Hiring in the age of AI

Your interview loop was built to measure effort. AI made effort free.

I rebuild hiring for technology and fintech companies — the process, and the profile you're actually hiring for. Because both changed, and most companies have only noticed the first one.

Book an intro call 2-minute diagnostic

CTO & co-founder, eXation.ai · Ex-Head of Data Science, Klarna
~100 data scientists hired in 3 years. None of them were ever let go.

How much signal does each stage still give you? 2019

Same loop. Seven years apart.

A process built when a polished CV and a clean take-home still told you something.

1 in 33 applicants reach an interview today. A decade ago it was 1 in 7. Industry hiring surveys, 2026

71% of engineering leaders say AI has made it harder to assess real technical skill. Karat survey of engineering leaders, 2026

22% of candidates already use AI tools during live interviews. Industry hiring surveys, 2026

26% of candidates trust AI to evaluate them fairly — while 87% of companies use it. Gartner · company figure, Disher Talent

The shift

Two things changed. Most companies have addressed neither.

Everyone noticed that candidates now use AI. Far fewer have noticed that this makes the job different — and therefore makes a different kind of person the right hire.

SHIFT ONE — THE PROCESS

Your filters stopped filtering

A candidate can generate twenty tailored CVs in an afternoon and produce a flawless take-home in twenty minutes. Every screening stage you built assumed effort was expensive. It isn't any more, so the stages that used to separate people now just pass everyone through.

The answer isn't banning AI from your process — it's unenforceable, and it's the wrong instinct. It's moving evaluation to the places where AI use reveals ability instead of hiding it.

SHIFT TWO — THE TARGET

You're hiring for the wrong profile

The skills that made someone a strong hire in 2019 are the ones now most easily automated. Meanwhile the traits that used to be nice-to-have — judgment under ambiguity, knowing when the output is wrong, taste — are the whole job.

Years of experience→Speed of learning
Recall of syntax→Quality of judgment
Producing the output→Auditing the output
Working alone, fast→Teaching the team what works
Follows the spec→Questions the spec

Everyone assumes AI makes soft skills less relevant. It's the opposite. When the technical work gets cheaper, what's left is communication — and whether someone brings the rest of the team along with what they've figured out. That's now the highest-leverage skill on an engineering team, and almost nobody is interviewing for it.
Elez Shenhar

What I do

Four engagements. Each ends with something your team uses on Monday.

No benchmark report. Every deliverable is built with your hiring managers in the room, so it reads like your company wrote it — because your company did.

00 — START HERE

Loop teardown

A week-long read on your current process. Low commitment, and it tells you whether the bigger work is worth doing.

You send your interview stages, rubrics and job specs
Two calls: one to understand context, one to walk you through findings
A written teardown naming what to cut, keep and rebuild

1 WEEK · $4,500 — CREDITED AGAINST AN AUDIT

01 — DIAGNOSE

AI-era hiring audit

Stage by stage: what still produces signal, what has been quietly zeroed out, and what it's costing you in bad hires and missed ones.

Teardown of your current loop with named owners per stage
Review of your last 20 hires — who worked out, and what the process actually predicted
Prioritised fixes and a business case for your exec team

3–4 WEEKS · $15,000

02 — REDESIGN

Interview loop rebuild

New stages designed for a world where candidates have AI — including formats where they're expected to use it, and you evaluate how.

Working sessions on novel problems, with rubrics for judgment not output
AI-pairing observation: how they prime the model, and how they verify what comes back
Debrief format that forces evidence instead of confidence

6–8 WEEKS · FROM $28,000

03 — REDEFINE

What good looks like, now

A hiring bar rewritten for AI-augmented teams: the competencies that predict performance when the routine work is already automated.

Competency set built from your own top performers, not a template
Role-family interview kits with questions and evidence to look for
Levelling guidance — what seniority means when output is cheap

4–6 WEEKS · FROM $18,000

04 — ADVISE

Leadership advisory retainer

A standing line to someone who has run these teams, for founders, CTOs and first-time heads of function.

Org design and team structure for AI-augmented functions
Hiring plans, levelling, and comp conversations
Available after an audit or redesign engagement

$5,000/MO · MIN 3 MONTHS · POST-ENGAGEMENT ONLY

How engagements run

Always in this order. Diagnosis before treatment.

I don't sell a redesign before I know what's broken, and I don't take a retainer with someone I haven't worked with. It protects your budget and my judgment.

Intro call30 MIN · FREE

You describe what's going wrong. I tell you whether it's a problem I'm the right person for — and if it isn't, who is.

Teardown1 WEEK · $4,500

A paid first look at your loop, small enough to approve without a procurement cycle. If you go on to the audit, the fee comes off it.

Audit3–4 WEEKS

Interviews with your hiring managers, recruiters and recent hires. Funnel and outcome data. Ends with findings presented to your leadership team — by both of us.

Rebuild4–8 WEEKS

We fix the highest-value thing the audit found. Your managers co-build it, run it under observation, and get coached on the first real loops.

Hand overONGOING, OPTIONAL

Train-the-trainer so it survives without me, then a light retainer if you want a sounding board. The goal is that you stop needing one.

Free · No email required

Is your loop still measuring anything?

Five plain questions about the process you're running right now. The read at the end matters more than the score.

AI-era hiring check

0 / 5 ANSWERED

Answer all five to see your read.Nothing is stored or sent anywhere.

Talk it through

Proof

The last time I owned this problem.

Three years running data science hiring at Klarna, at the scale where a process either holds or quietly falls apart.

