8️⃣ ITONICS Prism AI — Ideate, Auto-Create Elements, Ratings, RTE, and Segmentation

1. ITONICS Prism — general information

2. Master Context — the heart and soul of Prism

3. Auto-create trend and technology radars

4. AI use cases — how to interact with Signals and Elements

5. Auto-rating — speed of change

6. Limitations

7. AI models in use

8. Data security and protection — Prism

9. Smart Actions — autofill description, and tags

10. Auto-generate and auto-evaluate ideas

ITONICS Prism AI

When in the workspace, to access PRISM, press the dedicated button in the bottom right corner.

PRISM chat is based on Claude 4.0 Sonnet, a LLM similar to ChatGPT, but the internal functionalities are split up into different agents who all communicate between each other based on what prompt they are executing, sometimes referred to as multi-agent debates or deep reasoning.

So far the different agents are:

• Supervisor: responsible for coordination
• Signals Agent: responsible for Signals pool
• Elements Agent: responsible for elements
• Knowledge Agent: responsible for all questions regarding the ITONICS Innovation OS

If you are interested in what is being discussed after your prompt, clicking on the little grey arrow will show the internal reasoning process.

Master Context

Get the most value out of each prompt and set a Master Context for your workspace.

What is a "Master Context"?

Large language models work best when you provide them with relevant background information.

The Master Context is a persistent block of text that gets infused into every conversation you start with the AI. Instead of repeating your company’s mission, priorities, or constraints before every prompt, the Master Context ensures the AI always has this baseline understanding.

Think of it as the lens through which the AI interprets every question, idea, or suggestion, helping it generate responses that are aligned with your company’s identity and strategy.

Because the Master Context is especially reliant on your use cases and your companies needs, ITONICS can only help you so much with crafting the Master Context. This is why it is crucial that you collect feedback concerning Prism from your colleagues and refine it yourself.

Why is a Master Context important?

• Consistency: Keeps answers aligned with your goals across multiple sessions.
• Efficiency: Saves time by avoiding repeated instructions.
• Relevance: Ensures outputs resonate with your brand voice, strategic direction, and users’ needs.
• Quality: Increases the chance that suggestions are practical, feasible, and tailored to your use cases.
  
  • For example, if you are having issues with, on what basis Prism creates elements and would like more information to each element, include a section about what you would like to have included in Prisms response when creating elements.

How to craft the perfect Master Context

A strong Master Context addresses key dimensions of your organization. You can think of it as answering the following guiding questions:

1. Core Identity

   • Who are you as a company?
   • What values, principles, or cultural traits guide your decisions?
2. Strategic Goals
   • What outcomes are you working toward?
   • What KPIs or success metrics matter most?
3. Priorities & Challenges
   • What problems are you trying to solve?
   • Which trade-offs (speed vs. safety, cost vs. innovation) must be considered?
4. Technology Areas & Capabilities
   • What platforms, tools, or technologies define your domain?
   • Where do you innovate, and where do you rely on partners?
5. User & Market Focus
   • Who are your users or customers?
   • What do they expect in terms of experience, reliability, and trust?
6. Constraints & Non-negotiables
   • Regulations, compliance, safety, sustainability, or ethical standards that must always be respected.
7. Use Cases
   • Outline use cases for the AI, describe them as in-depth as possible and what Prism should do when prompted.

Important considerations — improving Master Context as you go

Getting the perfect master context is crucial to mastering Prism, it is not as easy as it seems. While getting a first version is pretty easy, by using a LLM for example, this will not be perfect, encourage coworkers to provide feedback to your workspace administrator, if they achieved a particularly good response with a certain prompt, ask the Internet and refine the Master Context as you go.

If you have different use cases for Prism, specify each one as accurately as possible, this will greatly affect the results that the AI will produce in regards to that specific use case.

As a basic example, you have a formula to calculate which ideas should proceed in the innovation funnel, something like:

If "expert rating" >= "high" then status = "next stage"

Include this in the Master Context:

Use Cases:
- Advancing Ideas for Idea Innovation Funnel: when prompted to advance ideas in the innovation funnel, use this formula and stick strictly to it, if you have issues, explain them in your response:
         "If expert rating >= high then change status to next stage" 

Now Prism will always use this formula to calculate that use case.

