Custom Variables: Personalize Your AI Interviews at Scale

Every research project is different. The company you're investigating, the industry context, the specific angles you want to explore—these details shape how your expert conversations should unfold.

Until now, customizing AI interview agents meant editing prompts for each project. That approach works for a handful of interviews, but it doesn't scale. And it introduces risk: a hurried edit might break something that was working.

Custom Variables change this equation entirely.

The Problem with One-Size-Fits-All Prompts

Consider a typical due diligence workflow. Your team might conduct expert interviews across:

• Multiple target companies: Each with different products, competitors, and market dynamics
• Various industries: Healthcare behaves differently than SaaS, which differs from manufacturing
• Different research angles: Customer sentiment, competitive positioning, operational efficiency, regulatory landscape

A single AI agent prompt can't effectively handle all these variations. You end up with generic questions that miss project-specific nuances, or you constantly edit prompts—introducing delays and potential errors.

How Custom Variables Work

Custom Variables let you create placeholders in your agent prompts that get filled in differently for each interview. Think of them as fields in a template.

In Your Prompt

You are conducting due diligence research on {{company_name}}.

COMPANY CONTEXT:
{{company_background}}

RESEARCH FOCUS:
Explore {{research_focus}} with particular attention to:
- Market position relative to {{competitor_list}}
- {{industry_context}}

QUESTIONS:
{{questions_list}}

For Each Interview

When you create an interview, you fill in the variables with project-specific details:

Interview 1: Enterprise SaaS Due Diligence

company_name: Acme Corp
company_background: B2B SaaS, $50M ARR, Series C
research_focus: Customer retention and NPS
competitor_list: Salesforce, HubSpot
industry_context: Enterprise software buying cycles

Interview 2: Fintech Startup Evaluation

company_name: TechStart Inc
company_background: Early-stage fintech, pre-revenue
research_focus: Product-market fit validation
competitor_list: Stripe, Square
industry_context: Regulatory requirements for payment processing

Same agent. Same underlying prompt structure. Completely different interview contexts.

Real-World Use Cases

Multi-Target Due Diligence

Private equity firms evaluating multiple acquisition targets can use the same agent framework across deals:

Required Variables:

• target_company - Company being evaluated
• deal_thesis - Investment hypothesis to validate

Optional Variables (with defaults):

• valuation_context - Defaults to "growth-stage private company"
• key_concerns - Defaults to "market position, unit economics, management quality"

Each deal team fills in target-specific details while the core interview methodology stays consistent.

Sector-Specific Research

Hedge funds covering multiple sectors can create sector-aware agents:

For Healthcare:

industry_context: "FDA approval timelines, reimbursement dynamics, clinical trial phases"
regulatory_focus: "Medicare/Medicaid implications, HIPAA considerations"

For Technology:

industry_context: "Cloud migration trends, enterprise adoption cycles, developer ecosystem"
regulatory_focus: "Data privacy regulations, antitrust considerations"

The AI adapts its questions and follow-ups based on sector-specific context.

Client-Specific Engagement

Consulting firms serving multiple clients can customize while maintaining methodology:

Required:

• client_name - For appropriate confidentiality framing
• engagement_scope - What the client is trying to learn

Optional:

• client_industry - Helps frame industry-relevant follow-ups
• prior_findings - Context from earlier research phases

Geographic Expansion Research

Companies exploring new markets can tailor interviews by region:

target_market: "Southeast Asia"
local_context: "Mobile-first consumers, fragmented distribution, diverse regulatory environments across countries"
competitive_landscape: "Local champions like Grab and Gojek, plus global players adapting to regional preferences"

Required vs Optional: Getting the Balance Right

Custom Variables can be marked as required or optional, which determines the interview creation workflow.

Required Variables

Mark a variable as required when the AI genuinely can't conduct an effective interview without it.

Good candidates for required:

• Target company name (essential for framing)
• Core research question (defines the interview purpose)
• Compliance context (necessary for legal reasons)

Signs you've over-required:

• Interviewers leave fields blank or enter placeholder text
• The AI performs fine even with generic values
• You're requiring information that's "nice to have" rather than essential

Optional Variables with Defaults

Optional variables with sensible defaults give flexibility without friction.

Example:

Variable: interview_style
Default: "conversational and exploratory"
Override when needed: "direct and structured" or "technical deep-dive"

The AI uses the default unless someone has a reason to change it. Most interviews proceed without extra input; special cases get customization.

The Right Mix

A well-designed agent might have:

• 2-3 required variables: The essential context every interview needs
• 3-5 optional variables: Useful customizations with reasonable defaults

More than that suggests the agent is trying to do too much. Consider splitting into multiple specialized agents instead.

Maintaining Consistency Across Your Team

One challenge with customization is ensuring consistency. Custom Variables help here too.

Standardized Vocabulary

By defining variables at the account level, you establish shared terminology:

• Everyone uses target_company (not company, firm, target, or organization)
• research_focus means the same thing across all interviewers
• New team members learn the standard approach through the variable definitions

Enforced Completeness

Required variables ensure critical information isn't skipped:

• No interview goes out without proper company context
• Compliance disclaimers are always included
• Research objectives are always documented

Audit Trail

Variable values are stored with each interview, creating a record of:

• What context the AI was given
• How that interview differed from others
• Why certain approaches were taken

Best Practices

After working with teams implementing Custom Variables, several patterns emerge.

Start Simple

Begin with one or two required variables for your most common use case. Add complexity only when you have clear evidence it's needed.

Week 1: company_name (required) Month 1: Add research_focus (required) and industry_context (optional with default) Quarter 1: Evaluate whether additional variables would improve outcomes

Write Good Descriptions

Each variable should have a clear description explaining:

• What information to provide
• How the AI will use it
• Examples of good inputs

Poor description: "Company background"

Better description: "2-3 sentences about the company: what they do, approximate size/stage, and any relevant recent news. Example: 'B2B SaaS company providing HR software, ~200 employees, recently raised Series B, expanding into enterprise market.'"

Test with Edge Cases

Before rolling out a new variable:

1. Run test interviews with typical values
2. Run test interviews with unusual values
3. Run test interviews with minimal values (for optional variables)
4. Verify the AI handles all cases appropriately

Review and Refine

Periodically review how variables are being used:

• Are required variables always meaningful, or do people enter boilerplate?
• Are optional variables actually getting customized?
• Do defaults still make sense given how interviews are going?

Getting Started

Custom Variables are available now in InsightAgent. To begin:

1. Navigate to Settings → Variables to define your account's variables
2. Link variables to agents by referencing them in your agent prompts using {{variable_name}} syntax
3. Create interviews and fill in variable values as part of the interview setup

For detailed instructions, see our Custom Variables documentation.

What's Next

Custom Variables are the foundation for increasingly sophisticated interview customization. Coming soon:

• Variable templates: Save and reuse common variable combinations
• Conditional logic: Show different questions based on variable values
• Variable analytics: See which variable configurations produce the best outcomes

We're building toward a future where every expert conversation is precisely tailored to its research context—without requiring manual prompt engineering for each interview.

---

Ready to personalize your AI interviews? Get started with InsightAgent or schedule a demo to see Custom Variables in action.