Use cases · A recruiter
Candidate search
A recruiter hands over a job description. Assistants find the people, in rounds, and the recruiter decides who is right.
The situation
Maya is a recruiter at a small agency, with six roles open at once. Each needs the same careful work: read the job description, turn it into searches, go through hundreds of profiles, and bring the client a shortlist worth their time. Her agency has written down, once, how it runs a candidate search, and every recruiter’s assistant follows it.
Handing it over
She opens her assistant and pastes the job description. She doesn’t say how to search; that is already written down.
You: Here’s the JD for the integration architect role at Harbourline. Start a LinkedIn search and find me ten good people.
Assistant: I’ve written a brief from it: five must-haves, three nice-to-haves, and two deal-breakers. Before I start: it says “MuleSoft or similar”. Is Boomi experience enough, or do they need MuleSoft itself?
You: Boomi is fine. They do need to have led a team.
Assistant: Added to the must-haves. I’m running six different searches and will stop when I have ten strong matches, or have used this round’s effort. Then you’ll get a shortlist to review.
It asked about the one thing the job description left open, and her answer went straight into the brief. Then it started.
While she does something else
Her assistant runs a page of each search and checks every result against the brief as it reads it. The plausible ones get a full reading, criterion by criterion. Searches that bring back nothing useful are dropped and replaced with better ones. Several assistants can do the reading at once, each taking the next profile nobody else has, so a long list goes quickly, and nobody is read twice.
None of this happens in a chat window she has to keep open. It carries on until the round has what it needs.
Maya decides
When the round stops, her assistant opens the review page: every match, strongest first, with a summary, any concerns, and the evidence for each requirement.
Round 1: ten matches to review
Ten matches from six searches, four of them strong. Two searches brought back nobody useful and were replaced.
- Priya RamanStrong
Led a team of six through a Boomi-to-cloud migration. Every must-have met.
- Tom KellerPossible
Deep MuleSoft work. Concern: has mentored, but not formally led a team.
- Ana SousaStrong
Owns integration architecture at a logistics firm.
Maya’s note for the next round: “Too many consultancy people. They need to have led a team themselves, not just mentored.”
She approves four, rejects two with a word on why, and leaves that note. Each decision is saved as she makes it. The note goes into the brief, the searches are adjusted, and the next round starts from there. The assistants propose; only Maya decides that the brief changes, that another round runs, or that the search is done.
The next morning
On the train, in a new conversation on her phone, she asks one question.
Next morning, a new chat, on a phone
You: How’s my search going?
Assistant: The Harbourline search is in round two. Seven of ten have matched so far, three of them strong. Nothing is waiting on you; I’ll open the review page when the round stops.
It knew which search she meant, because it is the one open in her space, and it knew where the work had got to, because the work was never in the chat.
The whole agency, one way of working
Every recruiter at the agency uses the same instructions, so every search is run the same way, and improving the instructions improves everyone’s searches. Each recruiter’s candidates stay in their own space, and Ruth, who runs the agency, sees how every search stands without asking anyone for an update:
Fourteen open searches
Five shortlists went to clients this week. One search hasn’t moved since Monday, and its recruiter’s assistant has a suggestion.
- Integration architect · Harbourline7 of 10
- Data platform lead · Calder BankWith client
- Cloud engineers ×3 · FerroStuck since Mon
The searches find almost nobody in Perth. Priya’s assistant suggests widening to remote.
- Head of security · LumenClosing
Better every week
Each session ends with a short note on how the instructions held up: what was awkward, what failed, what should change. Early on, one note spotted that very long keyword searches quietly return nothing. The instructions now keep searches short, and every session since has started out knowing it. Someone reads the notes and folds them in, and that someone can be an assistant too.
More stories
- A software team Every assistant knows what is waiting for its person. People decide what gets built, how it is split, and when it is done.
- What to post about The assistant keeps track of the topics in their area and what they have posted on, and drafts on what is worth posting today. They approve, edit or discard.
- Relationships at scale Says who counts. The assistant finds the people, writes each a personal note, and keeps track of every conversation. Nothing is sent without them.
- Looking for a job Says once what they want. The assistant finds and screens openings, drafts what each one asks for, and applies to the ones they approve.
- Watching competitors Names the products and the competitors. The assistant matches them across sites, checks every week, and shows what changed.
- Every page, reviewed Sets the goal and the rules once. The assistant goes over every page, records what it found and fixed, and brings back only what needs them.