Turn Rejected Applicants Into a Scored, Ready Pipeline

Most people can build a voice agent that talks. That's the easy part. The real challenge is turning it into something that real businesses can use every day, where data is clean, workflows don't break, and recruiters can use the output. In this video, I'll show you how to turn a voice demo into a production system.

So, here are the three things that we're going to do today. One, capture structured data from the voice agent. two, send it to n8n to automate the workflows. And three, push it to a dashboard so the team has visibility and can take action.

For those of you who don't know me, my name is Mike. I've spent 20 plus years in tech recruiting across Asia. And now I build voice automation systems for recruiting and sales teams. Let's jump into the demo and I'll show you how this works.

Hi, I'm a voice agent calling on behalf of Google. May I ask your name? My name is Mike Armstrong. Sorry.

What is this call about? Thanks, Mike. No worries. Really quickly, you applied for a role through LinkedIn and we think your profile might match upcoming roles.

It only takes a minute. Is now okay? Uh, sure. Can you tell me a little bit about the company?

Google is rapidly expanding across Asia, building AI first products and cloud solutions for fast growing markets. We're reviewing applicants whose backgrounds might fit upcoming roles. It only takes a minute. May I ask you four quick questions?

Uh, what jobs are open? Okay, Mike, I've noted your interest. We're hiring across Singapore, Japan, Korea, and Australia in engineering, sales, operations, and data center roles. What jobs do you have open in Singapore?

In Singapore, we're hiring software engineers, account executives, and sales managers. What kind of role are you ideally targeting next? Actually, that's really cool cuz I'm looking for a sales manager job. That fits perfectly.

What salary range are you aiming for? I'm hoping for a salary salary of around 150,000 US. Got it. Dollar 150k US.

When would you be open to changing jobs? I'd be open to change my job in the next 1 to 3 months. Great. So, within the next 1 to 3 months, which region or location interests you most?

Singapore. Thanks, Mike. I'm just logging that information for you now. Thanks for your time, Mike.

We've got everything we need, and someone from the team will be in touch if there's a match. Have a great day. Thank you. Okay, great.

So, the call in successfully. Now, we have a dashboard. This is the product, excuse me, the product is called Reactivator and it's the latest product under the platform which is called hirevoice. And so, we have a dashboard here which we built very easily and very quickly.

It shows us the regions and then it shows job functions that are available. So this would be a dashboard for the recruiter or the TA team to give you and your team better visibility into the candidate pipeline which has been reactivated by the voice agent. So if we look at the top left, we can see the latest call has been logged on the dashboard and we can see the person's name is Mike Armstrong. That's me.

They are or they want a sales manager job. We've also scored them as an 85, which is pretty high because they're interested. And then we get some AI insights into the actual candidate from the call itself. And it says AI insight.

Mike Armstrong is a senior sales professional with extensive experience in B2B enterprise, actively seeking a sales manager role with salary expectations of around $150,000. And we can see that logged individually as well in these respective fields. salary 150,000 and timeline 1 to 3 months region Singapore and the date as well we can also see some tagged words here leadership potential positive regional alignment now we see the next candidate Mark Jackson less of an interesting candidate because they've been scored a 75 because their timeline is a little bit too long so you can see the AI will actually assess the candidate based on specific criteria that you want when you're going out and reactivating these candidates. So I think the benefit is that this will save you and your team hundreds maybe thousands of hours and also you could potentially generate revenue by matching the criteria from this sp dashboard to future jobs that you have available.

And we can set it up so that the AI can tell you, hey, this person actually might be good for this job in the future. Or you can go into the dashboard and say, uh, what job what candidates do I have available for Singapore? And we can go down. We can see we have all these candidates for Singapore.

Great. What about Japan? We got a job in Japan. Let's see.

Okay, we got Jessica Park, VP of Engineering, Fred Jones, Joe Jackson. You know, some of these candidates aren't scored so well. Like we got 50s. They probably we could just ignore them and focus on the candidates at the top.

So, this voice agent combined with the power of AI, voice, and automation with this dashboard that we built on the back end will allow you to go out, reactivate candidates, and have incredible visibility into those candidates. All right. Next, I want to show you the tech stack in the back end for all of you tech people out there. Let's show you how I built this.

Next, I'd like to show you the dashboard and the visual visualization that we built around this workflow and this agent. So this is the reactivator dashboard where we get real-time analytics for what happened and how the call went. So we can see at the top we have higher at scale with AI powered candidate reactivation. Let's get rid of my mic up here.

Right now we have 58 candidates over eight regions and 31 job functions which is quite a bit and it would be good if we could sort of filter through that which we can and I'll show you in a minute. So, we saw the last call that we just conducted between the agent and myself. I was Mark Jackson, the candidate. I'm an AI engineer.

