AI agent to help remediate security related issues
AI agent to help remediate security related issues
Lead Designer
Lead Designer
Lead Designer
Hackathon project turned into initiative
Hackathon project turned into initiative
Hackathon project turned into initiative
Designed in 1 month
Designed in 1 month
Designed in 1 month
What is Security Agent?
What is Security Agent?
What is Security Agent?
Sarah is an IT administrator responsible for protecting her company’s sensitive data and minimizing security risks.
Sarah is an IT administrator responsible for protecting her company’s sensitive data and minimizing security risks.
Sarah is an IT administrator responsible for protecting her company’s sensitive data and minimizing security risks.
She activates Security Agent, a smart assistant that continuously monitors her company’s domain for potential issues each paired with AI-generated remediation suggestions
She activates Security Agent, a smart assistant that continuously monitors her company’s domain for potential issues each paired with AI-generated remediation suggestions
She sees a filtered dashboard of high-risk issues, like risky sharing behavior, inactive user accounts, unusual access patterns, or signs of ransomware.
She sees a filtered dashboard of high-risk issues, like risky sharing behavior, inactive user accounts, unusual access patterns, or signs of ransomware.
One issue highlights a public link exposing confidential files in a folder.
One issue highlights a public link exposing confidential files in a folder.
The Agent also identifies 44 more similar 'Public link' issues for other folder locations.
The Agent also identifies 44 more similar 'Public link' issues for other folder locations.
Sarah receives a recommendation to create a Content Safeguard Policy that fixes all 45 issues and blocks similar ones in the future.
Sarah receives a recommendation to create a Content Safeguard Policy that fixes all 45 issues and blocks similar ones in the future.
She reviews the policy, with actions mapped to her organization’s compliance standards and best practices.
She adjusts and approves it. The policy is now active and automated.
She reviews the policy, with actions mapped to her organization’s compliance standards and best practices.
She adjusts and approves it. The policy is now active and automated.
Instead of chasing every alert manually, Sarah now uses Security Agent to prioritize, fix, and bulk-remediate both current and future threats, saving hours and keeping her company secure.
Instead of chasing every alert manually, Sarah now uses Security Agent to prioritize, fix, and bulk-remediate both current and future threats, saving hours and keeping her company secure.
Instead of chasing every alert manually, Sarah now uses Security Agent to prioritize, fix, and bulk-remediate both current and future threats, saving hours and keeping her company secure.
How was it before?
How was it before?
How was it before?
How is it now?
How is it now?
How is it now?
What did I achieve?
What did I achieve?
What did I achieve?
I secured third place in the Egnyte Hackathon 2025 for the best use of AI, competing against 67 submissions. It was a group effort involving backend, frontend, and AI engineers.
I secured third place in the Egnyte Hackathon 2025 for the best use of AI, competing against 67 submissions. It was a group effort involving backend, frontend, and AI engineers.
I secured third place in the Egnyte Hackathon 2025 for the best use of AI, competing against 67 submissions. It was a group effort involving backend, frontend, and AI engineers.
My solution showcased how an AI-driven Security Agent can automate complex security workflows and deliver smart, actionable recommendations for stronger data protection.
My solution showcased how an AI-driven Security Agent can automate complex security workflows and deliver smart, actionable recommendations for stronger data protection.
My solution showcased how an AI-driven Security Agent can automate complex security workflows and deliver smart, actionable recommendations for stronger data protection.
How did I achieve this?
How did I achieve this?
How did I achieve this?
> Key design decisions taken
> Key design decisions taken
> Key design decisions taken
Prompt supports free-form natural language input to accommodate varied user mental models.
Design Process
1) Ethnographic study & defining goals
- We began by conducting ethnographic research to observe how users managed security issues and executed related workflows, helping us identify pain points in existing processes.
- These insights were further enriched through targeted interviews that revealed user expectations and concerns around adopting AI.
- Based on the findings, we defined clear business and user objectives, ensuring the product direction addressed both practical needs and user attitudes toward new technology and managing security risks related to their cloud data.
2) Enabling Open AI Model Development with Design Support
- Collaborated early with engineering as they began building the LLM-based AI agent.
