How Navigent works

From job ad to contacted shortlist, step by step. You paste the ad, Navigent finds and ranks the candidates and shows the evidence for every requirement, and you decide who goes forward. Only then does Navigent write to them, from your own LinkedIn and email.

The example all the way through is a senior controller in Copenhagen.

1. You paste the ad

Everything starts in one box. Paste a job ad as text, a link to the ad, or a file, PDF or Word. You can also paste a LinkedIn profile to find people like it, or simply describe who you're looking for. You don't need to clean up the text, and there's room for up to 20,000 characters.

As you type, five ticks show what Navigent can read: function, seniority, location, industry and years of experience. Navigent fetches a link itself, and reads a file once. The file isn't stored, only the text from it, and if you use a LinkedIn profile, neither the link nor the profile is stored.

2. Navigent reads the requirements

From the ad, Navigent writes down the requirements, each as a must-have or a nice-to-have:

• Must-haves that search data can show become filters: function, seniority, location and years of experience.
• Nice-to-haves such as industry or type of company decide the order and never filter anyone out.
• What search data can't answer, such as skills, languages and education, is set apart as something to clarify.

Navigent looks up the titles in Danish and English, so a senior controller is also found as a financial controller or a business controller. If the ad names the company, its current employees are always excluded, and the thread says so in one line.

You can see what Navigent is doing as it happens: a few steps with a timer, folded into one line when they're done. The panel on the right always shows where the role stands: the plan in four steps, the talent pool with its numbers, your searches, and a map of where the candidates live.

3. The free preview

Before anything costs you, Navigent shows how many people in the market fit the filters. These are the role's matches, always shown as an approximate number: "I've set these filters · about 1,243 matches".

The filters and criteria sit as chips in the thread. Remove a filter, and you get a new preview; remove a criterion, and the ranking updates. Both are free, and every change is noted in the thread, so you can always see what was adjusted.

With fewer than about 50 matches, Navigent widens the search by itself: from the city to the region, then to the whole country, and finally with more titles. Every step is shown in the thread.

4. The questions

If something is unclear, Navigent asks up to three questions before the search starts, for example: "71% of the matches work at audit and advisory firms. How should they be treated?" Each answer is written as what it does to your results, "Keep them, but rank them lower", and you choose with the keys 1 to 4. There's always "Other…" and "Skip".

Your answers adjust the filters and the requirements, and afterwards you can see every question and answer in one place.

5. The fetch

When the filters are right, you press Start search. The button tells you what it costs: "Start search · uses 1 of your 10 searches". Only now is a search used.

Navigent fetches about 1,000 unique candidates. Duplicates are removed before they're counted, and when the market has more, Navigent fetches new ones in their place, at its own cost. When the market has fewer, you get what there is. In the example it was 986: 257 duplicates were removed, and the market had no more.

You follow the fetch in the thread and in the panel, "Fetched 640 of about 1,000", and you can stop along the way and keep what's been fetched. If something goes wrong, you carry on where it stopped at no extra cost, and if the search stores no candidates, you get it back.

People who have asked to be removed from Navigent are never fetched, and neither are the people and companies on your do-not-contact list.

6. The gates

Every fetched candidate passes four gates. They remove the people who clearly don't fit, always with a tolerance, so no good candidate drops out on a technicality:

If a piece of information is missing, the candidate gets through. Missing data never filters anyone out. Industry and nice-to-haves only decide the order. In the example the gates stopped 52, and 934 went through.

Those stopped are listed under Excluded, with the reason, and you can bring one back in.

7. Calibration

While the rest are being fetched, Navigent shows three candidates in the thread: the highest ranked and two who differ from her in seniority, industry or type of company. You say yes or no to each, just as in the review: a yes puts the candidate on the shortlist, and a no rejects them. So you never decide on anyone twice.

Your answers become the examples every candidate is scored against again, so the right people rise to the top. Say yes to two of the three, or press "Looks right", and calibration is done. Otherwise a new round follows, three at most, and after a round without a single yes Navigent first suggests one change. You can also skip calibration.

8. The ranking and the longlist

Everyone who got through the gates is scored from 0 to 100 on how well their history covers each requirement: titles, employers, tenure, seniority and industry. The score gives a band: Strong from 75, Good from 55 and Possible below 55.

About 200 of the best become your longlist. When several are tied at the cut, they all come along, so the longlist is rarely a round number. In the example it was 207. The rest stay in the pool and can still be seen.

The AI model that scores never sees names, photos, email addresses or phone numbers, only the career. Gender, age, ethnicity and other sensitive traits are never guessed or used, and years of experience are only compared with what the role asks for.

9. The recommended

At the top of the longlist are the recommended, up to about 50. They're the ones who genuinely fit: a score of at least 55 and a clear match on the function, so nobody is recommended on nice-to-haves alone. It's never a number to fill. If 23 fit, 23 are recommended, and if none do, the thread says so and offers the longlist instead.

