Best practices for using AI to write great email outreach

We’ve recently published our latest guide, AI Sourcing Simplified. We’re sharing a few highlights in this article, but if you’d like to learn more, you can download the entire ebook here.

You probably don’t need convincing that AI can be a useful addition to your recruiting toolkit. But successfully adopting AI tools to supercharge your sourcing isn’t as easy as it might seem at first glance. It takes time and effort to learn how to prompt your AI tools to make the most of them.

In this article, we’ll share some of the best practices for writing effective prompts for AI sourcing. We'll focus on how being explicit, detailed, and intentional can drastically improve the quality of your results. You’ll learn how to refine and iterate your prompts for better sourcing outcomes. Let’s dive in!

The Amelia Bedelia Principle: Be Explicit and Literal

One of the most critical aspects of crafting effective AI sourcing prompts is understanding how AI interprets the information you give it. Unlike us humans, AI doesn’t read between the lines or infer meaning from context; it processes prompts exactly as they come. 

This phenomenon is best illustrated through the Amelia Bedelia Principle, named after the children’s book character known for her literal interpretation of instructions. When she was told to “draw the drapes,” Amelia didn’t close the curtains. Instead, she drew a picture of them.  (Source: Happy Annebeth)

Similarly, giving an AI-powered recruiting tool vague instructions like “find backend developers” may produce a wide range of profiles that don’t match your specific requirements. This is because the AI model relies solely on the words you provide and can’t make additional judgment calls outside the scope of what you’ve provided.

To improve your AI sourcing outcomes, it’s essential to write prompts that are explicit and detailed. Detailed prompts help the AI narrow down its search to profiles that precisely match your requirement criteria, saving you time and improving the quality of candidates it surfaces.

Good Prompts vs. Bad Prompts: The Importance of Detail

Crafting effective prompts for AI sourcing requires attention to detail and precision. When prompts are vague or general, AI tools struggle to identify the right candidates, often returning irrelevant or suboptimal profiles. In contrast, well-crafted prompts with specific, measurable criteria allow AI tools to function at their full potential, delivering higher-quality results.

Characteristics of Bad Prompts

Bad prompts are typically too vague, overly general, or filled with subjective descriptions. These prompts fail to provide clear instructions that AI can act on. Here are a few examples of bad prompts:

• Shows strong leadership skills

• Has development experience

• Is a high achiever

What’s wrong with these examples? They lack concrete, objective criteria. General terms like “leadership skills” and “high achiever” are subjective and difficult for AI to quantify based on resumes or profiles. As a result, the AI may surface candidates who loosely fit these descriptions, leading to irrelevant profiles and wasted time in the screening process.

Characteristics of Good Prompts

Good prompts, on the other hand, are specific, measurable, and concrete. They provide clear guidelines that AI can easily follow, ensuring more accurate results. Take a look at some examples:

For “leadership skills”

• Led a team of at least 5 people in delivering a cross-functional project.

• Managed a department budget exceeding $100,000 annually.

• Oversaw the onboarding and development of new team members, increasing retention by at least 15%.

• Coordinated multi-departmental efforts for project implementation, achieving results ahead of schedule.

For “backend development experience”

• Proficient in Java and Golang, with experience building scalable backend systems.

• Worked with NoSQL databases such as MongoDB and Redis to optimize data management.

• Developed APIs for integrating third-party services into mobile apps.

• Implemented security protocols in cloud-hosted environments using tools like AWS Lambda and Azure Functions.

For “high achiever”

• Surpassed sales targets for at least 8 consecutive quarters.

• Closed deals with an average value exceeding $250,000 annually. 

• Ranked in the top 5% of sales team for total revenue generated over the last 3 years.

• Recognized as Salesperson of the Year twice within a 4-year period.

Best Practices for Framing Qualifications and Examples of Effective Qualifications

To craft qualifications that deliver high-quality results, follow these steps:

1. Be measurable: Whenever possible, include quantifiable details. For example, rather than “experienced in sales,” use “3+ years of experience in SaaS sales with a consistent track record of exceeding quotas by at least 20%.”

