We Couldn't Afford to Review 900 Applications, So We Built an AI Interviewer

One of my companies had almost 900 applications waiting, and we could not afford a bigger team to go through them. We built an AI interview system, the prototype took two weeks, and within ten days of the prototype the queue was empty.

It costs us around $50 a month to run the system. That doesn't include the work to build it, and people still conduct the HR and department interviews, but the running cost was small enough for us to handle work we couldn't get through before.

We were growing quickly at the time, and sales, operations and marketing all needed people. Automating this process was essential for our growth because the delay in hiring was holding up work in the rest of the company.

So the question I would put beside the AI budget is, what happens to your business while that work keeps waiting? For us, the cost decision started there.

We had more applications than we could handle

The other departments were already angry that hiring was taking so long. They needed people to do the work, applications were coming in, and the queue kept getting larger because HR couldn't work through it.

That was demotivating for the team, with the queue growing while the departments waiting for hires needed the process to move faster.

You could look at that and say we needed more people in HR, and yes, we needed more capacity. But bringing new HR employees up to speed was difficult and would take a long time, and expanding the team was more than the department could afford then.

So asking the existing team to try harder wasn't going to solve it. We needed to find a way for them to handle the applications with the time and money we had, and that was our responsibility as owners.

The system we built analyzed applications, pre-screened candidates, conducted interviews and managed the candidate Kanban, basically the board showing where candidates are in the hiring process.

And when HR took over, the human interview was better. The candidate already had more information about the role and had asked some basic questions, while our interviewer knew more about the candidate before they spoke.

So the conversation was more thorough and useful to both sides, rather than simply being a faster way to move someone to the next step.

After the short interview with HR, an approved applicant moved to the department interview much sooner, already knowing more about the position, the work they would need to do and the team.

That's what I mean when I say we hire better and faster, and we received good feedback about the speed of the process. The system ran fine, while people still conducted the HR and department interviews.

The $50 didn't include the build

Those two weeks to build the prototype belong in the cost of the system, even though they don't appear in the monthly running bill. If you're considering something similar, you need to make room for that work before you can depend on the new process.

Someone has to decide what the system should do, set it up for your hiring process and try it with the people who will use what it produces. A low monthly price doesn't tell you how much of that work is included or who will do it.

And HR still needs to interview the candidate and decide whether they should meet the department. That's hiring work, not evidence that someone has to redo the AI's work.

For your budget, keep the build effort and running bill separate, then describe the work people will still do. In our case, that's the HR interview and the department interview, with better-prepared people coming into them.

I don't have a complete dollar cost or a measured payback period to give you for this system. And our two-week prototype is the time this build took, not a timetable I'm promising for yours.

Compare it with the work that stays waiting

We hadn't been paying a larger recruiting team and then stopped, so I can't count its salary as money we saved. What changed was our ability to get through applications and move people toward the departments that needed them.

A project can help your company do needed work without removing an existing expense.

If you already have someone handling the applications well and candidates reach the right interviews on time, you have a different decision to make. You would need to show how the proposed system improves that process enough to justify building or buying it.

In our case, the people waiting for those hires weren't all in HR. Sales, operations and marketing needed staff, so looking only at the recruiting budget would miss why getting through the applications mattered to the company.

That's the comparison I would want you to make in your own business. Name the work that's waiting and what it prevents other people from doing, then compare the realistic ways you could get it done.

In our case, we got through the application queue and improved the interviews that followed. That is the result I'm describing, not a claim that AI should make your final hiring decisions or that every growing company needs to build an interviewer.

Put the whole job in the budget

Before you buy or build, sit down with the person responsible for the waiting work and the person who needs its result. For a hiring process, that means understanding both what HR can handle and what the department needs before it can decide to hire someone.

Then write down which part of the work you want AI to take on and what the person taking over needs to receive. In our case, the human interview went better because both the candidate and the interviewer came into it knowing more.

Put the setup effort beside the monthly running budget, and name who will still make the decisions. If the proposal only tells you the software price, ask who will get it ready for your process and how much time your own team needs to give it.

When you try it, follow the work beyond the AI step. Look at whether candidates reach the human interviews better prepared and sooner, and whether the department gets what it needs to make a decision.

Before committing to the tool, agree with the people on both sides of that handoff what has to improve. That gives you something to judge the system by beyond a smaller queue.