Guide

A new paradigm

With the advent of AI, we’ve entered a new paradigm of recruitment. Previously unsolvable problems are now solvable - among them, the heavy admin that comes with the job.

Both in-house and agency recruiters spend between 20% to 40% of their day on administrative tasks like taking notes, writing job descriptions, updating the ATS, creating write-ups, and so on. A tremendous amount of time that could be spend on more impactful things.

But, we hear you thinking - this is all just part of the job, right? For most recruiters, it is.

Up until now, there hasn’t been a way around this “admin problem.” It’s been seen as a necessary evil of the job.

‍Up until now.

Enter Large Language Models (LLMs) and generative AI. This new technology gives computers the ability to understand language in a way they never could before.

And we’re not just talking about understanding simple keywords—computers have always been capable of that, as seen in keyword-based boolean search, for example. No, in this new reality, computers can“understand” the semantic connections between words and the context of language itself and use it to create new output. And this is a truly revolutionary development for recruiting.

For the first time in recruitment history, it’s possible to work alongside AI, delegate tasks to an AI assistant, and in effect create a zero-admin recruitment agency.

In such an agency AI is embedded into the workflow of recruiters, handling the repetitive, administrative tasks, so that they can focus entirely on connecting with clients and placing people into jobs they love.

And this shift brings numerous benefits.

Zero-admin recruitment agencies experience increased recruiter productivity due to less time spent on admin, higher candidate and customer satisfaction, better recruiter retention, and ultimately increased revenue growth.

In short - it’s a new paradigm worth exploring.But how do you get there?

Let’s see what it takes to become a zero-admin recruitment agency - both in theory and with a real-life example of an agency already using AI in this manner.

Let’s dive in!

The admin problem explained

Before we dive into the inner workings of AI and the concept of “the zero-admin agency”, let’s take a moment to understand the admin problem itself. And to do that, we need to dissect the recruitment process.

If you strip it down to its core, the recruitment process essentially consists of a series of interactions of a recruiter with hiring managers and candidates, leading to a placement. The role of the recruiter is to use the information gathered from these interactions to match candidates to the right roles.

On an abstract level, we can imagine the recruitment machine as a system with a series of inputs and outputs:

• The process starts with an intake call with the hiring manager (input). The context of this call is turned into job requirements, and job descriptions (output).
• The job description (output) becomes input for the candidate search step.
• Once a candidate applies to the open vacancy, the information in their CV and the assessment result(input) is used by recruiters to shortlist or reject them (output).
• If a candidate is shortlisted and is invited to an interview, the context of the conversation becomes input for the next stages of the process.The candidate interview will be the input, while the candidate write-up will be the output document that helps recruiters “sell” the candidate to the client, and so on.

Of course, every process is a bit different, but from an information transfer perspective, it’s a series of inputs used to create a series of outputs, leading to a hiring decision.

When a human being fully handles this process, all the steps required for turning the input information into output - for example, taking notes during an intake call to create a job description, or writing a candidate presentation based on an interview - are done manually. In this system, the recruiter acts as a translator, turning input into output.

So next to the actual conversations that are part of the process, recruiters have to handle all these administrative tasks that are essential for transferring information about candidates to hiring managers, and vice versa.

Now, if we look at effort versus impact, it’s clear that the steps of the process that are the most time consuming are not necessarily the most impactful. In other words, recruiters spend a big part of their time on tasks that - although necessary - are not “moving the hiring needle.”

Both recruiters and agency owners acknowledge this problem - and admit that all this administrative work is a “necessary evil”, as without it, the recruitment machine wouldn’t function. After all, you need a human being to take in the information, process it, and contextualize it to create output.

A necessary evil?

But now a question arises: if so many acknowledge the problem, why haven’t we solved it yet?

The answer is simple, and has to do with the format of the data used in the hiring process.

Most of the data used in the recruitment flow is what we call “unstructured data” - it’s written or spoken words that don’t follow a particular format or structure.