2. Focus on specific skills: Avoid general qualifications like “good communication skills.” Instead, opt for specifics, such as “proven ability to deliver effective client presentations and close deals with Fortune 500 companies.”

3. Clarify work history: Ensure the AI knows exactly what you’re looking for in a candidate’s work history. Rather than saying “has worked in the finance industry,” frame it as “5+ years in financial analysis at a multinational corporation.”

4. Frame interests clearly: To capture candidates with relevant interests, go beyond vague statements like “passionate about technology.” Instead, ask for specific indicators such as “published research in machine learning” or “active contributor to open-source projects.”

Now let’s look at a few more examples related to criteria that often come up in sourcing searches: skills, work history, and interests:

1. Skills:

   1. Proficiency in data analysis using Python, R, and SQL, with experience in creating predictive models.

   2. Strong knowledge of cloud platforms like AWS or MS Azure, with experience managing cloud infrastructure for enterprise applications.

2. Work History:

   1. 3+ years of experience as a software engineer in a startup environment, with demonstrated success in scaling backend systems.

   2. Previous experience leading cross-functional teams in product development, delivering at least two major projects on time and within budget.

3. Interests:

   1. Has published technical articles on cybersecurity or participated in speaking engagements at major tech conferences.

   2. Active contributor to open-source communities, with a focus on developing tools for blockchain technology.

These examples work well because they are highly specific and tied to measurable outcomes or concrete achievements. When you outline your desired qualifications in specific terms like these, AI tools can more effectively scan resumes and are more likely to return relevant results.

Leveraging AI Tokens for Personalization

One of the most powerful ways to enhance your AI-powered sourcing strategy is by using tokens to personalize your outreach efforts. Tokens like #{{company}} and #{{title}} allow you to dynamically insert candidate-specific details into messages, making them feel tailored and relevant without manually personalizing each message.

For instance, by using the #{{company}} token, the AI will automatically pull in the name of the company where the candidate is currently employed, while the #{{title}} token inserts the candidate’s job title. This type of customization helps you build a strong connection with candidates, as they see a message that reflects their specific situation rather than receiving something that feels generic.

Gem’s #{{reason}} Token

In addition to these commonly used tokens, Gem has introduced the #{{reason}} token, a powerful tool for delivering scalable, targeted messaging. The #{{reason}} token allows you to personalize messages based on why a particular candidate was sourced. 

You can use this token to insert a specific reason why the candidate is an excellent fit for the role, such as skills they’ve demonstrated or experience they have with particular technologies. This not only saves time but also improves the candidate’s perception of your message, showing them that you’ve put thought into why they would be a good match.

Best Practices for Token Use

Here are some best practices to maximize the effectiveness of your tokens:

1. Start simple: Begin with basic tokens like #{{name}}, #{{company}}, and #{{title}}. These are the most frequently used tokens and can immediately add personalization to your outreach.

2. Leverage the #{{reason}} token: For a more targeted approach, experiment with the #{{reason}} token. Use this to provide context for your outreach, such as mentioning a shared industry or highlighting a candidate’s specific skills that make them a great fit.

3. Test and iterate: As with any AI-driven strategy, it’s essential to experiment with different tokens and adjust based on performance. Test how various combinations of tokens affect response rates and continuously refine your approach to optimize results.

4. Avoid over-personalization: While tokens add a personal touch, avoid using too many in a single message. Overloading a message with tokens can make it feel robotic and insincere. Stick to the most impactful ones that are relevant to the candidate’s experience and background.

Avoiding Common Pitfalls in AI Prompt Writing

As we’ve mentioned already, AI thrives on concrete, measurable data but struggles with ambiguity and interpretation. Here are a few common pitfalls to avoid when it comes to AI prompt writing.

Pitfall #1: Subjective Qualifications AI Can’t Access

Subjective qualifications, such as “strong leadership qualities” or “great team player,” are complex for AI to evaluate since these attributes are often not explicitly stated on resumes. AI sourcing tools can’t infer soft skills from a candidate’s job titles or achievements without specific contextual data. As a result, using subjective terms in your prompts often leads to mismatched profiles.

Instead, focus on qualifications that are measurable and tied to achievements. For example, rather than asking for “strong leadership skills,” request specific indicators of leadership, such as “managed a team of 10+ people for three years” or “led cross-functional project teams for product launches.”

