment decisions in planning are being made under pressure to act rather than in response to a clearly identified problem. The organisations seeing returns are the ones that started differently.
The pressure is real. Your board has asked what you are doing about AI. Competitors appear to be moving. Vendors are presenting compelling roadmaps. And somewhere underneath all of that noise is a genuine question you have not yet been able to answer cleanly: what would AI actually do for your planning capability, given where you are right now?
The difficulty is not a lack of options. It is that most of the available guidance comes from people with something to sell, calibrated to organisations with data foundations and architectural maturity that bear little resemblance to yours. The result is a landscape where the gap between what AI can theoretically do and what it will realistically deliver in your specific context is large, and almost nobody is being straight with you about it.
The questions that matter at this stage are not about which vendor to choose or which use case to prioritise first. They are about whether your data foundations are good enough to support the investment you are being asked to make, whether the sequencing of capability investment is right for your current maturity, and whether the build versus buy versus wait question has been honestly answered before any commitment is made.
Those questions are not answered well by vendors or by benchmarks drawn from organisations at a very different maturity level. They are answered by practitioners who have navigated comparable decisions in comparable contexts — and by an honest assessment of what your organisation is actually ready to absorb.
What follows draws on BPC's corpus of recorded practitioner conversations at this specific decision stage — what tends to go wrong, what conditions tend to produce better outcomes, and what practitioners who have navigated comparable decisions say they would do differently.
The most consistent pattern is starting from the wrong question. Organisations are being asked what they are doing with AI — a supply-side question — rather than which specific decisions would genuinely be better if AI were applied to them — a demand-side question. The first question produces projects that are defensible in a board presentation but hard to justify once the initial enthusiasm settles. The second produces a shorter list of initiatives with a clearer path from experiment to embedded operation.
A related pattern is the pilot that does not reach production. The gap between a working experiment and a justifiable investment at scale turns out to involve considerably more infrastructure than the initial pilot suggests — including how decisions get executed back into the systems that run the operation day to day. That infrastructure is consistently underestimated in early pilots, and its absence is one of the main reasons promising experiments do not reach production.
Data readiness is also consistently underestimated. Not in the sense that organisations lack data, but in the sense that data exists somewhere without being available in usable form at the moment of decision. The distinction between existence and usability is where assumptions most often break down, and it is rarely visible until something external forces it into view.
The organisations seeing returns on AI investment in planning share a common characteristic: they identified a genuine operational problem first and worked back to whether AI could address it, rather than arriving with an AI tool and searching for problems that fit. That sequencing is straightforward to describe and surprisingly rare in practice.
What also tends to help is being honest about where AI is and is not currently well suited to the problem. Clean feedback loops, high data volumes, and bounded decisions are the conditions under which AI generates reliable returns in planning. Long horizons, sparse data, and judgment-intensive exceptions are the conditions under which human expertise still carries more weight. Knowing which conditions apply to the specific use case in question is more valuable than any vendor benchmark or analyst ranking.
On the data foundations question specifically: the inventory reduction argument tends to land better with finance than the time-saving or error-reduction argument, because it shows up on the balance sheet. For organisations struggling to get data governance funded on operational efficiency grounds, the working capital angle is often a more viable route into the same investment.
Timing: Wed 17 Jun · 15:00 BST · 60 minutes
Focus: Supply chain and planning leaders examining how to sequence AI investment in planning against their actual current position rather than an idealised future state.
Format: Consultant-supported discussion session grounded in practitioner experience
AI tooling is lowering the barrier to building internally and changing the architecture direction question before any vendor enters the room. This session examines what that shift means in practice for a mid-market supply chain function.
Examines the strategic context that should sit upstream of any capability investment, including operating model design, partner ecosystem constraints, and the shift toward AI-enabled best-of-breed components.