ICP-LLM
ICP-LLM: New version released – markedly easier to use, markedly easier to integrate
The Intelligence Collection Plan (ICP) is one of the core structured analytical techniques in NATO doctrine. Commander's Critical Information Requirements (CCIRs) are broken down into Priority Information Requirements (PIRs), these into Specific Information Requirements (SIRs), and finally into Essential Elements of Information (EEIs). The value of the method is not in question. The effort involved is considerable, however: a complete ICP can easily run to several hundred EEIs.
ICP-LLM, developed in cooperation between the Northern Business School, Strukturierte Analyse Deutschland and the NATO Civil-Military Cooperation Centre of Excellence, automates precisely these lower levels. Analysts can spend their capacity where it makes the greatest difference: sharpening and prioritising CCIRs and PIRs. The result is an ICP that is available sooner and more consistent in its structure.
What has changed in the new version
The feedback on the first release was unambiguous: the concept convinces, the handling needs to get simpler. That is where this version focuses. Every substantive change concerns usability and integration.
- New user overlay. A cleaner interface guides the user through configuration and execution. All settings sit in one place and are clearly labelled.
- Built-in version control. Earlier iterations of an ICP are stored automatically and remain traceable, so nothing is lost when a plan goes through several rounds of revision.
- Editing and extending existing ICPs. Plans already in use can be read in, expanded and carried forward instead of starting from scratch each time.
- Full ICP generation from CCIRs alone. The entire tree, from PIRs through SIRs to EEIs, can be derived from the top-level information requirements.
- Change highlighting. Newly generated and modified entries are marked, which speeds up the review of large plans considerably.
- Simplified model interface. Model access has been decoupled, so providers other than OpenAI, including locally hosted models, can be connected.
Why the barrier to entry is low
ICP-LLM runs inside Excel, which is the environment in which Intelligence Collection Plans are built anyway. It ships as a complete ZIP package: embedded Python runtime, all libraries and every required component are included. No installation, no dependencies, no programming knowledge. Download, extract, unblock the file, paste an API key, run.
Try it, and help shape it
The fastest way to form a view is a test run with one of your own CCIRs and a handful of PIRs. That takes a few minutes.
ICP-LLM is released under the Apache Licence 2.0. The source is open, and the tool may be adapted to your own requirements, extended and embedded into existing workflows. Feedback, bug reports and contributions are explicitly welcome and feed into further development. This release is the clearest evidence of that.
Contact: niklas.koch[at]i2sm.nbs.de
Note: ICP-LLM processes the content entered through the interface of the selected language model. When using external providers, no classified, personal or otherwise protected information may be entered.
Rethinking decision relevance: Using ICP-LLM to support intelligence work
Intelligence serves to give decision-makers in the military, security agencies, and private security an information and decision-making advantage. What is relevant for decision-making, for example, in strategic forecasting for the armed forces, in crisis management for a company, or in crime analysis for police case processing, should be defined by those who have to make the decisions themselves. For this to work, the provision of decision-relevant intelligence products is a prerequisite.
In security policy and criminal analysis practice, the Intelligence Collection Plan (ICP) is a key tool for targeted information gathering. It translates overarching information needs – such as Critical Information Requirements (CCIR) and Priority Intelligence Requirements (PIR) – into concrete Specific Information Requirements (SIR) and Essential Elements of Information (EEI).
By structuring and organizing the research and analysis process, ICP contributes significantly to more effective teamwork. It enables the targeted delegation of specific EEIs, supports debiasing in the analysis process, and makes it comprehensible and verifiable for third parties.
However, manually creating an ICP is time-consuming, error-prone, and often influenced by cognitive biases. Advances in generative artificial intelligence – particularly large language models (LLMs) – open up the possibility of partially automating this routine process. Generative AI can substantially simplify routine tasks in intelligence planning. Consistent human control ensures the responsible use of LLMs.
On November 3, 2025, the Intelligence Collection Plan – Large Language Model (ICP-LLM) will be officially unveiled. This innovative tool, developed in close collaboration between the Institute for Intelligence and Security Management (I2SM) at Northern Business School (NBS), Structured Analysis Germany, and NATO's Civil-Military Cooperation Centre of Excellence (CCOE), will then be available for free download here. Stay tuned!
The development of the ICP-LLM is continuing as part of a research project. Please feel free to share your comments, suggestions, and experiences with using our tool with us. We look forward to receiving your feedback: niklas.koch[at]i2sm.nbs.de