AI-Powered Fields

AI-Powered Fields enable automatic AI-based generation of insights, classifications, summaries, evaluations, and other supporting content throughout the innovation lifecycle. These capabilities help organizations reduce manual effort, improve consistency, and accelerate idea analysis and decision-making.

The AI can generate outputs automatically during workflow transitions, on submission, or manually through regeneration actions, depending on the field configuration. Supported AI-generated field types include:

  • Text

  • Number

  • Selection fields (Select, Radio, Checkbox)

The AI can also use additional idea fields, campaign context, and optional web search capabilities to improve result quality.

Common Use Cases

Executive Summaries

Automatically generate concise summaries for long or complex submissions.

Recommended when:

  • Idea descriptions are lengthy

  • Campaigns are broad or cross-functional

  • Large volumes of submissions require rapid triage


Example:
Generate a short executive summary highlighting the core problem, proposed solution, expected impact, and implementation complexity.

Idea Value Assessment

Use AI to evaluate the potential value of an idea before it progresses through the workflow.

The output can be:

  • Textual assessment

  • Numeric score

  • Automatic selection from predefined categories


Example:
Estimate the expected business value and strategic alignment of the idea.

Risk, Feasibility, and Impact Assessments

Generate structured evaluations to support review committees and decision makers.

Examples:

  • Risk analysis

  • SWOT summaries

  • Feasibility scoring

  • Strategic alignment classification

  • Expected cost vs. benefit evaluation

These insights can be triggered at different workflow stages. For example:

  • Executive Summary on idea submission

  • SWOT Analysis during advanced evaluation

  • Strategic Fit assessment before approval

How to Configure an AI-Powered Field

  1. Navigate to the relevant Workflow State settings.

  2. Open the "Additional Info Fields" page.

  3. Add a new field.

  4. Select one of the supported field types:

    • Text

    • Number

    • Select / Radio / Checkbox

  5. Set the field source to AI Generated

  6. Configure the AI settings and task instructions.

Supported field references and prompt enrichment capabilities are described in the AI-Powered Fields specifications.

Understanding How AI Prompts Work

The AI request is assembled into a structured prompt that includes:

  • General company and subsystem context

  • Campaign information (optional)

  • Idea title and description

  • The configured AI task instructions

  • Additional reference fields (optional)

  • Output formatting instructions

  • Language and validation rules

The system automatically enriches the prompt with relevant context, so there is usually no need to manually repeat information already available in the idea or campaign.

How Field Data Is Sent to the AI

When additional fields are selected as reference fields, their data is sent as structured text within the prompt.

Each referenced field includes:

  • Field title (label)

  • Field value

  • Field machine name / unique identifier

Field types are normalized before being sent:

Field TypeSent Format
TextPlain text
NumberIncludes numeric format and units/prefix/suffix
SelectionIncludes available values and internal term IDs

This allows the AI to understand both the semantic meaning of the field and the technical structure required for deterministic outputs.

Referencing Fields Inside Prompts

Reference fields are additional context which can be relevant for the AI to complete its task. 
You can and should reference fields by their titles directly in your instructions.

Example:

Analyze the "Scope" and "Timeline" fields to assess implementation feasibility.

This works because:

  • Field titles are explicitly included in the prompt

  • The AI receives labeled structured inputs

  • Internal machine names are included for deterministic mapping



Empty Fields Behavior

If a referenced field has no value:

  • The field is still included in the prompt with an empty value

  • The AI is expected not to invent missing information

  • If insufficient data exists, the field may remain empty

  • A data-related error message may be displayed

This behavior helps prevent hallucinations and unsupported conclusions.

Web Search Option

Administrators can optionally allow AI-generated fields to perform real-time web searches before generating responses.

This is useful for:

  • Market analysis

  • Competitive intelligence

  • Trend discovery

  • External validation

If enabled:

  • The AI may retrieve up-to-date external information

  • Citations are logged internally

  • Output formatting rules are still enforced

If you do not see the Web Search option, contact your Customer Success Manager.

Prompt Writing Guide

Well-structured prompts produce significantly more accurate and consistent outputs.

General Principles

Be Explicit

Clearly define:

  • What the AI should analyze

  • Which fields should be used

  • What type of result is expected

Avoid:

Analyze the idea

Prefer:

Analyze the "Problem Statement", "Timeline", and "Expected Savings" fields to assess feasibility and business impact.

Keep Each Field Focused

Each AI field should perform one specific task.

Good examples:

  • Generate an executive summary

  • Estimate implementation complexity

  • Classify strategic alignment

  • Identify key risks

Avoid combining multiple unrelated tasks into a single field.

Define Output Expectations

Specify:

  • Desired format

  • Length constraints

  • Allowed values

  • Fallback behavior

Example:

Return only one category from the provided list.
If insufficient information exists, return "None".

Prevent Unsupported Assumptions

Instruct the AI to rely only on provided information.

Recommended rule:

Do not assume missing information.

Recommended Prompt Structure

A reliable prompt structure is:

  1. Role (optional)

  2. Fields to analyze

  3. Task instruction

  4. Rules and constraints

  5. Output format

Example:

Act as a product evaluation analyst. 
Using the following fields: - Scope - Timeline - Expected Savings 
Assess the implementation feasibility of this idea. 
Rules: - Use only the provided field data - Do not assume missing information - If insufficient data exists, return "Insufficient Information" - Return a concise professional summary under 150 words

Best Practices

Use Reference Fields Strategically

Include only fields that materially improve the task quality.

Too many unrelated fields may:

  • Increase token usage

  • Reduce output precision

  • Introduce conflicting context

Use Field Titles Naturally

Field labels act as semantic anchors for the AI.

Example:

Compare the "Current Process" field against the "Proposed Solution" field.

Keep Instructions Deterministic for Selection Fields

For AI-generated selection fields:

  • Explicitly state whether multiple values are allowed

  • Define fallback behavior

  • Request exact matching values only

Example:

Select exactly one category from the provided terms.
If no clear match exists, return an empty response.

Selection field behavior and term handling are defined in the AI selection field specifications.

Define Numeric Expectations for Number Fields

For number outputs:

  • Clarify whether the value should be integer or decimal

  • Define the scoring scale or units

Example:

Return an integer score between 1 and 10 representing implementation complexity.

Avoid Overly Broad Instructions

Broad prompts tend to generate inconsistent results.

Avoid:

Analyze this idea completely.

Prefer:

Evaluate the expected operational impact and summarize the primary benefit in two sentences.

Monitoring and Troubleshooting

All AI requests and responses are logged internally.

Administrators can review:

  • Full prompts

  • Responses

  • Tokens usage

  • Web search activity

  • Errors and finish reasons

Navigate to:

Admin → System Logs → AI Audit Logs

This is useful for:

  • Prompt optimization

  • Troubleshooting failed generations

  • Reviewing AI behavior

  • Improving consistency

AI logging and audit capabilities are described in the platform specifications.

Important Notes

  • AI-generated results may occasionally contain inaccuracies.

  • Prompt quality directly impacts output quality.

  • AI-generated fields remain editable by authorized users.

  • Empty or insufficient input may result in no generated output.

  • Generated fields display an AI indicator on the idea page.

Continue refining prompts over time based on observed results and audit log feedback.