Saraschandra KaranamHuman - AI Interactions Researcher · Enterprise AI
Back to Research
RESEARCH PROGRAM • 2025-26

Enterprise AI & Agentic Systems

Investigating how enterprise security professionals build trust, delegate decisions, and collaborate with increasingly autonomous AI systems during complex firewall management.

Amount of PainUser comfort with delegation to AIStrategic OpportunitiesQuick WinsWorkflow AWorkflow BWorkflow CWorkflow DWorkflow EWorkflow FWorkflow G
Representative Visualization

Task Comfort × Amount of Pain

An anonymized representation of how enterprise security professionals calibrated willingness to delegate work to AI across representative workflow categories.

23+Experts
2Connected Studies
MixedMethods
MultipleProduct Releases
02 · WHY THIS RESEARCH MATTERED

Why this research mattered

Enterprise AI is evolving from answering questions to taking actions. As autonomous systems become capable of reasoning, planning, and executing workflows, organizations face a fundamental challenge: determining which decisions professionals are willing to delegate, where human oversight remains essential, and how trust should be established in high-risk environments.

Alert TriggerPhase 01AI ReasoningPhase 02Human ReviewPhase 03ExecutionPhase 04FeedbackPhase 05Continuous Feedback Loop
Simplified Human ↔ AI Collaboration Loop
RESEARCH QUESTION

"How do security professionals decide what to delegate to increasingly autonomous AI?"

Rather than treating trust as a single metric, this research program investigated it from multiple perspectives—including mental models, workflow analysis, design validation, and adoption behavior—to build a comprehensive understanding of human–AI collaboration in high-risk enterprise environments.

Stream 01

Mental Models

How experts reason about trust, risk, and delegation.

Stream 02

Workflow Analysis

Understanding where automation creates value—or introduces risk.

Stream 03

Design Framework

Evaluating whether a unified agentic design framework supports real-world needs.

Stream 04

Adoption

Understanding how design influences willingness to collaborate with AI.

03 · EMPIRICAL DEVELOPMENT

Research Program

Connected studies conducted progressively answered different aspects of the central research question.

Study 01

Identifying Agentic Use Cases

2025
Question

"Where across firewall management do professionals actually want AI?"

Methods
SurveyInterviews
Outcome

Prioritized agentic opportunities across seven workflow stages.

Study 02

Validating the Agentic Design Framework

2026
Question

"How should AI interactions be designed to align with user expectations?"

Methods
Framework ValidationInterviews
Outcome

Validated framework, identified coverage gaps, refined design principles.

Insights from the first study directly informed the hypotheses and validation strategy of the second study, allowing the program to evolve from identifying opportunities to validating design guidance.

04 · REPRESENTATIVE CASE STUDY

Validating the Unified Agentic Design Framework

4 Months8 Enterprise Security ProfessionalsInterviews

Background

Several product teams building agentic workflows operated under conflicting assumptions regarding task automation comfort. The central design framework needed empirical evidence and rigorous validation to drive aligned product roadmaps.

Research Questions

  1. How do enterprise professionals interpret different touchpoints within agentic workflows?
  2. Which principles of the unified design framework align—or conflict—with user expectations?
  3. Does exposure to well-designed agentic workflows increase willingness to adopt AI assistance?

Research Approach

InterviewsConcept WalkthroughFramework Validation
Trust variation across touchpoints50%40%30%20%10%0%Trust %34%22%29%23%41%33%17%10%Touchpoint 1Touchpoint 2Touchpoint 3Touchpoint 4Touchpoint 5Touchpoint 6Touchpoint 7Touchpoint 8
Representative visualization of varying trust levels of security professionals across various agentic touchpoints

Key Findings

Due to NDA and intellectual property considerations, specific qualitative findings and metrics are generalized or omitted.

Research Contribution

This case study generated empirical evidence that strengthened the unified enterprise agentic design framework, identified previously unaddressed interaction patterns, and informed future product direction across multiple enterprise AI initiatives.

05 · RESEARCH ASSETS

Research Artifacts

Empirical assets and frameworks generated to guide future enterprise-wide design and engineering direction.

Journey Maps

A comprehensive step by step flow for each of the seven stages of firewall management

Agentic Principles Framework

List of gaps and improvements for each of the touchpoints in the framework

Trust Calibration Matrix

Empirical thresholds plotting amount of pain against security professional's delegation comfort.

Touchpoints vs Agentic Principles

A matrix aligning agentic principles and agentic touchpoints

Interview Framework

Semi-structured protocol used to query expert mental models of automation.

Survey Instrument

Attitude scaling questionnaire deployed to validate quantitative comfort shifts.

06 · PROGRAM PORTFOLIO

Additional Studies in this Research Program

This representative deep dive is one part of a much larger program of research.

Study 01

Firewall Agentic Use Cases

Methods: Survey (64), Interviews (15)
Outcome: Prioritized automation opportunities across seven stages of firewall management.
07 · PROGRAM OUTCOMES

Research Impact

The structural, organizational, and methodological contributions generated by this program.

Product Contributions

  • Validated unified agentic framework
  • Prioritized enterprise roadmap
  • Identified high-value automation opportunities

Organizational Alignment

  • Shared across multiple product teams
  • Created common vocabulary
  • Aligned design decisions

Methodological Value

  • Established reusable evaluation methodology
  • Created trust metrics
  • Provided reusable research framework