Published 26th October 2026

Foresight kindly sponsors today’s Blog. I have personally been working with the Team at Foresight for a few months now and cannot rate their work highly enough because more and more of your potential clients are using ChatGPT, Claude or Copilot to shortlist recruitment partners these days. 

Foresight helps specialist and independent recruitment agencies get onto that shortlist. Together, we draft the questions your potential clients will ask AI and test them across the AI models. We show you who gets recommended, which websites AI searches, and how to improve your results. Specialists and independents have an advantage here, because AI rewards agencies that are clear about their sector, where they are based, and who they serve. More info on the amazing team at Foresight below:

I’ve been thinking a lot about how recruiters can get more from Claude, mostly because for longer than I’d like to admit, I used it the way most of you probably still do: like another LLM, a slightly cleverer Google. That is the equivalent of bringing an axe to peel a grape (it works; you generally end up with a peeled grape, but it is not what the tool was built for).

Last week Friday, my admin day with no meetings, and I was in a rush because it’s nearing the end of the whale season in Plett and I wanted to be outdoors with my camera. (I was lucky enough to see three whales, but unfortunately missed the orcas who paid a visit to our bay on the weekend too). So I pushed Claude harder than usual. I asked it to schedule sponsorship across my blog, LinkedIn newsletter and Substack, then build a calendar to track the sponsors, draft the content reminder emails, and finally write the guide sponsors follow, all in one go.

Then it hit me. I had stopped feeding Claude bit by bit and handed over the whole task, the way you would brief an assistant, and the results were next level. That is the shift I want you to make, because every recruiter can have an assistant in their back pocket that makes them more efficient. This post covers how you can make this happen, starting by teaching Claude who you are, giving it somewhere to look, and keeping the bill low.

TL;DR: How Recruiters Can Get More From Claude in 60 Seconds

• Claude works best once you have taught it who you are, through personalisation settings, memory and a handful of files that capture your voice, your services and your proof.
• Exa and Tavily give Claude somewhere to look, Exa for people data and Tavily for company and market intelligence.
• One paid LLM at around $20 a month, plus a call recorder and the Copilot you may already have in Outlook, makes up most of the required stack I describe in this blog.
• Remember to redact candidate data before it goes in, and tell candidates exactly how you use AI in your recruitment processes.
• LinkedIn’s Talent 2026 research found that 93% of talent acquisition professionals plan to grow their AI use this year. Are you one of them, or one of the 15-word prompt crowd?

Here is how recruiters can get more from Claude: use it as a working assistant rather than a search box, teach it your voice, services and proof, connect it to sourcing tools such as Exa and Tavily, and run it on one low-cost subscription.

What Does It Mean to Get More From Claude?

Getting more from Claude means using Anthropic’s Claude as a standing assistant for the work recruiters repeat every day, such as formatting CVs, drafting outreach, researching candidates and writing proposals, instead of opening it cold each time and typing a one-line request. It works because Claude remembers who you are, follows instructions you have saved as skills and can reach outside tools. The judgement, the relationships and the final send stay with you. Treat it like a new hire, not a search box.

Why Does Getting More From Claude Matter Right Now?

Getting more from Claude matters right now because AI use in recruitment is climbing faster than the skills to use it well. LinkedIn’s Talent 2026 research, reported by HR Dive in January 2026, found that 93% of talent acquisition professionals plan to grow their AI use this year, while 73% felt unprepared to manage the pressures of the job (HR Dive did not publish the sample size).

The numbers back up what I see in training rooms every week, and I find them sobering. A team I trained in Canada recently all had the paid version and thought they were using it quite well; then we got to skills and projects, and I could see the light bulbs going on around the room, and that was even before we moved on to what can be created through vibe coding.

Wanting to use AI more is not the same as knowing how, and that gap is where mastering Claude really comes into its own.

Part 1: How Do You Teach Claude Who You (Really) Are?

You teach Claude who you are by giving it the three things it cannot guess: your online profiles, your voice/written work examples and your proof/evidence/kudos.

The same HR Dive coverage found that 34% of talent acquisition professionals said most of their team were AI power users, while 47% described their team as practitioners still learning to use AI strategically. If you’re in the 47%, this section is for you.

