Introducing
Simi
SimiYour AI copilot for recruitment
Not a chatbot trained on help articles. Simi works inside the job — sourcing from your own database, writing the ideal candidate profile, scoring against it once you approve, moving the pipeline, publishing the role and editing the fee. You brief it. It asks before it acts.
You brief it. It asks before it acts.
Screening does not begin with a score. It begins with an Ideal Candidate Profile you can read, argue with and approve — so the shortlist is something you can defend to a client rather than a number you have to trust.
Find me senior Java devs who have done payments
Searched your database with AI Match rather than keywords, so people who describe the work differently still came back. Want me to score them against the role?
Yes — weight payments harder and make SC clearance a deal-breaker
Here is the Ideal Candidate Profile I would score against.
Ideal Candidate Profile · draft
- Must have: commercial payments experience
- Must have: active SC clearance
- Weighted up: payments over general backend
- Nice to have: regulated-environment delivery
Approve it, or tell me what to change.
Approve. Then add the top five to Shortlist
Scored and ranked against the approved profile, with the reasoning on each candidate. Top five moved to Shortlist.
Illustrative exchange. No boolean strings, no filters to configure, nothing typed twice.
What it does inside a job
Say the sentence. It does the thing.
Sourcing
“Find me senior Java devs”
Find candidates in your own database
AI Match runs a semantic search across your database, not a keyword filter, so the person whose CV never uses your phrasing still surfaces. Ask for “senior Java devs who have done payments”, or pull back what a previous run found without re-running it.
Screening
“Who’s the best fit?”
It writes the Ideal Candidate Profile first
Screening does not start with a score. Simi drafts an Ideal Candidate Profile for the role, shows it to you, and waits. Nothing is scored against criteria you have not seen.
Screening
“Add SC clearance as a deal-breaker”
And you argue with it in plain English
Refine the profile the way you would brief a resourcer — more weight on payments experience, clearance as a deal-breaker, drop the degree requirement — and it regenerates. Approve when it reads like the role you are actually working.
Screening
“Rank them”
Then it scores every candidate against it
Each candidate is scored and ranked against the profile you approved, with the reasoning attached, so a shortlist is something you can defend to a client rather than a number you have to trust.
Pipeline
“Add the top 5 to Sourced”
Move the pipeline by saying so
Add to a stage, shortlist, reject in bulk, promote the top few. The stages are your own workflow, so the instruction is the same sentence you would say out loud.
Publishing
“Is it live anywhere?”
Publish it, or take it down
Push the job to your careers site, the candidate portal or the boards — and ask where it is currently live. When the instruction is ambiguous it asks which channel rather than guessing.
The commercials
“Fee 20%, 90-day guarantee, staged 30/30/40”
Edit any field on the job, including the money
Description, skills, benefits, seniority, pay, openings, location, dates, sectors, client brief, hiring manager — and the commercials: fee percentage or fixed, currency, engagement type, guarantee period, staged and retainer payment plans. Several changes in one sentence, resolved against your own lists.
Comms
“Send that to the hiring manager”
Send the email without leaving the thread
Draft it, read it, send it from the job — and it lands on the record like every other message, so the next person to open the candidate sees it.
The lavender tiles are one loop: it drafts the profile, you refine it, it scores against what you approved.
Why this is not a support widget
| A generic assistant | Simi | |
|---|---|---|
| Answers from | Help articles and documentation | Your candidates, jobs, pipeline — and the docs when you want the docs |
| Knows about | Product features and pricing | Your account: your plan, your limits, your credit, your usage |
| Can take action | Open a support ticket | Source, screen, score, move the pipeline, publish, edit the job, send the email |
| Before it acts | Nothing to approve | Shows you the Ideal Candidate Profile and waits for your approval |
| When it cannot help | Repeats the same article | Hands you to a human, with the conversation intact |
Six places you will meet it
Inside the job
The copilot with the tools. Open any job and it already has the spec, the pipeline and the candidates in front of it — so “who’s the best fit” needs no further explanation.
The help chat, everywhere else
Ask how something works and it answers from the product documentation. Ask about your own account and it answers from your account — your plan, your limits, your remaining credit — rather than sending you to a pricing page.
A human, when you want one
Say so and it hands the conversation to a person with the whole thread attached. You never repeat yourself to the second responder.
From your very first minute
The same assistant walks you through signup — finding your company, verifying you, planning the migration — so the first thing you meet is the thing you will keep using.
In the browser, on LinkedIn
The Chrome extension carries it onto profiles you are already reading, with your CRM context and matching against your live jobs.
Through Claude, over MCP
Connect your CRM to Claude and ask in natural language. Read-only, OAuth 2.1, free on every plan.
It is only useful once it knows your data
Which is the whole point, and the reason to bring your database in first. Free plan, no card.