And most important of all, play around and try stuff out, as the token limit has been removed for everyone for now, feel free to test Prism, how you can best integrate Prism into your workflow and:

Setting the Master Context in the ITONICS Innovation OS

To set the Master Context, open PRISM via the designated AI button  in the bottom right.
Then locate the three dots [2] at the top right of the now expanded side panel and select "configure" [3].

Now write or paste the Master Context in the text field.

At the bottom of the popup you can see another dropdown menu [4] to select an element type. Select which element type best represents this workspace, this is not necessary, but if you already have a "strategy" element or something similar in your workspace, it is recommended to select that and if possible adjust the criteria below, under "advanced filter"[5], to select exactly the elements you want. The advanced filter works just like the regular filter.

Last but not least, make sure "Enable for new chats" [6] is enabled, press "Save" [7], and you are ready to go.

Segment and rate radars

After creating a radar with its designated element types, let PRISM do the job: Automate radar segmentation and let the AI rate the elements based on your, company specific, catered to your needs, master-context.

Click on the Prism button in the bottom right corner, a new side panel will open. In the top right corner you have the option to close Prism, expand the side panel to full screen or the three dot menu, with the option to set a Master Context for this workspace.

To segment your elements, simply ask Prism and don't forget to specify in your prompt which element type you want to segment.

Rating elements works similar to segmenting them. Specify which elements and fields you want to include in the rating process and the AI will do the job.
But be aware, rating is still based on generative AI, while Prism is much faster than a human at rating, the nature of generative AI is still to guess and therefore you may experience that ratings don't follow your standards.
It is recommended to use qualified personnel at your company to rate important elements.

Use Prism to interact with signals and elements

The Innovation AI can interact with signals and elements within the workspace to enhance your prompts. Below you find an overview of valuable prompts to execute different use cases:

Prompt playbook for signals:

Speed of Change index

To activate the speed of change index, head to the element type configuration of your workspace. You can either create a new rating type group [1] or use an existing one. Search for "Speed of Change" and add them from the dropdown [2]. 

You can rearrange rating criteria by dragging and dropping and hiding the ones you do not want to use anymore by clicking on the eye icon. Once you are happy with your rating setup, hit Save Changes. 

You can also compare the speed of change across multiple content elements when you head over to the technology or trend radar. Open the radar view configuration, select Distance, and choose either Speed of Change [90 Days] or Speed of Change [180 Days].

You can now compare the trends and technologies on the radar by the Speed of Change of the last 90/180 Days.

When you watch a content element, you will receive an email notification if the speed of change changes. 

Limitations

While PRISM is a powerful assistant for creating, summarizing, and analyzing innovation content within the ITONICS Innovation OS, there are still certain technical and functional limitations to be aware of. Understanding these helps you get the most accurate results and avoid unexpected behavior.

Element agent capabilities

The Element Agent is responsible for creating and updating elements in your workspace. However, its abilities are currently focused on specific fields and operations.

Prism can:

• Create new elements within your workspace
• Update existing elements, including the following fields and property types:

  Title, summary, rating properties, dropdown properties, numerical properties, tags

Prism cannot (yet):

• Update text fields including regular, and RTE fields, from within the chat window
• Add or manage attachments
• Set or modify relations between elements
• Update all elements in bulk
• Perform property-based searches (see below for details)

Search behavior

The element search with Prism is structured in two steps.

1. General search
   1. Prism searches for elements that fit the general criteria, searching only text fields (titles, summaries, regular text fields, RTE fields, element type).
2. Specific search
   1. After that, Prism accesses all properties of these, selected, elements and chooses only these who fit best to the initial search criteria specified by the users prompt.

Implications — keep in mind

• Since the initial search is not property-aware, there is a high likelihood that relevant elements are not found simply because they did not match the text search criteria.
• Searches may sometimes miss relevant data if your workspace relies heavily on structured fields rather than textual descriptions.
• For now it is recommended to provide an accurate description of each element in its respective summary or description field.

Understanding and context limitations

Prism aims to execute every request as completely as possible, but unlike a human, it cannot always judge how well it has understood your intention. If a request is ambiguous or incomplete, PRISM might still try to perform the action rather than asking for clarification, which can lead to unexpected results.