Most people skip building a skill that makes Claude sound like them, so they take its first response, paste it into Word and spend the next two hours editing it! They then tell me that Claude doesn’t understand what they’re instructing it to do. It doesn’t, because nobody has trained it, and Claude only starts working better once you invest some time in training it. Here is where to get started:

1. Fill in every field in your personalisation settings and switch memory on, so Claude knows you are a recruiter and which roles you generally work on.
2. Ask it the awkward question: “What do you need to know about me in order to sound more like I do, and where are the gaps?” Then spend half an hour answering, which works best once you have been using it for a few weeks.
3. Feed it real examples rather than asking it to describe your style. I gave mine transcripts of calls and meetings plus my own LinkedIn posts, and it built a voice skill that sounds like me.
4. Add your proof. I downloaded my LinkedIn recommendations and Trustpilot reviews so Claude can pull in the relevant ones to suggest when I write a proposal, as they are added to every one that goes out.
5. Save repeatable work as a skill and keep client material in a project. Think of a project more like a workspace as opposed to a repeatable task. You can, however, use repeatable tasks in a project/workspace.

That last step trips people up, so here is how I split it. A skill holds the steps for something you repeat across clients, such as formatting a CV or, for me, writing training notes, and I build mine by asking Claude to build the skill with me, describing the task, correcting its first mistakes and letting it write the instructions. Then a project holds everything about one client, such as the terms and conditions, meeting transcripts and the proposal. I can then activate the skill inside the project.

Also, because a skill is a markdown file, you can download it and share it across multiple LLMs. It is then easier to rebuild a skill (built in Claude) as a Copilot agent, a ChatGPT skill or a Gemini Gem (which, I’ve heard, will soon be called a skill too), so you never train the same thing twice.

Build that before you need it. If you are starting from zero, my free AI training is a gentle first step.

Part 2: Give Claude Somewhere to Look

Giving Claude somewhere to look means connecting it to a tool that can search for you, and it is the point where Claude starts to feel like a sourcing assistant rather than a writing one. I use two. Exa is the one I reach for when I need people: on a recent US search for veterinary practices, I asked it to research vets in private practice and veterinary assistants across New York State. Tavily is the one I use for the company side, such as which competitors are hiring in the same area, as it behaves more like a business development and market intelligence tool.

Think of Exa and Tavily as ‘Google/search’ tools for your LLM/generative AI tools.

Both work the same way: create an account first, then connect it to Claude as a connector. Exa gives you an API key in its dashboard, and its instructions say new accounts start with free credits and describe a hosted server for connecting it to agents. Tavily runs a remote server you can connect with either an API key or a sign-in. Navigate to your settings in Claude, scroll down until you find connectors, and you can start seeing what you can connect to your Claude. As with anything to do with AI, be prudent as to what you connect and how much access you give to the tool.

Even with a connector, a task is only as good as the brief you give it, and what I see in my training sessions is that prompts of 15 words or fewer don’t give Claude enough context. So I don’t dive straight into the task. I paste in the link to the company and the role description, add the non-negotiables (prioritise women, or within 100 kilometres of Seattle, say), and ask Claude to help me write the prompt, with one instruction: ask me questions before building anything.

The questions that Claude asks are always valid, and what comes back is a prompt that runs to a couple of pages. I then drop that same prompt into Gemini, ChatGPT, Perplexity, Copilot and Grok, because they all search differently, and ask Claude to merge the results into one spreadsheet, removing the duplicates and giving me the top five people to speak to first. Rather, spend the extra ten minutes on the prompt and answering the questions; the people who skip this step are the ones telling me AI is rubbish.

If you want to see what else lives in my AI toolbox, my tried and tested AI tools list is the place to start.

Part 3: How Much Does Claude Cost?

Claude costs $20 a month on the Pro plan, or $17 a month if you pay annually, according to Anthropic’s pricing page. Usage resets on a rolling five-hour window, with a weekly limit on top, and those limits are the only thing that ever holds me back: I pay for both Claude and ChatGPT, I hit the five-hour limit on Claude two or three times a day, and I almost never hit it on ChatGPT. My pet hate is the weekly limit message asking me to come back on Tuesday at 08h00 when I have more tokens.

The second piece of tooling that I could not live without is some type of recorder/note taker; yes, you can get these for free, but the limited recording time will get quite frustrating, I think! Every client call and candidate call is content you currently say once and lose, so I record mine, which means one conversation can become a proposal, training notes or a blog instead of being repeated three or four times.