The average interview, for example, is not a series of “yes” or “no” questions. On the contrary, it consists of more complex exchanges - open questions and answers that together make up the context needed by the recruiter to create output.

Taking this type of unstructured data, processing it, interpreting the context, and turning it into usable information requires human-like language understanding abilities. Until now, no technology has brought such capabilities to the market.

Yes, we’ve had rule-based automation for quite some time, along with boolean and keyword-based search, and we have been able to speedup certain stages of the hiring process with that. But that is different from actually understanding language.

As a result, even with pre-AI automation in place, the bulk of the manual work has remained, because of this simple problem: no technology before AI was able to mimic the human ability of processing, interpreting, and working with unstructured data like the one from interviews and intake calls.

Impact on recruiters, candidates, and customers

The impact of this ‘old way of doing things’ cannot be overstated. The admin problem is the root cause of many of the other issues faced by all stakeholders in the recruiting process, and is one of the main reasons why modern hiring is often called “broken”.

‍For recruiters, the admin problem is most significant. They want to connect with candidates, get to know enough about them to make sure they fit the open roles perfectly, and build strong relationships with clients.They want to keep everyone in the process engaged and informed, and make fast placements.

In practice, though, they only have about half their weekly time for such activities.

The other half is filled with admin, leading to a range of problems: Disengagement and a high risk of burnout, prolonged work hours but decreased effectiveness, and high employee turnover due to the increased workload and the repetitive nature of the job.

And this leads to a range of challenges downstream.

For candidates, the recruiting experience suffers when recruiters aren’t able to dedicate enough time to interviews, follow-ups, and in-between checks. Long response times, limited attention from the recruiter, and slow progression from one stage of the process to the other, all lead to candidate dissatisfaction.

For customers, working with recruiters who are spread too thin means that they don't always get the best candidates presented to them; after all, speed is key in hiring the best people. Furthermore, the relationship suffers and trust is eroded when recruiters can’t swiftly respond to requests, or can’t adapt to changing requirements.

By now it’s probably clear to see why solving the admin problem has long been a challenge for recruiters.

What’s exciting about all this, is that generativeAI is emerging as a game-changer, offering a much-needed breakthrough.

So let’s dive deeper into why AI is poised to solve this problem, and how it actually works in practice.

How AI can finally solve the admin problem

Artificial intelligence—particularly in the form of large language models (LLMs)—represents a fundamental shift in capability, and is causing us to rethink the way we search for, engage with, and hire people. These models, unlike earlier technologies, can understand human language, generate new content based on context, and display human-like creativity. LLMs are trained on massive datasets of text.

For example, GPT-4, which powers ChatGPT, is trained on hundreds of billions of words from various sources, such as books and websites, covering almost every topic imaginable. These words create a large, interconnected web of ideas and concepts that AI uses to learn the nuances of language, including idioms, tone, and implied meanings.

This deep understanding helps AI interpret ambiguous language, go beyond simple keyword matches, and grasp the intent behind questions or statements. The way these models are built allows for:

• ‍Associative thinking: AI can make connections between seemingly unrelated concepts, much like human creativity.
• Contextual weighting: The importance of different factors can be dynamically adjusted based on the overall context.
• Fuzzy matching: Instead of exact matches,AI can identify similarities and relationships across a spectrum.
• ‍Analogical reasoning: AI can apply knowledge from one domain to another, finding relevant patterns and solutions.

To visualize this, imagine a vast multidimensional space where every word, concept, and idea has its own coordinates. When processing information or generating responses, AI navigates this space, finding the most relevant and meaningful paths between different points.

For instance, if you ask "Can you suggest some good team-building activities?", the AI doesn't just look for the keyword "team-building." It understands the context and can suggest relevant activities like escape room challenges or collaborative problem-solving exercises, ensuring the suggestions are contextually relevant.

This fundamental difference in how AI operates translates to significant practical advantages in recruitment.

AI can understand the subtle differences between job requirements and candidate qualifications, going beyond simple keyword matching. It can draft personalized messages that take into account the candidate's background, the job specifics, and the overall market context.

