Effective Prompting Guide

​

Why Prompt Quality Matters

• Understand your specific situation and constraints
• Draw on relevant AI industry expertise
• Provide more actionable and targeted guidance
• Save time by reducing follow-up questions
• Generate more useful documents and recommendations

​

The Anatomy of an Effective Prompt

​

1. Clear Objective

• “I need to evaluate vendors for a generative AI project”
• “I’m creating a requirements document for our data science team”
• “I need to prepare for a meeting with an AI vendor tomorrow”

I need to develop evaluation criteria to 
compare three enterprise LLM vendors (OpenAI, 
Anthropic, and Google) for our customer service use case.

​

2. Specific Instructions

• Create a specific document type
• Compare multiple options
• Analyze a particular vendor’s offering
• Suggest questions for a specific scenario

Please create a comparison matrix with rows
for each vendor and columns for: model capabilities, 
data privacy approach, enterprise features, 
pricing model, and implementation complexity.

​

3. Contextual Background

• Organization context: Industry, size, regulatory requirements
• Technical environment: Existing systems, integration requirements
• Team context: Stakeholder roles, expertise levels
• Procurement stage: Where you are in the buying process
• Timeline and constraints: Budget, deadlines, or other limitations

We're a financial services company with 
2,000 employees handling sensitive customer data. 
We currently use Zendesk for customer support 
and want to augment it with an AI solution. 
Our team has limited AI expertise, and we need 
to present a recommendation to our executive team 
in three weeks.

​

4. Examples (When Applicable)

• Similar documents you’ve created before
• Formats you prefer
• Style or tone preferences
• What “good” looks like to you

We previously created a vendor evaluation 
for our CRM system that was well-received. 
It used a 1-5 scoring system with clear criteria 
definitions and included both quantitative 
metrics and qualitative assessments.

​

5. Exception Handling

• Risk factors specific to your industry
• Internal requirements or policies
• Particular concerns from stakeholders

Data residency is a critical concern for us as 
we operate in the EU and must comply with GDPR. 
Also, our security team has expressed concerns 
about prompt injection attacks.

​

6. Tone and Interaction Preferences

• Do you want a direct answer or a guided exploration?
• Do you prefer technical depth or executive summaries?
• Are you looking for coaching through a process?

I'd like you to guide me through this evaluation
process step by step, as I'll need to explain 
the rationale to stakeholders with varying levels 
of technical understanding.

​

Putting It All Together

Objective: I need to develop a proof of concept (POC) 
plan for evaluating LLM integration into our customer 
support system.

Instructions: Please help me create a structured POC 
plan document that outlines the testing approach, 
success criteria, and timeline.

Context: I'm the IT Director at a healthcare company 
with 1,500 employees. We're considering implementing 
an AI solution to help our support agents respond to 
common patient questions more efficiently. We currently 
use ServiceNow for our ticketing system and have a team 
of 20 support agents. We're at the validation stage of 
our procurement process and have shortlisted Microsoft 
Azure OpenAI Service and Anthropic Claude. 
Our budget is approximately $50,000 for the POC phase.

Example: We previously ran a successful POC for our 
patient portal that included 4 weeks of testing with 
clear success metrics (20% reduction in response time, 
90% accuracy rate).

Exceptions: We need to consider HIPAA compliance 
requirements throughout the testing process, and 
we cannot use actual patient data without additional 
security measures.

Tone preference: I would appreciate a collaborative 
approach where you guide me through the key decisions 
for the POC design, explaining the rationale for 
different testing approaches and helping me anticipate challenges.

​

Refining Your Prompts

1. Clarify your objective: “That’s helpful, but I’m specifically looking for…”
2. Add more context: “Let me provide some additional background…”
3. Request adjustments: “Can you format this differently to focus more on…”
4. Ask for alternatives: “What other approaches should we consider?”

​

Prompt Templates for Common AI Procurement Scenarios

​

Problem Identification Stage

I need help defining the business problem we're trying to solve with AI technology. 
We're a [industry] company considering [general AI use case]. Our main challenges 
include [specific pain points]. Can you help me create a structured problem statement 
that will help align our stakeholders and quantify the potential impact?

​

Requirements Gathering Stage

I'm working on requirements for an AI implementation for [specific purpose]. 
Our organization is [brief context]. The key stakeholders include [roles]. 
Can you help me develop a comprehensive requirements document that covers 
technical capabilities, data needs, integration points, and governance considerations?

​

Supplier Selection Stage

We're evaluating [vendor names] for our [specific AI use case]. Based on our 
requirements, which include [key requirements], could you help me develop a scoring 
framework and suggest specific questions to ask each vendor during demos to uncover 
their true capabilities and limitations?

​

Validation Stage

We're planning a proof of concept for [AI solution] before finalizing our purchase. 
Our use case is [specific use case]. What should we include in our POC plan to ensure 
we're thoroughly testing the solution's capabilities, limitations, and fit for our 
environment? Please help me design specific test scenarios and success criteria.

​

Best Practices

• Start with guided tasks when available, as they’re designed to walk you through specific procurement activities
• Be specific about what you’re trying to accomplish
• Include relevant details but avoid overwhelming with unnecessary information
• Use project-based chats rather than standalone conversations for ongoing procurement activities
• Upload relevant documents to your project to provide additional context
• Follow up with clarifying questions if Kowalah’s response isn’t quite what you needed

​

Common Prompting Mistakes to Avoid

• Being too vague: “Help me with AI procurement” provides little guidance
• Omitting crucial context: Not mentioning industry-specific requirements or constraints
• Asking complex questions without breaking them down into manageable parts
• Expecting mind-reading: Not clearly stating your objective or expectations
• Overlooking your audience: Not specifying who the output is intended for

​

Next Steps

• Explore example AI procurement questions
• Learn about the six procurement stages
• Start your first project