To minimize misunderstandings:

• Write clear and specific prompts
• Include relevant context or workspace details
• Confirm updates before performing large-scale actions
• Configure the master context accordingly

Message context

PRISM uses only the last three messages of the conversation as active context for generating its next response.

This means:

• If your new question references something discussed earlier than three messages ago, PRISM may no longer remember it.
• If the last three messages are particularly long, PRISM might attempt to interpret your latest prompt in relation to those messages only, sometimes leading to results that appear unrelated or inconsistent.

To ensure the best performance, try to keep conversations focused on one task at a time, or restate key information from earlier in the chat when continuing a previous topic.

AI models in use

Since different AI models have vastly different use cases, Prism is using six different models, each for a different use case:

At ITONICS, data protection and privacy are at the core of how Prism AI operates. Every interaction with Prism is designed to keep your organization’s information secure, private, and fully under your control.

Secure structure — tenant, workspace, and user

Each organization runs in its own private Tenant Space within the ITONICS Platform.
This ensures your data remains isolated from other customers and accessible only to authorized users.

• Tenant — Your company’s dedicated, secure environment where all data and AI features (like Prism AI) are contained.
• Workspace — A protected area for team collaboration. Data and AI contexts are encrypted and access-controlled.
• User — Each user is authenticated and granted only the permissions they need. All communication is encrypted end to end.

Secure AI processing

Prism AI works within a secure AWS environment to handle every request safely and efficiently.

1. Encrypted communication — All messages and data exchanges are protected with modern encryption (TLS 1.2+).
2. Verified access — Every request is validated by tenant, workspace, and user identity before processing.
3. Context-aware responses — Prism AI uses workspace-specific instructions to tailor outputs, without exposing data outside your environment.
4. Isolated AI agents — Specialized agents (for Signals, Elements, and Knowledge Base) operate inside a private, access-restricted environment.
5. Data privacy — No prompts or responses are stored or used for AI model training. Processing happens only within regional (EU/US) AWS environments.

Infrastructure and privacy protection

Prism is built on a multi-layered security framework:

• All data is encrypted at rest and in transit.
• Every operation runs inside a private, access-controlled AWS network.
• Full audit logging ensures transparency and compliance.

Your data stays yours.

Prism never shares, sells, or exposes customer information — ensuring your innovation activities remain private, compliant, and secure.

Smart Actions — autofill description and tags

We help you reduce the burden of adding descriptions, or tags by our smart action feature. 

Autofill description

Go to any content detail page and find the smart action button below the content title or in a rich-text-field. Choose Autofill Description from the Smart Actions  [1] in the header or navigate to the Description section [2]. A modal will open and create a suggested description for you [3]. 

When designing a campaign, there is even the option to decide on a preferred storytelling template [3] for auto-filling the description. You can choose between the following options which will generate a description based on your provided Summary: 

• Campaign Canvas: This canvas will produce a structured description covering the following points: Campaign description, Objective of the Campaign, Stakeholders & Audience, How To Participate, Guiding Questions that help idea submitters understand what's expected and a short and engaging closing statement.
• Storytelling Tactics:
  • The golden circle: This Storytelling method starts with "Why" (the purpose or belief), then "How" (the process or values), and finally "What" (the product or result) to inspire and connect with an audience.
  • Conflict and Resolution: This Storytelling method introduces a central challenge or problem (conflict) and then presents a solution or outcome (resolution) to engage and inform the audience. The resolution in that case is the campaign itself.
  • In media res: This Storytelling method begins with a narrative in the middle of the action, often at a critical point.
• Default: This method will create a structured idea description. 

To increase the quality of the response enter a few more information. Click the Create button to get the description.

Autofill tags

Go to any content detail page and find the smart action button below the content title or navigate straight to the tag section. Choose Autofill Tags and the system will return a suggestion of tags based on your provided Summary. This means that suggestions may include both existing tags from your platform and entirely new tags. If you select a suggested tag that already exists in your platform, it will automatically be merged with the existing tag, ensuring consistency and avoiding duplicates. Not fitting tags can be removed by pressing the "x" icon next to the respective tag. Insert your tag selection.

You can also add tags to an element via the chat. Specify which element or elements you have in mind and ask PRISM to generate tags for you.