- Provided initial UI designs and conversation flow mockups to give engineers a clear reference for user interactions, context, and expected outputs.
- This helped ground the model’s development in real user scenarios and ensured alignment between the AI’s capabilities and the intended user experience.
3) Iterative Prototyping & Feedback
- Ss the LLM agent evolved, created interactive prototypes and tested them with stakeholders.
- Collected feedback on usability, clarity, and trust, then refined both the interface and conversational flows to better match user expectations and business requirements.
4) Launch & Continuous Improvement
- Supported implementation and rollout by collaborating with engineering, product and growth teams.
- Monitored initial adoption, gathered user feedback, and identified opportunities for further enhancement to ensure the AI agent delivered measurable value and built user trust.
Design Process
2) Open AI Model development with design support
A flowchart mapping the integration of user interactions with LLM responses, clarifying how the AI agent would guide users through analysing and resolving security issues.
Developed actionable user personas based on ethnographic findings.
Jessica, a security analyst from Austin, Texas, represents our core user group, seeking a secure, efficient, and transparent way to report incidents.
Her persona guided key design decisions, ensuring the solution addressed real user needs and concerns about privacy and AI adoption.
1) Ethnographic study & defining goals
Conducted ethnographic interviews to observe and understand how users currently report and resolve security issues.
Sample questions explored users’ experiences with process bottlenecks, information needs, and their comfort with adopting new digital tools.
Insights from these sessions directly informed our product goals and design strategy.
AI prototyping tools (Like Vercel, Figma Make, Bolt, Stitch and others): Initial design artifacts provided to engineering, illustrating proposed user flows and key interaction points. These visuals anchored model development in real user scenarios, ensuring the AI agent’s capabilities matched user expectations.
3) Iterative prototyping & feedback
Proof of Concept was shared with stakeholders for usability testing and feedback. Iterative testing helped refine both conversational flows and interface elements to better meet user needs.
Their feedback led to improvements in AI explanations and escalation pathways, increasing user trust and task completion rates.
4) Launch & continuous improvement
High-fidelity screens of the launched product, showcasing the intuitive resolving issues interface and admin dashboard designed for clarity and efficiency.
'Feature Awareness' guide was created using 'Pendo' application to onboard users and drive feature adoption, increasing discoverability and reducing support requests.
The 'Feature Awareness' campaign led to a 49.37% increase in first-time user engagement within two weeks of launch, compared to the engagement at release.
in just 2 weeks
Design Process
2) Open AI Model development with design support
A flowchart mapping the integration of user interactions with LLM responses, clarifying how the AI agent would guide users through analysing and resolving security issues.
Developed actionable user personas based on ethnographic findings.
Jessica, a security analyst from Austin, Texas, represents our core user group, seeking a secure, efficient, and transparent way to report incidents.
Her persona guided key design decisions, ensuring the solution addressed real user needs and concerns about privacy and AI adoption.
1) Ethnographic study & defining goals
Conducted ethnographic interviews to observe and understand how users currently report and resolve security issues.
Sample questions explored users’ experiences with process bottlenecks, information needs, and their comfort with adopting new digital tools.
Insights from these sessions directly informed our product goals and design strategy.
AI Prototyping tools (Like Vercel, Figma Make, Bolt, Stitch and others): Initial design artifacts provided to engineering, illustrating proposed user flows and key interaction points. These visuals anchored model development in real user scenarios, ensuring the AI agent’s capabilities matched user expectations.
3) Iterative prototyping & feedback
Proof of Concept was shared with stakeholders for usability testing and feedback. Iterative testing helped refine both conversational flows and interface elements to better meet user needs.
Their feedback led to improvements in AI explanations and escalation pathways, increasing user trust and task completion rates.
4) Launch & continuous improvement
High-fidelity screens of the launched product, showcasing the intuitive resolving issues interface and admin dashboard designed for clarity and efficiency.
'Feature Awareness' guide was created using 'Pendo' application to onboard users and drive feature adoption, increasing discoverability and reducing support requests.