The recommended are assessed requirement by requirement first, and the thread tells you when they're ready: "41 recommended ready for review". You can have the rest of the longlist assessed 50 at a time, free.

10. The review

Every candidate you review has a card with two sides.

• On the left: name, title, employer, town and distance, total experience, time in the current job and average per job, signals such as Promoted and Long tenures, and the work history, job by job.
• On the right: the band, a two-sentence summary, one row per requirement with Met or Missing and one sentence citing the jobs it rests on, what's left to clarify, and strengths and concerns.

Hover over a citation, and the job lights up in the history. Hover over the band to see the score. The card also says what data it rests on, and when it was fetched.

You decide from the keyboard:

On your phone, swipe right for yes and left for no. A person makes every decision, and every decision is saved.

If the workspace has photos turned on, the card shows the candidate's profile photo. Photos are never sent to an AI model, and if you turn them off, they're neither shown nor fetched.

11. The shortlist

The shortlist is only the people you've said yes to. Nothing else is called shortlist.

In the list view, every tab has its count, the same number as in the talent pool: Recommended, Longlist, Shortlist and Rejected. You can sort by rank, distance or experience, filter by band, town or signal, and decide on many at once. A decision can always be changed. The longlist and the shortlist can be exported as CSV, and every export is logged.

What's left to clarify, you check on the candidate's LinkedIn profile once she's on the shortlist, or with the Chrome extension.

12. Campaigns

When the shortlist is ready, Navigent can write to the people you've said yes to, and only them. You add candidates to a campaign from the shortlist, one or many at a time, and a campaign can take new ones along as you say yes. A campaign belongs to one role and sends from one person's LinkedIn and email all the way, so the candidate hears from the same person throughout.

The sequence is a flow:

• an invitation on LinkedIn, without a note by default,
• a message once the invitation is accepted,
• a follow-up if the candidate doesn't reply,
• and an email from your own Gmail or Outlook if the invitation isn't accepted.

Every step has its own draft. You insert variables such as first name and title, and an AI sentence on why this particular person fits, built from the assessment and the ad and written in the ad's language. Every message can be read before it's sent, and if the candidate replies, the rest stops at once. A note on the invitation can be turned on per campaign, at most 200 characters, LinkedIn's limit.

Navigent sends on weekdays between 8:00 and 16:00 in the sender's time zone, with careful limits: 15 invitations a week to start with, rising with the acceptance rate to at most 80. There's no open or click tracking. People and companies on your do-not-contact list are never contacted, and anyone who replies that they don't want to be contacted is added to it automatically.

A free LinkedIn account can run the whole sequence when the invitation goes without a note. For a note on every invitation, LinkedIn requires Premium, Sales Navigator or Recruiter.

13. The inbox

Every user connects their own LinkedIn and their own Gmail or Outlook, once, through a secure sign-in. Navigent never sees the password.

Every reply lands in one inbox for the workspace, LinkedIn and email side by side, and you can filter by account, by campaign or to your own accounts. Navigent sorts the replies into interested, not interested and auto-reply, so you know who to talk to first. The label only sorts: it's the reply that stops the sequence, never the label. Anything waiting for a person is collected under Needs action.

You reply straight from the inbox, and your answer goes through the same account. Only conversations with candidates in your campaigns are stored. Nothing else in your inboxes ever reaches Navigent.

14. The Chrome extension

On a LinkedIn profile, click Navigent in Chrome or press Alt+Shift+N, and the extension shows Navigent's assessment of the profile against one of your roles, in Chrome's side panel. Because it sees the whole profile, it can also answer what was left to clarify, such as skills, languages and education. The assessment costs no search.

With "Add to campaign" you say yes to the candidate and put her in a campaign's queue, so she is contacted from your own LinkedIn when it's her turn.

The extension reads only the profile you have open, and only when you click. It never navigates, scrolls, clicks or sends anything itself. You sign in to it with your Navigent account in one click.

You can always adjust

The chat stays open on the role. Write "prefer food industry experience", for example, and Navigent changes the weighting and ranks again, free. Write "also Aarhus", and you get a new free preview. If something needs fetching again, Navigent asks first, because it uses a search.

With fewer than 50 candidates left to review, you can fetch about 1,000 more for the same role. It costs one more search, and Navigent asks first: "This uses 1 more search on Senior controller. You have 7 left."

Together as a team

Roles are shared with everyone in the workspace, and every role has an owner. You write notes on candidates and mention each other with @, and the activity feed shows what has happened. Analytics show the numbers per role, per campaign and per person. Everyone draws on the same pools of searches and email addresses.

Users are owners, admins or members, with access to every workspace or to selected ones.

The words

Every word has one meaning, in Navigent and on this page.

Get access when we open

Navigent opens soon. When it does, you get 14 days free with 2 searches and 50 email addresses, and you pay nothing until the trial ends.

Get early access · See pricing