Planning means making trade-offs under constraints, based on data and signals that are sometimes incomplete. AI in planning is becoming increasingly widespread, but the use cases that deliver real operational value are still not always clearly identified. Guilhem Delorme and Hugo Van Straaten share their perspectives on organisational maturity, the capabilities available in Anaplan and the evolving role of planners. Their analysis brings together three dimensions: business decisions, the contribution of AI and the technology architecture that makes it possible.

The four approaches to AI in planning

Not all organisations are progressing at the same pace. Guilhem Delorme identifies four approaches, reflecting different levels of ambition and maturity.

• General-purpose chatbots: A conversational assistant is made available to employees. This approach allows organisations to explore potential use cases, but remains disconnected from actual planning decisions. An assistant that has no knowledge of the company’s master data, business rules or performance indicators can only provide generic answers.
• Business-specific use cases: AI is applied to a specific process, such as demand forecasting, assortment recommendations or detecting signals that could affect volumes. The benefits become measurable because they are linked to a clearly identified decision and a tracked performance indicator.
• Agent Studio: Users configure their own agents and make them available to their teams. This approach broadens the range of use cases, but requires a clear framework, with defined responsibilities, validation of results and control over the data being used.
• Domain-wide transformation: AI is integrated into the processes, roles and decision-making practices of a business function. This approach brings processes, organisational structures and technology choices together within a shared transformation roadmap, following the progressive methodology described in our approach to planning projects.

In all four cases, the starting point remains the same: identifying the decision or action to improve. The next steps are to assess the available data, clarify the user’s role and define how the resulting benefits will be measured.

Anaplan and AI: current and upcoming capabilities

Anaplan‘s AI capabilities fit into this progression, although they are at different stages of maturity. They address different needs, and their relevance depends on the targeted process, data quality and the maturity of the teams involved.

• Machine learning for forecasting: Available for several years. Models use historical data and available signals to support demand forecasting. They enhance analysis without replacing knowledge of commercial and operational events.
• AI assistants embedded in applications: Available today. Users can query their business models directly from the interface, understand variances more quickly and explore assumptions without leaving their working environment.
• CoModeler : Available today. This capability is designed to build applications and models from natural language instructions, without going through traditional modelling processes.
• Agent Studio : Coming soon. This component orchestrates agents, platform-native skills, user instructions and connections to the company’s broader agent ecosystem.
• AI Gateway (MCP) : Being rolled out. This capability makes planning data and models available to the wider enterprise AI ecosystem, helping improve decision-making beyond individual system silos.

« Agent Studio will orchestrate both agents and skills native to the platform, prompts created by users, and connections to a broader environment of agents within the company’s ecosystem. » Hugo Van Straaten

These capabilities expand the ways users can interact with planning models. They do not replace functional design or the robustness of a planning platform, where business rules and the allocation of responsibilities remain decisions for the organisation.

What AI is actually changing in planning projects

The impact is first visible in project delivery. AI can handle document preparation, some project management activities and test script generation. These use cases reduce the time spent on repetitive tasks and free up capacity for scoping and business process design.

AI also simplifies how plans are analysed. It generates alerts, identifies variances and highlights weak signals, allowing planners to adjust their plans accordingly. However, this capability must remain part of a clear decision-making framework: AI-generated results need to be validated, explained and linked to an operational action.

One of the most significant changes concerns the nature of the data being used. Planners work with structured data from enterprise systems. They also need to interpret emails, meeting minutes, commercial information and market signals, which are often valuable but difficult to incorporate into traditional planning models.

« The vision, which is already a reality in some companies today, is to give planners AI tools that bring together both worlds: structured and unstructured data, finally combined in a single tool, readily available to them. » Guilhem Delorme

This combination is transforming the planner’s role. Planners spend less time searching for and consolidating information. Instead, they devote more attention to analysing the causes of variances, comparing scenarios and preparing decisions involving Sales, Finance and Supply Chain. Technology accelerates and enriches processes, while business expertise gives meaning to the results.

Real-world example: a demand forecasting project deployed in four months

An ongoing acquisition, a forecasting methodology that needed replacing and just a few weeks to scope the project. This was the context in which IRIS by Argon & Co and OneHive joined forces for a global food manufacturer on a demand forecasting project in Canada.

The sales forecasting application was deployed in four months. The project combined the configuration of the Demand Planning process in Anaplan, statistical and algorithmic modelling, and the integration of specific parameters such as promotional effects.

A project of this kind involves much more than a forecasting engine. Sales history must be integrated, data quality assessed, promotions accounted for, models calibrated and business adjustments organised. The results must then be made available through an application that teams can use every day, with performance monitored after go-live.

The project brought together two complementary areas of expertise. IRIS led the statistical modelling work, from data preprocessing to advanced modelling. OneHive accelerated the platform configuration, allowing more time to be dedicated to business-specific adaptations.

« We were able to measure an improvement in forecast accuracy at go-live that actually exceeded expectations. The results confirmed that performance was in line with what we had tested on historical data. And once planners had made their adjustments, we achieved even better results.» Hugo Van Straaten

This experience reinforces the lessons we have identified regarding the key success factors for APS and EPM projects: clearly defined roles, the right expertise brought in at the right time and a well-controlled functional scope. Model performance and user adoption improve together.

Conclusion: a transformation across three dimensions

AI brings new analytical, forecasting and automation capabilities to planning teams. Its value depends on the business framework in which it operates, the quality of the available data and the robustness of the underlying technology architecture.

The most successful projects address these three dimensions from the scoping stage: the decisions to improve, the relevant AI use cases and the components needed to integrate them into existing processes. This approach helps distinguish quick wins from more far-reaching transformations.

For planners, the challenge is less about delegating decisions to AI than about having access to better-prepared, broader and more actionable information.

Watch the full interview (in French)

Guilhem Delorme and Hugo Van Straaten share their perspectives on AI in planning in this seven-minute joint interview.