The 'Feature Awareness' campaign led to a 49.37% increase in first-time user engagement within two weeks of launch, compared to the engagement at release.
in just 2 weeks
Design Process
2) Open AI Model development with design support
A flowchart mapping the integration of user interactions with LLM responses, clarifying how the AI agent would guide users through analysing and resolving security issues.
Developed actionable user personas based on ethnographic findings.
Jessica, a security analyst from Austin, Texas, represents our core user group, seeking a secure, efficient, and transparent way to report incidents.
Her persona guided key design decisions, ensuring the solution addressed real user needs and concerns about privacy and AI adoption.
1) Ethnographic study & defining goals
Conducted ethnographic interviews to observe and understand how users currently report and resolve security issues.
Sample questions explored users’ experiences with process bottlenecks, information needs, and their comfort with adopting new digital tools.
Insights from these sessions directly informed our product goals and design strategy.
AI Prototyping tools (Like Vercel, Figma Make, Bolt, Stitch and others): Initial design artifacts provided to engineering, illustrating proposed user flows and key interaction points. These visuals anchored model development in real user scenarios, ensuring the AI agent’s capabilities matched user expectations.
3) Iterative prototyping & feedback
Proof of Concept was shared with stakeholders for usability testing and feedback. Iterative testing helped refine both conversational flows and interface elements to better meet user needs.
Their feedback led to improvements in AI explanations and escalation pathways, increasing user trust and task completion rates.
4) Launch & continuous improvement
High-fidelity screens of the launched product, showcasing the intuitive resolving issues interface and admin dashboard designed for clarity and efficiency.
'Feature Awareness' guide was created using 'Pendo' application to onboard users and drive feature adoption, increasing discoverability and reducing support requests.
The 'Feature Awareness' campaign led to a 49.37% increase in first-time user engagement within two weeks of launch, compared to the engagement at release.
in just 2 weeks
Design Process
"
What information do you wish you had at your fingertips when reporting or tracking an issue?
Painpoints
"
Can you walk me through the last time you resolved a security issue? What steps did you take?
Understanding Current Workflows
"
Can you recall a time when a new technology/app was introduced to your workflow? How did you feel about it
Attitudes Toward Technology and Change
"
What would make you feel confident that your security issue is being handled properly?
Expectations/Concerns
2) Open AI Model development with design support
A flowchart mapping the integration of user interactions with LLM responses, clarifying how the AI agent would guide users through analysing and resolving security issues.
Developed actionable user personas based on ethnographic findings.
Jessica, a security analyst from Austin, Texas, represents our core user group, seeking a secure, efficient, and transparent way to report incidents.
Her persona guided key design decisions, ensuring the solution addressed real user needs and concerns about privacy and AI adoption.
1) Ethnographic study & defining goals
Conducted ethnographic interviews to observe and understand how users currently report and resolve security issues.
Sample questions explored users’ experiences with process bottlenecks, information needs, and their comfort with adopting new digital tools.
Insights from these sessions directly informed our product goals and design strategy.
AI Prototyping tools (Like Vercel, Figma Make, Bolt, Stitch and others): Initial design artifacts provided to engineering, illustrating proposed user flows and key interaction points. These visuals anchored model development in real user scenarios, ensuring the AI agent’s capabilities matched user expectations.
3) Iterative prototyping & feedback
Proof of Concept was shared with stakeholders for usability testing and feedback. Iterative testing helped refine both conversational flows and interface elements to better meet user needs.
Their feedback led to improvements in AI explanations and escalation pathways, increasing user trust and task completion rates.
4) Launch & continuous improvement
High-fidelity screens of the launched product, showcasing the intuitive resolving issues interface and admin dashboard designed for clarity and efficiency.
'Feature Awareness' guide was created using 'Pendo' application to onboard users and drive feature adoption, increasing discoverability and reducing support requests.
in just 2 weeks
The 'Feature Awareness' campaign led to a 49.37% increase in first-time user engagement within two weeks of launch, compared to the engagement at release.
Designs
Designs
Designs
(Swipe using your mousepad to view more)
(Swipe using your mousepad to view more)
(Swipe using your mousepad to view more)
Receive the Security Agent's recommendations to fix cloud data risks on the Home page, within the Tasks section (To-do list). With a single click, usears can create a policy to resolve these issues and prevent future risks.