On top of this, AI can guide interviewers with relevant questions based on the candidate's responses, adapting in real-time to the conversation flow. When evaluating candidates, AI can consider a wide range of factors simultaneously, weighing them dynamically based on their relevance to the specific role and company culture.

Furthermore, the amount of data an AI system can store and derive information from is virtually endless. In recruitment, an AI assistant can remember everything a candidate or client said, from the broader talking points to the least significant details. And it can do this with every candidate and client.

Finally, as AI systems process more data, they can identify trends and patterns that humans might miss, providing valuable insights into the recruitment process.

The importance of context

In essence, AI has the potential to transform the entire recruitment process and make it human again. It brings a level of understanding, adaptability, and insight that closely mimics human cognition, but at a scale and speed that far exceeds human capabilities.

However, there’s one catch: for gen AI to truly transform recruitment operations, it requires the right context. In other words, to be valuable in your recruiting process, AI needs to have access to company-specific information, in order to generate company-specific output.

For example, your company description can be used by AI to write job descriptions and to ensure candidates fit the company culture. When creating candidate write-ups, AI can borrow the tone of voice used by a recruiter in a previous presentation document, or in an intake call or interview to make sure its version mimics the original.

But it needs to have access to this context.

Without context, the output generated by artificial intelligence tools will be generic. And while this can work in some scenarios, most clients and recruiters prefer their AI to be customized to sound like them.

For this to happen, AI systems need to be integrated into your existing workflows and tech stack. This means including AI in the ‘input’ parts of the process and providing access to documentation where context can be derived from. For example meetings, notes, resumes, and so on.

And if you go one step further, you can give the AI access to your ATS, CRM, talent pool, and other sources of information, to supercharge your tech stack with AI capabilities. With such integrations, true zero-admin recruiting becomes possible.

The zero-admin agency

If you combine all these concepts, it’s possible to establish a zero-admin agency where the administrative tasks traditionally handled by recruiters are fully automated through contextually enriched artificial intelligence.

This shift enables recruiters to focus almost entirely on the high-value aspects of their roles, such as building relationships with candidates and clients, rather than being bogged down by time-consuming administrative duties.

How it works

In an AI-enabled zero-admin agency, recruiters and AI work alongside each other, each doing what they’re best at.

While recruiters are focusing on building relationships with clients and candidates to strengthen the employer brand, build pipeline, and grow revenue through placements, AI takes care of all the admin work.

AI assistants work tirelessly in the background,24/7, executing tasks that were assigned to them by their human counterparts, or initiating new tasks, based on predefined workflows.

For example, a recruiter might decide to delegate all the sourcing to AI, while another recruiter might prefer their AI assistant to only handle the interview- and intake-related admin- things like preparing interview questions, transferring call notes to the ATS, or writing follow-up emails.

The level of customization is up to each individual recruiter. But in principle, in a zero-admin agency, recruiters no longer have to do admin tasks.

Through integrations across the tech stack, AI assistants can be embedded in all the stages of the recruitment process, from sourcing and screening to interview scheduling and hiring decisions, as follows:

• In the sourcing phase, AI can scan various job boards, social media platforms, and databases to find potential candidates that match the job requirements, and provide recruiters with a long list of potential matches.
• During screening, AI can analyze resumes, assess qualifications, and rank candidates based on their fit for the role, shortening the list to only the top candidates.
• AI can also engage candidates through instant messages, qualifying them, answering their questions, and providing more information about the role.
• Once candidates reach the interview stage, AI can handle scheduling, prepare interview questions, and transfer call notes to the ATS. AI can coordinate interview times that work for both the candidates and the interviewers, and generate follow-up emails.
• ‍Throughout the hiring process, AI can assist with background checks and reference calls, ensuring all necessary information is gathered efficiently.
• Finally, AI assistants can support recruiters in preparing offer letters and onboarding documentation once the hiring decision has been made.

By integrating AI at each stage, recruiters can focus more on strategic tasks and relationship building, while AI handles the administrative workload.