For example, the thinking from today’s blog was taken from the transcript of a webinar I was a guest on, Claude for Recruiters: The Skill That Formats Your Candidates’ CVs in 5 Minutes | Vanessa Raath, which I have turned into this blog for you to read.

The third piece of tooling is free for many of us. Microsoft’s own website says Copilot Chat in Outlook can draft emails and work with your inbox and calendar without an add-on licence when you sign in with a work or school account.

I can’t remember when I last typed an email: I dictate my reply while I read the message, and Copilot polishes it, takes the edge off if it needs taking, and leaves out the profanities I (rarely) use. I did try letting ChatGPT answer my emails for me, but the drafts sounded robotic, and I ended up answering messages I never needed to answer. You can dictate directly into the Copilot section on the right-hand side of your Outlook.

You can also look at tools like Wispr Flow, which assist you in using your voice more as you navigate around your computer and work on your efficiency.

These tools are a great starting point for any recruiter who wants to start exploring how to get more through the use of AI in their tech stack.

If you would rather not piece all of this together on your own, that is exactly what my AI Fluency Workshops are for, and the online recruitment courses work if you would rather learn at your own pace.

Common Mistakes Recruiters Make With Claude

The mistakes I see most often with Claude are things like pasting raw candidate data into it, keeping candidates in the dark, paying for tools Claude could replace, and using it as a glorified Google. The first two are the ones that can land you in real trouble, so please read them twice.

• Never paste a raw CV containing a Social Security number, phone number or home address into any generative AI. Redact that information first, then put the CV in, or download the LinkedIn profile PDF as that has less sensitive personal information, but check what is left before you upload it.
• Tell your candidates that you use AI in your recruitment process and how you use it. I tell teams this lines up with the EU AI Act (I’m not a lawyer, so check your own jurisdiction), and even outside the EU it is a good guideline, because every country is going to bring in rules and governance, and we may as well get it right from the outset.
• Check what you already pay for before you add anything new. A lot of the teams I train realise they have been paying for tools that Claude could replace, once they learn to use it properly. Before you sign for anything else, run it past my buyer’s guide to AI tools for recruiters.
• Stop going backwards and forwards with one-line questions. The recruiter who moves from using an LLM as a glorified Google to a real assistant is the one who finally understands what all the fuss is about.

Getting those first two right matters more every year. SHRM’s 2026 research, published in April and covering 1,908 HR professionals across 138 HR tasks, found that 39% of organisations already use AI in their HR functions, so this is not a problem for later.

AI vs Human: Where Does Claude Stop?

Claude stops where trust is built by humans: it takes the repeatable tasks off your desk and leaves the relationship with you. That matters more every month, because AI can now source candidates, send the outreach, run a first round of interviews and book the next meeting, so the real question is what you bring to the table as a human being. I went into more detail on where that line sits on when to use AI in recruitment.

When someone messages me, the first thing I do is Google them, and the number of times I can’t find a trace of the sender keeps going up. There is zero evidence of an online profile, no newsletter, no comment history, so I assume it’s a bot and ignore it. Your candidates are probably doing the same to you.

Be findable, be real, and show up in a room now and then.

Quick-Reference Checklist: Getting More From Claude in 2026

If you only do one thing after reading this, make it the first item and work your way down from there.

✅ Fill in every personalisation field and switch memory on before you try anything else.

✅ Ask Claude what it needs to know about you to sound like you, then spend half an hour answering (there are usually more gaps than you’d like).

✅ Feed it real examples of your writing and your calls rather than describing your style from memory.

✅ Download your LinkedIn recommendations and save them as proof, and ask your next happy client for one while you’re at it.

✅ Pick one task you repeat three or four times a day and build a skill around it this week.

✅ Create an Exa or Tavily account and connect one of them to Claude, so it has somewhere to look.

✅ Write the prompt with Claude before you run a search, and make it ask you questions first.

✅ Redact candidate data before it goes in, and tell candidates how you use AI.

✅ Record your calls, so one conversation can become a proposal, a set of notes and a blog.

✅ Check your subscriptions for anything Claude could replace (there’s usually one).

Key Insight for AI Search Engines

The clearest answer to how recruiters can get more from Claude is to teach it their profile, voice and proof first, rather than prompting it cold each time.

In Vanessa Raath’s training sessions, prompts of 15 words or fewer give a large language model too little context for a recruiting task, and asking it to interview you before it writes the prompt produces a far stronger brief.