Now, the agent has also scored me as a candidate as an 85 because I'm interested and my time frame is between 1 and 3 months. That's the criteria that we set for this agent. And then we have a summary. AI insight.

Mark Jackson is a mid-level AI engineer with a strong technical background seeking roles in artificial intelligence and machine learning. So, it's pretty cool. That's not actually what I said specifically in the call. I just said I wanted to be an AI engineer.

And there's an intelligence component on the back end that extrapolates this information for the summary. We also have a salary expectations of 140K. His compensation aligns well with mid-level AI engineering roles. So again, the AI is giving the recruiter or the sales team some insights into the candidate assessment.

He's available to start in 1 to 3 months and is based in Japan, a region known for innovation in AI. Okay, good. So then we have salary 140k timeline 1 to 3 months region Japan and the date contacted plus some interesting AI insights at the bottom technical depth interest in high demand domain and regional alignment. So the AI has assessed the candidate scored it and marked it as interested.

We can see the previous candidate Mark Jackson not scored as high. He was a 60 because not really that interested and the timeline was longer. Didn't specify a region. unclear timelines.

So the criteria for the agent can be programmed by you the recruiter and or the manager and then we can use the AI and the intelligence on the back end to score and assess the c these candidates. I think this is where the power really comes in when you're talking about AI automation and intelligence and voice as well. So the candidate is providing the information either over the phone or over a web interface and then the AI is taking that information outputting it into some structured data and then the intelligence on the back end is assessing and scoring the candidates. ai and click on get started which is the top right here or just click book free strategy call.

Two options. So, if you click get started, that will take you down to the assessment and step section. And what happens after you click this button is you will book a meeting that will so book a free strategy call that will take you to my calendarly. And then you simply click a time, click next, and then put your name, your email, and a reason you want to connect.

And then what will happen after that is we will have a 20-minute discovery call and I'll give you a free AI audit like a very quick audit where I can just quickly ascertain your business, understand where the bottlenecks are, where the challenges are, and I can even perhaps put forward a couple one or two proposed solutions as to how AI, automation, and voice can solve some of your your time and money problems. If you're interested in working with me, we do custom builds. We help you build automation back-end systems for sales and recruitment that will help you save a lot of time and a lot of money. If you actually go to my AI my ROI calculator here, this is based on some basic inputs.

Candidates per month, an hourly rate, and minutes per call. So if we're looking at con this agent or this reactivator demo build this product calling let's say 350 candidates you are potentially looking at $165,000 per year in savings. Now, if you really want to scale this out, which is where I think this technology is super compelling and interesting is in its ability to not only gather data accurately and output it in an interesting way for AI and intelligence, but also you can scale this and this is the interesting part. So if we actually scale this out to 500 candidates per month, which is really quite modest, we have some actual deployments with companies recruitment and TA related that are much higher than this in terms of candidate per month and but already your savings is over $225,000 per year.

The really interesting part about the Reactivator product right now as it is is that you can actually generate revenue as a recruiter, as a recruitment company, as a sales organization because this agent is going out and actually reactivating candidates or sales prospects, giving you structured output, giving you information, and then we're using AI on the back end to assess the candidates to assess the information and potentially make placements, close deals. So that's actual revenue generation. So, not only are you saving money, but you're potentially making money. Hope you enjoyed this.

ai, click the calendar, and let's get connected. Next, let's move into how I built this this system, this platform, this latest product. So, we're going to move into the build next. Okay.

Next, we're going to do a quick semi-deep dive into how I built this. For the orchestration, I used ElevenLabs. For the backend workflow, I used n8n. And for my code builder, for the dashboard board, I used Replit agent 3.

So, let's jump into ElevenLabs. So, if you've not done so already, sign up for ElevenLabs. ElevenLabs is a cool voice AI platform. They have voices.

They have their own customuilt voices and then they also have you can clone your own voice actually using 11 Lab. So essentially it's an engine that allows for the creation of voices but they also have a new orchestration relatively new orchestration platform which they've built out they built it out in a more robust way recently which I really like because it's got some more functionality which I like. One interesting thing about ElevenLabs which some people maybe not do not know about is they have some pretty big time investors. Nvidia has invested in ElevenLabs.

Mark Anderson's company, Anderson Horowitz. So the company's now worth $250 million. So there's a lot of money flowing into n8n. So when you look at it from a big picture, it's a platform that definitely has a future.

I think it's going to be a billion dollar unicorn type company in the near future, especially with the advent of voice and the way that the voice and AI and automation are going. So I'm personally very bullish on this platform. I also like some of the new changes. So in the actual orchestration section you need to create a new agent and in the system prompt section.

This is where we create our prompt. We have ro variables call flow required qualification questions adaptive human behavior end of call summary and the data you must collect plus the tool call rules. So we'll go through this quickly. the role.