Pitfall #2: Focusing Too Much on Soft Skills

Soft skills are essential but best evaluated in interviews rather than through AI-powered sourcing tools. Skills like communication, adaptability, and creativity are often showcased through in-person interactions, and they can be tricky to determine from a LinkedIn profile or resume.

To avoid overreliance on AI for evaluating soft skills, concentrate your AI sourcing prompts on hard skills, certifications, and measurable accomplishments. For instance, instead of using AI to find a candidate with “strong problem-solving skills,” request evidence of problem-solving by asking for qualifications like “resolved complex client issue, reducing churn by at least 15%.”

Pitfall #3: Being Too Vague

If your prompts are too vague, you will likely get a broad range of irrelevant results. Try to think in terms of specific skills, experiences, and achievements that make a candidate a good fit for a role. 

Let’s look at two examples:

Vague prompt: “Find experienced backend developers.”

Specific prompt: “Find backend developers with 5+ years of experience using Python and Django who have contributed to open-source projects and implemented RESTful APIs.”

In the vague example, we haven’t defined what we mean by “experienced” or even what we mean by the term “backend developers.” This means the AI sourcing tool is likely to come back with too many candidates that don’t match our actual criteria.

In the specific example, we’ve called out the exact number of years of experience as well as coding languages and types of projects that form our ideal candidate’s profile. 

Iterating and Refining Your Prompts for Continuous Improvement

One of the most powerful aspects of AI sourcing is its ability to evolve alongside your recruiting needs. You can refine and adjust the results generated from your prompts to improve the quality of candidates. Here are a few tactics to try out.

Use Results to Refine and Adjust Prompts

The feedback loop between your AI tool and the results it produces is vital for continuous improvement. After running a sourcing campaign, review the candidates the AI surfaced. Are they closely aligned with your needs? Are specific qualifications too broad or too narrow? These insights allow you to fine-tune your prompts for better outcomes. For example, if you consistently receive candidates with too much experience or lacking certain skills, adjust your criteria to be more specific or inclusive as needed.

Over time, minor refinements—like narrowing the required years of experience or adding specific technologies—will help the AI focus on exactly what you’re looking for, increasing the overall quality of matches.

Test Different Criteria and Tokens

A critical part of this refinement process is experimenting with different criteria and tokens to see how they influence the quality of candidates. For instance, testing variations of role-specific tokens, like #{{title}}, or using Gem’s new #{{reason}} token can significantly improve how the AI surfaces candidates. It’s often helpful to run A/B tests on different versions of your prompts to determine which token combinations yield the best results.

Adjusting the weight of certain qualifications can help your AI sourcing tool prioritize the most important skills or experiences. For example, if a particular certification is non-negotiable, ensure the prompt reflects that by emphasizing this qualification over others.

Commit to Ongoing Refinement

As your talent needs evolve, your prompts should, too. By continually refining and iterating on your prompts, you ensure that the AI adapts to changes in your hiring strategy, industry trends, and specific job requirements.

Ongoing prompt refinement also helps maintain a high level of accuracy in candidate matching over time. The more data you feed into the AI, the smarter and more efficient it becomes, ultimately delivering higher-quality candidates. This proactive approach to AI sourcing results in a more streamlined and effective hiring process, reducing time-to-hire and improving your long-term recruitment outcomes.

Elevating Your Sourcing Strategy with Gem’s AI Recruiting Platform

Gem’s AI-powered recruiting platform is designed to empower recruiters by offering tools that streamline the creation and refinement of prompts, ensuring more accurate talent sourcing. With its ability to process vast amounts of candidate data and interpret detailed, resume-specific qualifications, Gem enables you to focus on the right candidates while saving time.

Integrating Gem’s AI sourcing tools into your recruiting strategy can reduce the time you spend manually reviewing profiles while improving the overall quality of the candidates surfaced.

Whether you’re sourcing for highly specialized roles or managing large candidate pools, our AI adapts to your needs, delivering faster and more precise results.

If you’re curious to learn more, get in touch to see how Gem can help you maximize productivity, hire faster, and save money.