Most recruiting tools promise "AI-powered sourcing" but give you the same generic LinkedIn searches with a chatbot wrapper.

I built my own AI sourcing agent instead — and it outperforms every vendor tool I've tested.

Here's what's different: Instead of one-shot prompts that return garbage results, I created a multi-step workflow using Claude's code capabilities.

The agent starts with ecosystem mapping — identifying companies, communities, and publications where your ideal candidates actually spend time. Then it moves to individual sourcing with built-in quality checks.

30% of initial candidates get filtered out automatically based on criteria I define. No more manually sifting through hundreds of irrelevant profiles.

The key insight: You need an agent that can write code, not just search databases. Once it can code, it can build custom tools for any workflow you throw at it.

I broke this down step by step in this week's video, including the full technical walkthrough and all the code you need to build your own.

https://lnkd.in/eyVwvgSW

What's your biggest frustration with current sourcing tools — the volume of bad matches or the lack of customization?

#AIforHR #TalentAcquisition #HRAnalytics #PeopleAnalytics

I've been collaborating with Claude Code for several months on developing a comprehensive recruiter sourcing screener.

Key features include:

- Sourcing candidates from platforms such as Meetup, GitHub, LinkedIn, and other repositories, including Kaggle.
- Scoring candidates to identify the best fits.
- Engage candidates from companies that are actively posting jobs for the roles I aim to fill.
- Eliminating contractors from the candidate pool.
- Providing a concise shortlist of candidates for direct communication.

This innovative approach streamlines the recruitment process and enhances the quality of candidate selection.

Confession: my team and I use AI a lot in recruitment. Like… a lot. If there’s going to be “brain atrophy,” we might be early adopters 😭

For each role, we’ve built custom agents to dig into candidate experience. It’s sharp. Efficient. Slightly terrifying for a Gen X like me.

Until recently.

We had a candidate who was a perfect match. Almost word for word aligned. Clean, structured, strategic.


A tad too perfect, but the companies she’d worked in didn’t match the sophistication she described. The story was polished, but the environment didn’t support it.

We caught it early. But this is where things are heading.

Right now, recruiters are using AI to screen smarter. Soon, candidates will use AI to answer smarter. It’s an arms race.

And the issue isn’t AI. It’s misrepresentation.

If your process can be reverse-engineered, you’re probably testing outputs, not thinking.

Real experience has texture. Operational detail. Trade-offs. Even contradictions that still make sense. AI can generate great answers, but it struggles with lived coherence. For now.

So the game shifts.

Not “who uses AI better,” but who can spot gaps. Who can probe depth. Who can tell when the environment doesn’t match the expertise.

AI can prep the file. But instinct still catches what doesn’t add up.

And if recruiters outsource their thinking completely, that’s the real risk.

The next era of hiring won’t reward the best prompts. It’ll reward the sharpest discernment.

Let’s not get lazy just because we got efficient. 💡

Recruiters! I'm 50/50 terrified/in love with AI. But is it making us terrible recruiters? 👇


Anthropic tested 52 junior engineers learning a new Python library. The group using AI scored 17% lower on comprehension.

The ones who asked AI "why does this work?" scored above 65%.
The ones who just asked it to write the code scored below 40%.

METR found that developers using AI took 19% longer to complete tasks. They expected a 24% speedup. After experiencing the slowdown, they still believed AI had made them faster by 20%.

They got slower. They thought they got faster *Luke raises his eyebrows in suspicion*

Wharton tested 1,372 people across 9,593 trials. Acceptance of incorrect AI answers hit 79.8% because people just... stopped checking. They also felt more confident because they had checked AI.

SO...The million dollar question is, would any of this "cognitive surrender" spill over into recruiting?

How about your sourcing instincts? AI runs the search. You stop building market knowledge. When the results are rubbish, you can't tell because you've lost the ability to judge what awesome looks like.

Your outreach craft. AI drafts your messages. You stop practising the skill of writing something a real person would actually reply to. Reply rates drop and you blame the market.

Your candidate assessment. AI screens CVs. You stop developing the wee internal alarm that fires when something doesn't add up.

The research also shows AI as a tutor accelerates learning. Students in AI-powered learning environments scored 54% higher. So this isn't about ditching the tools. It's about whether you're using them to think better or to avoid thinking altogether.

Two things you can do this week:

1️⃣ When you use AI, ask it to explain it's reasoning before you accept the output. "Why did you rank this candidate higher?" forces you to engage critically.

2️⃣ Spot-check one AI output per day against your own judgement. If you consistently can't tell when it's wrong, that's atrophy.
What's one recruiting skill you've noticed getting weaker since you started using AI?


I'm a constant AI user, its embedde dinto every part of my working life so I'm not saying "Boooo AI" I'm challenging us to properly discern where AI makes us baetter and where AI. makes us worse.

I'd LOVE a massive multi-sided chinwag on this one, do you agree? how do you choose what goes to AI and what stays with you squidgy organic things between your ears? Tell me in the comments!


Hi 👋 I'm Luke. I empower recruiters with data. Want to get data-driven for free? Link in my profile for my weekly newsletter.

AI gets hiring wrong more often than people realise.
I use AI in my workflow every day.
But it’s not deciding who’s a strong candidate.

If anything, it’s where mistakes can happen.

On paper, AI can highlight matches.

In reality, it misses context.

It doesn’t understand what someone actually built, how deep they went, or how they think under pressure.

That’s where experience and pattern recognition come in.

Over time, you build a feel for what “good” actually looks like.

Without that, you miss the candidates who don’t perfectly fit on paper but are actually the right hire.

Where AI does help me:

→ Structuring my screening calls
→ Highlighting areas to dig into early
→ Making sure I’m asking better, more targeted questions from the start

Which means I get to signal faster.

And give better insight to both candidates and clients.

AI is useful.

But it’s a tool around the edges.

The judgement piece is still very human.

Engineering leaders
Where have you seen AI get it wrong in hiring?

Engineers
Do you feel like these tools are helping or hurting how you’re being assessed?

I’ve been using AI in HR for months — here are the tools that actually changed my workflow
As a Global HR Director with years leading people operations across aviation, corporate shared services, and high-growth enterprises, I used to spend hours on policy drafting, reporting, content creation, payroll, and research.

I integrated these AI tools into my daily HR practice. The result? Faster strategic decisions, 95% grievance reduction, 50% lower cost-per-hire, 100% POSH compliance, and far more time for actual people leadership instead of admin drudgery.

Here are the AI tools that genuinely transformed my workflow:

1. Gamma (Branding & Communication + Content & Presentation)
Instantly creates stunning employer branding carousels, LinkedIn hiring posts, executive presentations, and training decks — from full day to 10–15 minutes.

2. ChatGPT + Claude + Gemini (HR Digital Transformation & AI)
My go-to for drafting HR policies, NDAs, job descriptions, grievance responses, and POSH training content. Cut document turnaround by ~70% while maintaining full compliance.

3. NotebookLM + Julius (HR Analytics & Reporting + AI Productivity)
NotebookLM turns reports/meeting notes into instant summaries and audio podcasts. Julius supercharges data analysis for workforce metrics and MIS reporting.

4. Grok + Perplexity (Research & Insights)
Real-time, accurate research on labor laws, market benchmarks, and HR trends — without paywalls. Perfect for quick leadership advisory.

5. Merlin + WhisperFlow + Formula. dog (AI Productivity & Knowledge Management)
Meeting transcription, action-item extraction, and workflow automation. Freed up hours every week.

6. Naukri Maestro + LinkedIn Recruiter + Zwayam + DoSelect (Talent Acquisition & ATS)
AI-powered sourcing, screening, and candidate matching that shortened hiring cycles
dramatically and lowered cost-per-hire by 50%.

7. Yoodli + Playground (Employee Experience & Learning + Branding)
AI communication coaching and creative visuals that boosted engagement and employer branding.

8. Debugcode (Technical & Automation Skills)
AI-powered code debugging and workflow automation for custom HR solutions.

Bottom line: AI didn’t replace the human side of HR — it amplified it. I now spend more time on crisis resolution, leadership advisory, culture building, and strategic transformation.

These AI tools helped deliver measurable outcomes: 95% faster grievance closure, 50% reduced cost-per-hire, 100% payroll accuracy, and stronger employer branding.

If you’re in HR or talent, which AI tools have actually moved the needle for you in the last year? Drop your favorites in the comments!

AI almost ruined one of my favourite hires.
We were hiring for a senior engineering role. The AI screening tool ranked one candidate near the bottom. His CV was non-linear, no clean keyword match, no obvious pattern.

On paper, he looked like a risky bet.

But something in his profile and side projects caught my attention, so I decided to reach out anyway.

In the conversation, it turned out he had exactly the depth we needed. He had solved similar problems before. He just was not someone who optimizes his CV for algorithms.

He is now one of the top performers on that team.
The AI was not wrong. It just did not have the full picture.
And that is the part that worries me.

Because if I had trusted the tool and moved on, I would have never spoken to him. And neither would anyone else.

That experience changed how I use AI in recruitment.

I let it help me prioritize my time. I do not let it make the final call.

AI is great at sorting. It is not great at understanding context, career detours, or the kind of depth that only shows up in a real conversation.

The more we automate screening, the more important it becomes to know when to override it.

If you are using AI in hiring, here is the question I would ask yourself:
Are you treating it as a smart assistant, or as a gatekeeper that quietly closes doors on great people?

AI can build a beautiful shortlist on paper.