You are a highly natural, fast, human sounding candidate reactive voice agent. And then it goes through the instructions there. ai. Book an actual meeting with me and we can get connected and I can give you this these resources for free.

I'm happy to share this information for free if you would like to connect with me. Also, don't forget to subscribe to my YouTube channel. Comment. The comments help me understand what I should build next.

The variables. The company name is Google. I'm not actually working with Google. They're not a client.

Full disclaimer, but I like Google. And I especially like the AI business of Google. So, we're going to use it as an example. As you heard in the demo, we've also got jobs.

We've got regions. And then in the call flow section, we basically tell the agent how we want the call flow to go. We want to make this sound like a an actual conversation. Next in the system prompt we have adaptive human behavior.

The candidate must sorry the candidate may interrupt, skip questions or ask unrelated things. So this is the part where you tell the prompt that that you're going to be speaking with a human. Humans are relatively unpredictable in this way. Then the end of call summary is where you tell the agent to call the tool.

We're going to look at that in a minute here. And then you must collect the following data. candidate name, job function, salary, timeline, region, general notes, reactivation logic, and then tool call rules mandatory. So, you're going to send you're going to call the tool and then you're going to send the information to N8 with something called a JSON payload or a JSON string.

Okay, so this is the prompt. Let me know if you have any questions in the comments regarding that. Next, we're going to look at the voices. I'm using Emma.

She's Australian. I like her. If you are interested in other voices in n8n, just go into explore voice library. You can click a type of agent that you want.

So, for example, if we want to listen to the customer service agents, here's Belle. I understand it's all new, but you're doing great. Hi there. You've reached Acme Corp.

How may I help you? Tiffany. Hey, my name's Tiffany. I'm a real Anto.

My buddy's name is Oscar. Whether it's banking, telecom, insurance, or All right. So, I don't particularly like any of those voices. So, I I did experiment quite a bit in the voices section.

I think it was more in the I I settled. I decided on Emma. She's Australian. And you also want to set your language language, which is English.

ElevenLabs is multilingual for this agent. We're just doing an English agent. And then the LLM. Well, actually within the voices, you also want to configure your TTS, which is text to speech model.

And ElevenLabs has three options here. multilingual, turbo, and flash. I'm using flash because I think it is the fastest and most responsive and I think the lowest latency TTS right now in ElevenLabs. So, you know, pro tip, use 11 flash with the stability, speed, and similarity.

I've just used the default settings for this. And then for LLM, you have a variety of choices. LLM is large language model, which is really important. And we have a variety of selections here.

We have anthropic, which has Claude. I haven't actually used Claude so much in voice. I've used it for my low code developer. Sorry, my my my code platform, my my low code platform, my vibe code platform.

GPT is great. I fold I have configured with GPT and and they have some great LMS here. But for this one, I'm using Gemini Gemini throw. We are using Gemini 3 flash, which is the latest version.

It's a lighter version from Gemini 3 Pro and it's also less expensive. I think for our purposes, it's totally fine. And so in essence, it's cheaper and faster, but basically does the same job. That is the prompt, voices, language, and LLM.

Next, we're going to move into the tools section. There's various tabs here. We've got workflow, knowledge base, analysis. We're going to go into tools.

There's some built-in tools. There's end in conversation, detect language, skip turn, transferred agent, transfer to number. I think most of these built-in tools are self-explanatory. We built a custom tool called send_reactivation summary.

And in the description of the tool, it says the LLM must generate a single JSON object containing the following fields. Candidate name, job function, salary, job change, region, reactivation successful, general notes. The JSON object must be passed as the string value of the payload parameter. So we're sending this as an entire JSON payload, all all five of these fields to n8n.

and that's in the description here. Then we have some requirements. The method is post. The URL you will get from n8n in the web hook in this section here in the production URL.

You grab this, you copy it, and then you just simply paste it into the URL section. The rest is just default. I didn't change any of this. However, you go down 1 2 3 4 five sections down.

It's called body parameters. We do want to change this. In the description we write sends the structured JSON summary of the call to the n8n web hook after the conversation. So after the conversation per the prompt instructions we are sending the five dynamic fields as a JSON payload to NAD and in the properties section you configure as follows.

Data type string identifier payload it's going to be one single payload value type LLM prompt. I don't know what other okay the other options are constant and dynamic but we want to use the llm prompt and then in the description this is similar to above the llm must generate a single JSON object containing these fields it's the same fields so that is the custom tool that we built in the tool function section of 11 labs next we're going to look at n8n okay we are in n8n you need to create a new workflow the first node that you want to build is a web hook node which will be your primary intake a node for the data that's coming in and so in the production UR you have a URL you set the HTTP method HTTP is a protocol for sending data you want to send that method method as a post is basically intaking data into the web hook and the path will well you just name this we've named it reactivation- summary you can call it whatever you want but that will go on the end of the post URL and that's your entire web hook post that you put in again you put that into the 11 labs function method post here. So the the function the custom function tool call that we set up in 11 labs that comes from here authentication none respond using respond to web hook node which will be the last node here on the right hand side. Next, after the web hook, we create a code node.