2,000+Interviews conducted — personality, technical, skills, take-homes — across multiple organisations

~100Data scientists hired at Klarna over three years

0Of them ever let go

7Teams led, 30+ people, Payment Methods domain

Who you'd be working with

I ran the hiring I'm now telling you to change.

Elez Shenhar · Founder

NowCTO & co-founder — eXation.ai
WasHead of Data Science, Payment Methods — Klarna
Led30+ people across 7 teams
Hired~100 data scientists over 3 years
Interviewed2,000+ candidates across multiple organisations
WasHead of Machine Learning — MotionTag, Berlin
Built inData science, ML, quantitative research, software
BasedPortugal · works remotely, globally

I spent 17 years building data and ML teams — most recently as Head of Data Science at Klarna, where I also owned hiring. Now I fix hiring processes that AI has quietly made useless.

At Klarna I led 30 people across seven teams in the Payment Methods domain, and hired around a hundred data scientists over three years. I inherited the interview process and rebuilt it around one idea: put candidates at ease early, so you see who they actually are instead of who they perform being under pressure. It worked better than I expected. Not one person hired through that pipeline was ever let go.

The hire I'm proudest of had no degree at all. He was talented and he wanted it, and every conventional filter would have removed him before anyone spoke to him. That's the part most companies still get wrong, and AI has made it worse — the credentials and the polished application are now the cheapest things a candidate can produce, and they're still what most loops screen on.

The belief I hold that most of my peers don't: AI is making soft skills more important, not less. When the technical work gets cheaper, what remains is judgment, communication, and whether someone teaches the rest of the team what they've figured out. Almost nobody is interviewing for that.

Before Klarna I was Head of Machine Learning at MotionTag in Berlin, working on mobility data for European transport operators, and before that I worked in quantitative research and software. I left because I was enjoying the hiring work more than anything else I was doing, and I wanted to reproduce that result at more than one company. I’m also CTO and co-founder of eXation.ai, where we build generative-AI systems for other companies — which means I’m not theorising about what AI does to technical work. I watch it happen to my own team every week. I work alone on the advisory side — you get me personally on your calls, not an associate — and I hold to three engagements at a time so that stays true.

Fit

Worth being blunt about this.

We'll work well together if…

You're hiring technical or analytical people and the last few hires have surprised you
Someone senior owns this and can decide without a committee
You want your team running it afterwards, not depending on me
You're open to hearing that the problem is your criteria, not your funnel
You'll give me access to real data and real people, not a curated sample

We won't if…

You need someone to fill roles — I'm not a recruiter and don't place candidates
You want to buy an AI screening tool and call it done
The goal is to stop candidates using AI rather than to evaluate how they use it
You want a benchmark report for a shelf
Nobody on the exec team thinks hiring is their problem

Writing

How I think, before you pay to find out.

EVALUATION

Watch how they verify, not what they produce

Anyone can get a good answer out of a model now. The signal is in what happens next: whether they build a check that fits the specific task, or trust the output because it reads well.

THE PROCESS

After the intro call, the CV is noise

A CV is worth one conversation about someone’s history. Past that it tells you nothing AI can’t manufacture — so evaluate the work you watch them do, and stop letting paper into the room.

METHOD

The first five minutes decide the interview

Ask about the thing they actually love doing. A candidate who is at ease thinks clearly for the next two hours — and you spend those hours assessing ability instead of nerves.

Start here

Tell me what's going wrong.

A 30-minute call, no pitch. If I'm not the right fit I'll say so on the call and point you somewhere better.

01You send a few lines about the situation
02We talk for 30 minutes, within 2 working days
03You get a written note on what I'd do and what it costs — free either way

Or email directly: elez@shenharconsulting.com · LinkedIn

Name

Work email

Company & headcount

What's going wrong?

Goes straight to my inbox. No newsletter, no sequence.

Legal

Who is responsible

Shenhar Consulting & Holding, Inc. (Delaware, USA) is the controller of personal data collected through this site. Contact: elez@shenharconsulting.com.

What this site collects

The contact form. Your name, work email address, company and headcount, and whatever you write in the message field. Nothing on the form is required except a name and an email address.
Email you send directly. Held in the same way as a form enquiry.
Server logs. The hosting provider records standard technical data, including IP addresses and browser type, to run and secure the site.

There are no analytics, no advertising tags, no tracking pixels, and no cookies set by this site. Fonts are served from this site itself, so no third party sees your visit.

The diagnostic tool

Answers to the hiring diagnostic are processed entirely in your browser. They are never transmitted, recorded, or seen by anyone. Closing the page discards them.

Why this data is used

Solely to reply to you and to discuss possible work together — steps taken at your request before entering a contract, and a legitimate interest in responding to business enquiries. Your details are never sold, rented, or used for marketing. There is no newsletter and no automated sequence.

How long it is kept

Enquiries that do not lead to work are deleted within 24 months. Records relating to actual engagements are kept as long as required for contractual, accounting, and tax purposes.

Who else sees it

Only the providers needed to operate the business — email, hosting, and the form service — acting on instruction. Data is handled in the United States, where this company is established.

Your rights

If you are in the EU, EEA, or UK you may request access to your data, correction, deletion, restriction or portability, and you may object to its use. Write to the address above and you will get a reply within 30 days. You may also complain to your national supervisory authority — in Portugal, the Comissão Nacional de Proteção de Dados.

Children

This site is aimed at businesses and is not directed at anyone under 16.

Last updated 24 August 2026. Material changes will be posted here.