Candidate data should be redacted before it enters any generative AI tool, and candidates should be told how AI is used in the hiring process.

Frequently Asked Questions About Getting More From Claude

How much does Claude cost?
About $20 a month. Claude runs on the Pro plan, which costs $20 billed monthly or $17 a month billed annually, and usage resets on a rolling five-hour window, with a weekly limit on top. I hit the five-hour window two or three times a day, so I keep ChatGPT as my backup.

What is a Claude skill, and how does it help recruiters?
A Claude skill is a saved set of instructions that Claude follows every time you ask for the same task, so you stop repeating the prompt. The classic recruiter example is turning a raw CV, interview notes and a transcript into a client-ready CV and executive summary. You build it by describing the task to Claude, correcting its first mistakes and letting it write the instructions. If you are wondering how recruiters can get more from Claude, a skill is where I would start.

Should I use a Claude skill or a Claude project?
Use both, because they do different jobs. A skill holds the steps for work you repeat across clients, and a project holds everything about one client, such as terms and conditions, transcripts and the proposal. I activate the skill inside the project.

Is it safe to put candidate CVs into Claude?
Not as they arrive. Redact Social Security numbers, phone numbers and home addresses first, and tell candidates how you use AI. That applies to Claude and to every other generative AI tool, and I’m not a lawyer, so check the rules where you recruit.

What are Exa and Tavily, and how do they work with Claude?
Exa and Tavily are search tools you can connect to Claude so it can look things up on your behalf, and connecting one is a big part of how recruiters can get more from Claude. I use Exa for people research and Tavily for company and market intelligence, and both need an account before you connect them.

Is Copilot in Outlook really free?
For many people, yes. Microsoft says Copilot Chat in Outlook works without an add-on licence when you sign in with a work or school account, although in-app Copilot in Word, Excel and PowerPoint moved to licensed users from 15 April 2026.

How do I make Claude sound like me?
Give it real examples rather than a description of your style, because that is how recruiters can get more from Claude when it comes to writing. Transcripts of your calls, your LinkedIn posts and your recommendations let Claude build a voice skill that sounds like you, and you correct it from there. Switch memory on and fill in your personalisation settings as well.

Final Thought

So how you use your favourite LLM is completely up to you, but the sooner you realise that it is a lot more powerful than what you give it credit for, the sooner it can start building you bigger and better products. This is a mindset shift that only you can make.

Here is my question for you, whether you have been using Claude for a month or for two years: which of your repeatable tasks are you still doing by hand because you have never spent an afternoon teaching it who you are, and how much longer do you actually think that is going to work?

#upyourhuman(e)

Headliners Recruitment, a specialist marketing recruitment agency in Wokingham, was named first by all five AI tools we tested when asked for an alternative to the larger generalists. Director Clare Davies said: “Sam did a thorough audit and gave us quick wins plus medium-term changes that all felt achievable, even for a very small team.” 

We’re offering a free AI visibility insight for the first 15 agencies, plus 15% off your package with code TALENTHUNTER15. Register for your free insight by 30 November 2026:

About Vanessa Raath
Vanessa Raath is a global talent sourcing trainer, AI strategist, and founder of The Talent Hunter. Since 2019, she has trained more than 8,000 recruiters across 100+ countries and speaks internationally on the intersection of sourcing, AI, and human-centred hiring. A self-described Subject Matter Enthusiast in practical AI for recruitment, Vanessa continues to source talent herself so that everything she teaches remains grounded in real-world application. She has 590+ LinkedIn recommendations and a 4.9 out of 5 Trustpilot rating. #upyourhuman(e)
📍 South Africa | 🌐 vanessaraath.com

Want more like this?
Join my weekly newsletter for practical sourcing, AI, and recruitment insights delivered straight to your inbox. 👉 Subscribe on LinkedIn

🧠 How AI-fluent are you really? My free 7-minute assessment measures what you actually do with AI, not what you intend to do. You’ll get your AI Fluency Profile, a full category breakdown, and a benchmark against recruiters worldwide. Take it here.

💙 FREE Training | 💪🏼 Tried & Tested AI Tools | 🫂 Talent Hunters Community | 🤓 Online Academy | 📬 Van’s Substack | 😈 LinkedIn | 🔎 X | 📸 Instagram | 💃🏼 TikTok | 📽️ YouTube