Using AI in Tines, you can take action and access proprietary data with Workbench, and with the AI agent action, bring advanced AI capabilities directly into your workflows.
Note: AI in Tines uses credits. You can see the real-time credits used per team in the AI settings section of the tenant Settings menu. Learn more about AI credits here.
Creating effective prompts for AI, or prompt engineering, can seem like a daunting task at the outset, but by focusing on writing clearly and concisely you’ll be able to leverage the most out of AI.
Following these 5 tips will help you to write better prompts, and get better results faster.
Write clearly and provide specific details
Provide context and use personas
Be specific about outputs and constraints
Clearly explain your goal and success criteria
Refine and adjust your prompts iteratively
1. Write clearly and provide specific details
Structure your prompt logically with clear instructions and include specific examples of what you're looking for. Avoid vague requests and instead provide concrete details about your expectations. When writing your prompt, use plain language to explain what the task is.
❌ Vague prompt:
Help me analyze this log file.
✅ Clear, structured prompt:
I need to analyze this Cisco firewall log for potential security incidents.
Please:
1. Identify any failed login attempts (more than 3 from the same IP)
2. Flag any connections to known malicious IPs (especially those from countries X, Y, Z)
3. Summarize the findings in a table with columns for IP, event type, and recommended action
4. Include a brief explanation of the security implications
2. Provide context and use personas
Give the AI relevant background information where necessary and consider asking it to adopt a specific role or perspective. When using personas in AI, it’s better to tell it that it is something, rather than to pretend that it is something. This will help to structure the response in a way that aligns with your needs.
❌ Limited context:
Write an email about the security incident.
✅ Rich context with persona:
You are an experienced CISO communicating with the executive team.
Write an email about yesterday's ransomware incident that:
1. Summarizes what happened (encrypted files in the finance department)
2. Explains current containment status (isolated to one subnet, backup restoration in progress)
3. Outlines next steps (security audit, employee training)
4. Maintains a calm, professional tone while conveying appropriate urgency
5. Keeps the email under 300 words for readability
3. Be specific about outputs and constraints
Clearly define what format you want the output in, any limitations to consider, and for complex tasks, ask the AI to show its reasoning process. Being specific with these details will help to reduce the amount of back and forth clarifications needed. However, you don’t want to make your request too busy or complex, be specific but succinct where you can.
❌ Ambiguous request:
Analyze this network traffic data.
✅ Specific request with output format:
Analyze the attached network traffic data for potential data exfiltration attempts.
Constraints:
- Focus only on outbound traffic to non-corporate domains
- Consider any single connection transferring >100MB as suspicious
- Flag connections occurring outside business hours (8am-6pm local time)
Please show your analysis step by step, and present your findings as:
1. A markdown table of suspicious connections with timestamp, destination, data volume
2. A brief explanation of why each flagged connection is concerning
3. Recommended next steps for investigation
4. Clearly explain your goal and success criteria
Help the AI understand what you're ultimately trying to accomplish and how you'll evaluate whether its response meets your needs. The more clear and direct you are with a goal the better. When you’re clear on this before writing your prompt, it’ll make understanding whether the response is right, or where it’s gone wrong much easier.
❌ Goal-less request:
Create a script to check our AWS environment.
✅ Goal-oriented request:
I need to create a Python script that audits our AWS environment for compliance with our security policy.
The goal is to identify and remediate any S3 buckets with public access enabled.
The script should:
1. List all S3 buckets in our account
2. Check each bucket's permission settings
3. Generate a report of non-compliant buckets
4. Include remediation commands to fix each issue
A successful script will correctly identify all public buckets in our test environment and provides accurate remediation steps that can be executed without manual modification.
5. Refine and adjust your prompts iteratively
Treat prompt writing as an iterative process. If you don't get the desired results, provide feedback and refine your prompt based on what worked and what didn't.
❌ Initial prompt:
Help me create a dashboard to monitor security alerts.
✅ Refined prompt after initial response:
Thanks for the overview. I'd like to create a more specific security alert dashboard in Tines.
I need to visualize:
1. Alert volume by severity (Critical, High, Medium, Low)
2. Top 10 alert sources
3. Alert trend over the past 7 days
Can you help me with the specific Tines formulas I'll need to aggregate this data from our alert JSON?
Learn more: