Getty Images/iStockphoto

Case study: GenAI saves time for Landing Point's recruiters

Executive search firm Landing Point uses GenAI to format resumes and job postings. The project saves every recruiter four hours per week -- all while ensuring data security.

Generative AI is reshaping IT strategies across industries, including recruitment.

At Landing Point, a New York-based executive search and recruiting firm, lead AI engineer Faizel Khan implemented generative AI (GenAI) to cut operational inefficiencies while prioritizing security. The firm's IT strategy had a clear objective: reduce administrative friction so recruiters could focus on building relationships. By automating repetitive tasks, such as resume formatting, candidate bio creation and contract drafting, the IT team freed recruiters to reclaim valuable time. The strategy emphasized speed in deployment and strong governance to protect sensitive client and candidate data.

To support this, the team used AWS private cloud infrastructure for secure data storage and fine-tuned large language models (LLMs) for specific business needs. They integrated AI tools directly into the firm's existing applicant tracking system (ATS), enabling seamless adoption without disrupting workflows. The initiative has delivered measurable gains: four hours saved per recruiter each week, a tenfold increase in job postings and faster time-to-placement.

In this interview, Khan discussed the architecture, challenges and lessons that shaped the GenAI project.

Editor's note: The following interview was edited for length and clarity.

What pain points pushed the company to adopt GenAI?

Faizel Khan: Our main motivation has been administrative friction. We believe that we should spend more time engaging with clients and candidates to get a better match between them. Our recruiters must do a lot in the background -- they must format resumes, create bios and create several documents, like contracts. That used to take a lot of time. We decided to tackle this using AI, because with GenAI, we could have natural language generation build these documents that our recruiters were previously doing.

When building out Landing Point's GenAI architecture, what considerations did you prioritize -- speed, scalability, governance?

Khan: First was speed and governance. We didn't have to tackle scalability because we're a small firm. We have less than 100 employees. So, we first wanted to have speed. How quickly can we get these resumes formatted? How quickly can we get the job description out on our website? At the same time, we wanted to maintain privacy regarding our client and candidate data.

What was the biggest challenge during the project?

Khan: The biggest challenge during the development was the data. Before I joined in, there was no internal IT person who would be able to govern the data, manage it and have some data modeling capacity to think about how we should gather it. But we had an ATS -- as we use in recruiting firms -- which had some structured data. But when I joined in, I had to deal with a lot of unstructured data, such as resumes and client notes that we keep. All the products that we were trying to build were going to come out of this unstructured data.

We had to hire people to label this unstructured data and turn it into structured data. That was a big pain point initially, but the deployment was easy enough. With the improvements in AI architecture these days, many companies offer fine-tuning and deployments right out of the gate on their platform.

Did you need to purchase any third-party vendor products?

Khan: When we started, we owned the data already, so we did most of the stuff ourselves in-house. One vendor we had to use was an LLM, because we cannot build our own LLM. We looked at all different kinds of LLMs, fine-tuned them and selected the one that was giving us the best outputs -- and there's actually not only one model. There are different models based on the use cases. So, I would consider that a third-party vendor.

The second is AWS. We had to store our data and applications on a private server, and we chose to go with AWS as a virtual private cloud.

What guardrails were most critical to build trust and meet compliance needs?

Khan: We have personal information on thousands of candidates, so we certainly have to prioritize security. We always use our virtual private cloud and private infrastructure to store data, and all access is rule-based, so users can only access what they need.

So far, I only mentioned to you the tools that recruiters are using with the ATS, but they sometimes want to create some type of content, such as an email, using client information. We built our own chat application on top of an open source LLM and secured it through guardrails, such as role-based permissions, serving it in a private cloud and using zero data retention. Even if you use OpenAI, the zero data retention policy means OpenAI will not save any information. We built a private chat app that allows recruiters to use client and candidate information to generate content.

In terms of guardrails, we perform extensive prompt validation to ensure that outputs meet our criteria. Also, nothing gets out without the review of a recruiter.

How did you drive adoption among recruiters and ensure that they trusted the AI outputs?

Khan: This is a big factor when you're building a product. As I said, we focused on administrative friction. This friction was felt by every single recruiter on a day-to-day basis. So, when we started with the resume generator -- the formatter we built -- it saved recruiters so much time that it motivated them to use it independently.

The second thing is that we focused on bringing the tools to recruiters' workspaces. We did not create our own application. Rather, we built the tools in the background -- in the cloud -- and then connected them in the existing ATS workspace.

Have you seen clear ROI or other business benefits as a result of the project?

Khan: We focus on speed and tempo. In terms of ROI, we look at how fast we are moving and how much more we can do in the same amount of time. So, not in the financial terms -- I don't think we have a formula to get that right. But in terms of time, we have seen at least four hours per week returned to every single recruiter. We have also seen -- in terms of the jobs we can produce with the job formatting tool -- at least 10 times the number of jobs we used to produce.

Did your IT or data teams need to acquire new skills for this project?

Khan: A few, yeah. As we are building AI tools, there are always some new AI architectures coming into the market -- MCP, A2A -- things like that. So, we must keep up with that. I'm literally signed up for like 10 newsletters on the technical side. We are constantly looking at different architectures coming out, learning from them and trying to implement them if they suit our use case.

What lessons did this project teach you about aligning AI initiatives with business strategy?

Khan: The most important lesson is to meet the end users. What are their pain points, and are the AI tools really solving their problems? In our company, we have taken the initiative to even go outside of recruiting and ask administrators to write down all the pain points they have in their workday -- let's look at how we can solve those using AI.

Tim Murphy is site editor for Informa TechTarget's IT Strategy group.

Dig Deeper on IT applications, infrastructure and operations