But who’s having the conversations that actually create that shortlist? 🗣

As a marketing recruiter, I use AI. It’s fast. It’s efficient. It can surface relevant CVs in seconds. I'd be silly not to use it.

But AI doesn’t:
• Sell the opportunity in a way that excites passive talent
• Sense culture misalignment before it becomes a costly mistake
• Translate a CV full of buzzwords into actual commercial impact
• Challenge a candidate on vague metrics and surface real numbers
• Identify untapped potential beyond job titles
• Apply market context: salary shifts, team restructures, brand reputation

The list goes on.......and on.......

The strongest shortlists aren’t built from keywords.
They’re built from conversations.

Real discussions about:
➡️ Brand ambition
➡️ Commercial targets
➡️ Team dynamics
➡️ Leadership style
➡️ What success actually looks like in 12 months

The best marketing candidates?
They’re rarely mass-applying. They’re busy. They need context. They need challenge. They need a reason to move.

That doesn’t come from an algorithm.
It comes from insight, credibility, conversations and proper headhunting.

AI can filter.
Recruiters create clarity.

And clarity is what secures interviews, and better hires.

So today, I'm asking - where do you think AI stops and human judgement starts in recruitment?

Kin Collective Recruitment

I’ll be honest.

I did not want an AI recruiter.

We’re in the people business.
Our reputation is built on relationships.
And I’ve spent 20+ years telling clients that great recruiting is instinct, judgment, and nuance.

So the idea of introducing a digital, agentic recruiter into Tier4 Group?
It felt… risky.

What if candidates hated it?
What if it diluted our brand?
What if it made us feel transactional?

But here’s what leadership really is:

It’s not protecting what’s comfortable.
It’s testing what’s possible.

So we built Taylor, powered by AlexAI.
And we measured everything.

Here’s what actually happened in our first 12 months:
11,500+ completed interviews
38,000+ invitations sent
30% completion rate
84% of candidates rate the experience 4/5 or higher

That 84% stat stopped me in my tracks.

Because I care deeply about candidate experience. It’s personal. Our name is on it.

What I’ve learned:

Taylor (AI) doesn’t replace our recruiters.
It replaces friction.

It doesn’t remove humanity.
It removes back-and-forth scheduling emails at 9:30 at night.

It doesn’t make decisions.
It creates structured insight faster so our humans can make better ones.

And here’s the part no one talks about enough...

Sometimes as founders, we resist change not because it’s wrong… but because it challenges our identity.

I built my career on being the person in the room reading between the lines.

But scaling that instinct requires tools.

Taylor hasn’t made us less human.
It’s made our humans sharper.

More time advising.
More time closing.
More time building trust.

We’re still learning. We’re still refining.

But I’m glad I didn’t let fear win.



Oh, and here are some stats I asked our team to pull on the time it would have taken a human to complete all of those interviews and the tasks associated with them. Go ahead and tell me we made the wrong move... I'll wait.

10,650 hours = about 266 full-time weeks of work (at 40 hrs/week)
That’s about 5.1 FTE-years (assuming ~2,080 hrs per year per person)

I DID IT. I AUTOMATED MY RECRUITING BACKEND WITH CLAUDE CODE.
And ironically, it made me more confident in the future of recruiting, not less.

I recently learned about the "doorman fallacy". When automatic doors were introduced, many hotels assumed they would no longer need doormen. They believed the job was simply opening the door. What they failed to recognize was that the visible task was not the value. The value was the relationship. The doorman greeted guests by name, built familiarity over time, hailed cabs, carried luggage, and created a sense of trust and hospitality that no automated door could replicate.

That framing has been sitting with me because over the last two days I have nearly automated my backend recruiting workflow using Claude Code. I built systems that identify candidates in my vetted network, surface them to me for open roles, screen my inbox to determine who else I should be speaking with from the VC talent community, who I have not followed up with, and draft succinct notes pulled from my extensive candidate notes explaining to founders why specific matches from our network make sense. It is far from perfect. It is scrappy, occasionally messy, and could be significantly optimized. But it works, and it will only get better as the models improve. As a talent partner supporting an entire portfolio, that speed matters. More importantly, it means I am spending less time clicking through profiles one by one and more time going deep with founders, hiring teams, and our candidates.

This reinforced something I have believed for a long time. The mechanical parts of our jobs are not the reason we have them. Searching, tagging, sorting, summarizing, and writing repetitive emails were necessary steps, but they were never the moat. The moat is trust built between humans. It is a founder calling you first because they need someone in their corner, not just in their inbox. It is a candidate being honest with you because they know you will be honest back. It is pattern recognition that only exists because of years of real conversations with real people who chose to let you in. The moat is not what you do. It is who you are to people.

As a recruiter, if you define your role by opening the door, automation feels threatening. If you define your role by building the relationship, automation is leverage.