Navigate to a filtered list of issues flagged by the Security Agent and review them before enabling the policy.
Users can also view the Security Agent's recommendations on the Issues page, allowing them to take contextual actions.
Ask contextual questions about the task created by the Security Agent. The AI offers expert guidance to help you understand the recommendation.
Egnyte Copilot Dashboard: View available AI Agents, Knowledge Bases, and Smart Apps
Customize the Security Agent to fit your requirements using agent templates. You can modify the configuration at any time.
Create a security-specific prompt that will be displayed contextually when users are viewing security-related information.
Use the new Security Prompt while reviewing the issues list. AI also suggests questions related to the prompt.
Security Agent Dashboard is a task- and expert-based agent. It provides an overview of the impact of using AI recommendations to resolve cloud data risks and offers expert guidance on security-related topics
Deep Research Agent helps users dive deeper into security-related topics. It also provides summaries with actionable tasks that can be configured using the Security Agent or Prompts.
Check out the research summary generated by the Deep Research Agent.
Check out the activities performed by the Deep Research Agent and the sources it used to generate the research summary.
Receive the Security Agent's recommendations to fix cloud data risks on the Home page, within the Tasks section (To-do list). With a single click, usears can create a policy to resolve these issues and prevent future risks.
Navigate to a filtered list of issues flagged by the Security Agent and review them before enabling the policy.
Users can also view the Security Agent's recommendations on the Issues page, allowing them to take contextual actions.
Ask contextual questions about the task created by the Security Agent. The AI offers expert guidance to help you understand the recommendation.
Egnyte Copilot Dashboard: View available AI Agents, Knowledge Bases, and Smart Apps
Customize the Security Agent to fit your requirements using agent templates. You can modify the configuration at any time.
Create a security-specific prompt that will be displayed contextually when users are viewing security-related information.
Use the new Security Prompt while reviewing the issues list. AI also suggests questions related to the prompt.
Security Agent Dashboard is a task- and expert-based agent. It provides an overview of the impact of using AI recommendations to resolve cloud data risks and offers expert guidance on security-related topics
Deep Research Agent helps users dive deeper into security-related topics. It also provides summaries with actionable tasks that can be configured using the Security Agent or Prompts.
Check out the research summary generated by the Deep Research Agent.
Check out the activities performed by the Deep Research Agent and the sources it used to generate the research summary.
Receive the Security Agent's recommendations to fix cloud data risks on the Home page, within the Tasks section (To-do list). With a single click, usears can create a policy to resolve these issues and prevent future risks.
Navigate to a filtered list of issues flagged by the Security Agent and review them before enabling the policy.
Users can also view the Security Agent's recommendations on the Issues page, allowing them to take contextual actions.
Ask contextual questions about the task created by the Security Agent. The AI offers expert guidance to help you understand the recommendation.
Egnyte Copilot Dashboard: View available AI Agents, Knowledge Bases, and Smart Apps
Customize the Security Agent to fit your requirements using agent templates. You can modify the configuration at any time.
Create a security-specific prompt that will be displayed contextually when users are viewing security-related information.
Use the new Security Prompt while reviewing the issues list. AI also suggests questions related to the prompt.
Security Agent Dashboard is a task- and expert-based agent. It provides an overview of the impact of using AI recommendations to resolve cloud data risks and offers expert guidance on security-related topics
Deep Research Agent helps users dive deeper into security-related topics. It also provides summaries with actionable tasks that can be configured using the Security Agent or Prompts.
Check out the research summary generated by the Deep Research Agent.
Check out the activities performed by the Deep Research Agent and the sources it used to generate the research summary.
If you would like to explore the released product, please try Egnyte application Start free trial
If you would like to explore the released product, please try Egnyte application Start free trial
If you would like to explore the released product, please try Egnyte application Start free trial
© 2025 All Rights Reserved | Parmi Mehta
© 2025 All Rights Reserved | Parmi Mehta
© 2025 All Rights Reserved | Parmi Mehta