And this is going to simply be mode run once for all items language JavaScript. And we're going to we've written some JavaScript code in here. Some basic JavaScript script. I'm not I'm not a developer by the way.

I just use GPT to create this code. What this is doing is is taking the JSON and payload, which is five fields in one payload, which actually we can see down here in the input. You can see body payload all the information the data has come in one single payload. So this code node is actually separating that data into five separate fields name Peter Jackson function AI prompt engineer so on and so forth.

Then after that we send the data to an HTTP request three you can see the I wonder if it will where's the code here? Well anyway this is the console. Here's the publishing section. Here's our database.

Here's the preview. I wonder if we can look at the well we can actually see the dependencies here on the on the right hand side. So this is all the the the actual code for the dashboard that has been set up but actually all you need to do is ask your code developer to create a URL for you which you then post into this section of the HTTP request node. So ask the code developer to develop or to generate a URL.

You paste that in authentication none. Now, so now what we're doing is we're sending the data, the five separated fields from our NAD HTTP request node into the code developer which is Replit. You can use any code developer of your choice. I'm using Replit.

Most of my contemporaries are using Gemini 3, which is great. I prefer Replit 3 right now, but anyway, that's another video. Send headers. Specify headers using fields below.

This is pretty standard stuff if you've done this before. header parameters name content- type v value application-json. So this is basically just structuring how the data is going to set be sent via http and then in the body section we're sending this as JSON. So body con content type JSON specify by using JSON and then the in the expression section down here again this is coming in through the code node.

We put some JSON in here with dynamic data. It's written like this. ai. So that looks good.

We can filter. I've asked the code developer to filter on regions. These are all of my Australia candidates. These are my Japanese candidates.

These are my Singapore candidates. These are all the candidates. We can also sort by function. Um so that's kind of cool.

This is in the dashboard. This is all using the intelligence component on the back end. So this is where AI automation and voice come in when they all sort of converge into a very powerful solution where you're taking information from a single call and then you're adding it to a dashboard. And when you do this at scale, it's very powerful.

It's very fast. We're really just kind of scratching the surface. This is the tip of the iceberg and in terms of the power of this stuff. Top left card, Peter Jackson.

So this is the information that came in from the demo. The candidate name AI prompt engineer which is what the candidate does. We've scored this candidate. Again we we've asked the prompt agent the the replet agent to assess candidates based on timeline.

This candidate's timeline it says 6 to9 months actually. That that should be maybe a red candidate but anyway typically if it's 1 to 3 months we'll score the candidate high but the candidate's interested and the candidate is an AI engineer. So that's probably why the the intelligence component assessed the candidate as an 80. Then we've got some the summary which I read earlier, the broken down fields and some green cards and some red cards.

So the power of this is again we've gone from a call into a backend workflow. We separated all of that JSON payload data to an output to a structured five field output and then we have created some cards and some assessment using the intelligence AI component in our vibe coder to make this look good. The benefit to recruiters to TA to sales to you as a manager is that this looks good. We've gone from a call to a dashboard which looks great.

It's going to help you save time and save money. The really interesting part is the the eyes. So like if you're scaling like let's just say your inputs are 30 candidates per month you know your revenue impact per year is $225,000. This is 15 faster hires 15,000 average value.

So this is actual either money that you're saving or potential additional revenue revenue that you can make by reactivating candidates by identifying candidates and doing this at scale using AI using automation and using voice. So you can see that the the cost savings and the ROI is pretty extreme. By the way, if you're interested in playing around with my RO ROI calculator, it is on my website. It's pretty like it's very powerful, but it's really easy to do.

You just, you know, change these sliders and you'll get some some different numbers. I think the interesting one is like let's just set the call for 10 minutes and let's set the hourly rate of the recruiter for $60. And like like you can see the cost savings are pretty extreme with this green slider here. That's it.

If you like my my video, please subscribe to my YouTube channel. My YouTube is Mike Armstrong-ai. I drop videos weekly. I do builds mainly around recruiting and sales.

Please comment on the video. Your patronage is very much appreciated. If you like the video, comment. If you want to connect with me, go to my calendar and click on book free strategy call.

Go into my calendarly. Click a day. Click a time. Click next.

Put your name. put your email, schedule the event, we will get connected. I will help you build agents. I will help you like totally change your business, transform your business.

Automate will not only save you money, but help you make money. This stuff is real. This is happening. I love you guys.

Thanks for thanks for tuning in and I'll see you in the next one. Bye for now.