Organizations choosing a planning platform spend months comparing features. Far fewer ask what the platform will run on. That question is where many planning programs stall, and artificial intelligence (AI) is making it impossible to ignore.
Business will increasingly run on intelligence rather than on individual software tools, and planning is where that shift shows first. The architecture around a planning tool now matters more than the tool itself. That is the view of two Keyrus leaders who see the problem from opposite sides: Robin Bulman, RVP of Sales Benelux and Germany, who works with finance, supply chain and workforce leaders on planning, and Tamrat (Tam) Ogubai, who leads data and AI work across the Benelux, DACH and the Nordics.
Most planning programs stall on data they assumed was ready
Robin often sees the pattern before a project starts. A manufacturer wants to plan in Anaplan, but its bill of materials data sits across several ERPs, with hierarchies and definitions that differ by country.
"Many organizations are ready to plan in Anaplan before they have the data to plan with. The ambition is there, but the inputs aren't." Robin Bulman, RVP of Sales Benelux and Germany, Keyrus
The business case gets built around the planning scope, and the data work is penciled in as a first phase. It rarely stays that small. In one recent case, a manufacturer had to put its planning project on hold before it began.
The gap is organizational before it is technical
Planning teams start with the business question. Data teams start with the technology that could answer it. Neither instinct is wrong, but when each side works alone, the business ends up with two partial answers to one problem, and nobody owns the whole.
"When planning and data work sit with different parties, problems quickly turn into a blame game, and the business ends up caught in the middle." Tamrat Ogubai, Data and AI Lead, Benelux, DACH and Nordics, Keyrus
Technology amplifies. Culture differentiates. The organizations that close this gap share one language, name one owner for the full setup, and involve both disciplines early.
Good architecture puts each job in the right place
In most planning programs, the bulk of the requirements fit the planning platform well. A smaller share, such as anomaly detection or reconciling data definitions across many source systems, can be better served by a data platform the organization already runs. Building on that existing investment is usually a stronger design than forcing everything into one tool or reopening the platform choice.
The planning platform's role stays central. Anaplan is where functions plan together and silos come down, and its own capabilities, such as Anaplan Data Orchestrator (ADO), handle much of the work of bringing data into the plan. A data platform complements it where data volumes, sources or transformations go beyond what planning itself needs.
"Not everything needs to live in the planning platform. The question is which tool is best placed for each job." Robin Bulman
The alternative is tools chosen one at a time, each with a single purpose, which recreate the chaos they were bought to fix.
AI raises the stakes for the planning core
Many people will never log into a planning platform. They will ask a question in their company's AI assistant and expect a correct answer. Large language models are good at that: natural language, summaries, explanations and quick exploration. The risk is letting a probabilistic interface become the place where planning logic lives.
Enterprise planning needs what a prompt cannot guarantee: repeatable calculations, controlled business logic, approvals, security and a full audit trail. Cost matters too. Running heavy planning calculations through a language model spends tokens on every question, with results that can shift from one run to the next. A planning engine calculates once and consistently, and the AI layer only has to explain the result.
"When someone asks their AI assistant a planning question, the answer should come from the planning engine's calculations and the company's actual data." Robin Bulman
The answer is to give each layer a clear job, and each group of users the layer that fits how they work.
In short: calculate in the planning platform, explain through AI, decide with humans, execute through workflow. AI makes planning easier to access. The planning platform makes it reliable enough to run the business. Tam calls this the difference between experimental AI (demos and pilots that rarely reach production) and industrialized AI, which starts with process redesign and clear ownership of results.
The goal is a human-orchestrated organization
Every thread here points to the same destination. Keyrus calls it the Human Orchestrated Model™ (HOM™): an organization where intelligence runs on a governed foundation, people stay in command of decisions, and performance is steered against measurable outcomes. HOM™ describes where an organization arrives. Planning platforms, data platforms and AI are the building blocks.
Tam compares it to building a house: a sound foundation lets value compound instead of resetting with each new tool.
"It is better to build step by step and deliver value at every stage than to over-promise and fall short." Tamrat Ogubai
Designing that system is the role of an architect of intelligence.
Operationalizing intelligence starts with the architecture
Planning has always depended on its inputs. As AI changes how people reach and question their plans, it depends on them even more. Before the next planning decision, ask two things: what will this platform run on, and who owns the layers below and above it?
To talk it through, reach out to Tam Ogubai on data and AI at tam.ogubai@keyrus.com, or Robin Bulman on planning at robin.bulman@keyrus.com.