Starting with AI, in plain English.

A guide for recruitment agency owners who keep hearing about AI and want to know what it would actually do on their desk. No jargon. Ten minutes to read.

What AI is, for a recruiter.

The AI people talk about is an assistant you write to in plain English. ChatGPT is the best known. We use Claude, made by a company called Anthropic. You type what you want, the way you'd ask a capable new starter, and it has a go, which you check.

On its own, it's good with words. It reads a CV in seconds, drafts an ad, summarises a long email chain, or turns your rough notes into a tidy candidate summary.

What it can't do on its own is see your database. So the useful half of the job stops at copy and paste, and your consultant goes back to clicking through JobAdder for the rest.

That's the gap we close. Once Claude is connected to JobAdder, the same conversation that reads the CV can find matching candidates, add the note and move the application along, all in your system and under the consultant's own name.

Three ways to use it, from easiest to most useful.

It's easy to stop at the first. The real time savings are at the second and third.

  1. Claude on its own

    Writing ads, summarising CVs, drafting emails and LinkedIn messages. Useful straight away, but everything is copied in and out by hand, and it knows nothing about your candidates or clients.

    Good for: writing. Anyone can start today.

  2. Claude connected to JobAdder

    Now it can search your candidates, read their notes, build shortlists, log calls and set reminders, all inside JobAdder. Each person signs in as themselves and sees only what they already could.

    Good for: the daily grind of searching, updating and keeping records. This is what NativeMCP does.

  3. Routines for the repeat jobs

    The things your people do every day the same way, like building a call list, get saved as a routine. One question does what used to take a morning.

    Good for: the biggest time savings. At one agency, a call list went from about 1,000 lookups to one request.

What you need.

Less than most people expect. No new software to install, and nothing changes in how JobAdder works.

  • JobAdderOur connector is built for it. If you use something else, ask us what's possible.
  • A Claude Team plan per personPaid to Anthropic. We recommend Team because Anthropic doesn't train its models on Team data.
  • Someone who owns itUsually the owner or an operations lead. They make the calls on what to switch on and what to fix first.
  • A little of everyone's timeA short session with each person on their real work. After that, they learn by using it.

Your first month with us.

A typical plan. Yours follows your team's pace. Or skip ahead with a free 14-day trial of the connector on your own data.

Try it free

  1. We look

    Short sessions with a consultant, an owner and whoever runs the back office. We look through your database and put real numbers on what's slowing your team down.

  2. We connect and teach

    Claude goes into your JobAdder. Each person connects once with their own login. We train them on their real work and leave a handbook written in your agency's words.

  3. We measure

    Your first weekly report: who's using it, for what, what got easier and what didn't. If someone's stuck, we help them.

  4. We take the next job off your plate

    We turn the most repetitive work into routines, and fix anything in the database that's making searches worse.

What to watch out for.

AI is useful, not magic. These are the things we teach every team.

It can be wrong

Claude can misread something or sound more certain than it should. Check anything that goes to a client or candidate before it's sent, the same as you would with a new starter's work.

Don't paste candidates into free AI tools

Free and personal AI plans may use what you type to train their models. Candidate details belong on a business plan like Claude Team, connected properly, under your own settings. That's how we set it up.

A messy database gives messy answers

If every candidate is marked Active, AI will happily hand you people who retired years ago. Part of our work is fixing the rules underneath, so the answers improve.

Tools nobody uses cost money

The risk isn't that AI fails. It's that people try it once and drift back to the old way. That's why we measure every week and say so when adoption slips.