How do these AI assistants work?

AI assistants explained

Each assistant - or AI agent - can work independently, specializing in a single type of task, or interact with other agents to ensure the quality and consistency of output.

The customization of each AI agent is done through prompt pipelines, which are structured sets of instructions meant to guide the AI model in producing specific types of responses.

For instance, if you want the AI to write job descriptions, your prompts might include details about the structure or the output, as well as the sections to be included: A short intro about your company, followed by the job requirements and key responsibilities.

The objective of the prompt pipeline should also be included. This could be generating content in a particular style or tone, or performing the task with a certain level of detail.

Each prompt may build on the previous one, refining the output progressively. For example, the first prompt might gather basic information about the company and role, the second prompt could focus on expanding that information, and the third prompt might refine the content for clarity and style.

The AI model uses these prompts to generate responses according to the instructions provided, and to ensure the output meets quality standards through a series of checks and balances.

Together, these custom AI agents form an"assembly line" where each agent has a specific role, as explained above: one handles sourcing, another agent takes care of interview admin, and another agent engages candidates in the talent pool.

By giving the AI systems access to the tools they work with on a daily basis, from calendars to ATS systems, recruiters ensure the AI has all the context needed for creating quality output and for taking over the admin tasks.

In this new reality, every recruiter ends up with their personal AI workmate that joins them where they work, takes in the context, and autonomously executes work the recruiter shouldn’t do.

Getting started: Becoming a zero-admin agency

Now that all the concepts are covered and clear, let’s move to the practical side of things: How to actually get started with incorporatingAI into your workflows.

At Carv, we believe that a successful implementation requires a holistic approach that takes into consideration not only the tech side of things, but also the human component.

This means that we encourage a phased approach, where we help recruiting teams achieve some quick wins while they’re still getting used to working alongside their new AI workmates.

But before the implementation work can start, a prerequisite is understanding the current state of things and the end goal. If the goal is to enable zero-admin recruitment, what’s the starting point?

Some agencies are still using old-school pen and paper, others have a mix of manual and digital processes, while others are almost fully automated.

As there’s no standard setup, we always start by mapping out the current recruitment process, to gain a clear understanding of both the candidate journey and the recruiter experience.

We map out not only the process steps, but also the tools used in each phase of the process, and the data that gets transferred from one system to the other.

This means that we look at all the input and output documents or data sources used in the process, to understand what will be the responsibility of AI, and what steps of the process should remain under human control.

Next, we quantify the potential effect of AI by identifying the bottlenecks and high-impact areas where AI can significantly improve the recruitment process.

Then, we choose a starting point for the solution design. For example, an agency could choose to have all their candidate profiles written by AI, based on the pre-screening calls or video interviews.If this is the step that takes most of recruiters’ time, delegating this task to AI will already make a difference.

In this scenario, the AI workflows will be customized to create specific output documents, tailored to the needs of the agency or end client.

The structure, tone of voice, format, and length of the candidate profile can be customized, and the data can be automatically pushed to the ATS or attached to an email, for following up with the hiring manager.

In another potential scenario, an agency could choose to delegate all the sourcing to AI, next to the admin tasks related to the core process. This would free up more time for recruiters, but would require a deeper setup too, with more custom workflows and personalized AI agents.

For example, an AI agent could automatically and continuously scan a talent pool or candidate platform, engage with candidates and pre-screen them via AI-led phone calls or instant messages, and shortlist the best ones for interviews with the recruiter.

Because the possibilities are endless, we generally recommend starting small, experiencing some first wins as the team is getting used to working in a new way, and scaling from there.

Our experience shows that starting with the core process and automating first the internal recruitment workflows - so anything related to intake calls and interviews - is a good approach, as it helps teams get up to speed in a very short time.

Once the “Zero-admin” status is achieved in the core process, agencies can choose to expand the AI implementation to their sourcing, candidate pool engagement, and pipeline building processes.

And it all starts with a first discovery conversation. Book one here.