H
H·
Higher
Demo Workspace
ML
M. Lindqvist
New Role · Step 1 of 3

Define the role profile.

Set what matters most for this hire. Higher weighs every candidate against this profile and explains every match.

1 Role basics

What are you hiring for?

The role title, department, and seniority frame the entire pipeline.

2 Required skills & competencies

Mandatory hard requirements.

Candidates without these are auto-classified as Category C.

Python × PyTorch × MLOps × PhD or equivalent × 5+ years production ML ×
Desirable
Causal inference × Distributed systems × Published research × JAX / TensorFlow ×
Languages
English
Swedish
German
French
+ 137 more
3 Match weighting

How should Higher weigh each dimension?

The weighting determines how the final match score is composed. Drag to adjust — the AI explains every decision based on these weights.

Hard skills & experience
40%
AI interview signal
25%
Assessment results
20%
Personality & culture fit
15%
Summary · Senior AI & Data Scientist

Higher will start screening as soon as the first application arrives.

PipelineAuto · 24/7
LanguagesEN · SV · + 2
Mandatory skills5
Desirable skills4
Estimated category splitA: 8% · B: 24% · C: 68%
i
Your team gets the dashboard the moment the first qualified candidate (Category A) is identified. No daily check-ins needed.
H·
Higher
Demo Workspace
ML
M. Lindqvist
Senior AI & Data Scientist · 247 candidates

Today.

Higher narrowed 247 applications down to four people worth your time. One is ready for your decision now.

Ready for your decision
SC
Sarah Chen
Stockholm · Advanced AI & Data Science · 4 weeks notice
96%
Match

Sarah completed her assessment last night. She's the strongest match in your pipeline — verified credentials, top-percentile structured reasoning, and clear cultural alignment with the team you described.

Hard skills
98
Personality
92
Culture fit
95
Documents
Assessment completed 14 hours ago
Up next in your queue 3 candidates
MR
Marcus Roussel
Interview scheduled · Wed 14:00
91%
AB
Anna Bergström
Assessment in progress
89%
LM
Léa Martin
Documents under review
88%
247 applied
21 strong match
3.2h avg. triage
99.2% documents verified
Show full pipeline
Total candidates
247
+18 this week
Category A · strong
21/247
8.5% match rate
Avg. time to category
3.2h
vs 9 days manual
Document verification
99.2%
3 flagged for review
Filter
All categories
Category A
Category B
Score 90+
Salary < 50k
Available now
Add filter
Reset
Category A · Strong recommendation
4 candidates Surface to leadership for final human interview
Candidate Match Score Salary Experience
LJ
Lania Johansson
Göteborg · 6 weeks notice
Advanced AI & Robotics
94
€48,000
11 yrs
NK
Niclas Karlsson
Malmö · Available immediately
Senior ML Engineer
93
€44,000
9 yrs
ZW
Zoé Werthen
Stockholm · 8 weeks notice
Advanced AI & Vision
91
€42,000
8 yrs
Category B · Possible candidate
17 candidates Strong signal, gap on one or two dimensions
DL
David Lindberg
Uppsala · 2 weeks notice
AI Research Engineer
85
€38,000
7 yrs
EV
Elin Vendel
Stockholm · Available immediately
Product Manager · ML
82
€40,000
8 yrs
MD
Michael Davies
London · 12 weeks notice
ML Platform Architect
81
€52,000
14 yrs
Category C · Not recommended
226 candidates Filtered out — gap on mandatory requirements
PB
Per Bergström
Lund · Self-employed
General Software Engineer
62
€35,000
5 yrs
H·
Higher
Demo Workspace
ML
M. Lindqvist
Back to candidates
AI synthesis · 1-page summary

Sarah is a senior practitioner with twelve years shipping production ML — most recently leading a four-person team building causal inference systems at a Stockholm fintech. Her hard skills are in the top 2% of all candidates Higher has assessed for similar roles. She communicates with unusual clarity and demonstrates strong leadership through her interview responses.

Synthesized from 7 documents, 42-minute interview, 3 assessments
96%
Match · Senior AI & Data Scientist
Strengths
  • + 12 years production ML, with verified output at three companies including a high-growth fintech.
  • + PhD from KTH Royal Institute · publications match the role's research dimension.
  • + Interview shows high communication clarity and concrete examples of cross-functional leadership.
  • + Available within 4 weeks · salary expectation within budget.
Risks & questions
  • ! Limited distributed systems exposure — a desirable, not mandatory, requirement.
  • ! One reference contact pending response. Two have already replied positively.
  • ! Has expressed preference for hybrid (3 days office) — confirm with our policy.
PhD Thesis · Causal Inference
KTH · 2018 · 247 pp
✓ Verified · 99.9%
MSc Computer Science
Uppsala · 2013
✓ Verified · 99.7%
Reference · CTO Klarna
Direct verification · Dec '25
✓ Replied positive
Code samples · GitHub
3 repos · 14k stars
✓ Authenticated
AI Interview · 42 minutes

Excerpt — leadership & problem-solving

Highlights
Full transcript
Audio
Question 4 of 12 · Behavioral · Leadership
"Tell me about a time you led a team through a technical decision where the team disagreed with your initial recommendation."
"We were choosing between two architectures for a customer churn model — a simpler logistic baseline vs. a gradient-boosted ensemble. I initially pushed for the ensemble. Two of my engineers pushed back: they argued the baseline would ship in two weeks vs. eight, and we needed signal in production fast to prove the project's value. I asked them to walk me through their reasoning, then I changed my mind. We shipped the baseline first, learned the actual feature importance was very different from what we'd assumed, then built the ensemble against real production signal. The final model was better than what I'd originally proposed because it was informed by data we wouldn't have had otherwise."
+ Demonstrates intellectual humility
Strong outcome reasoning
Speech clarity
9.4 / 10
Confidence
8.8 / 10
Structured reasoning
9.1 / 10
Facial-recognition analysis intentionally disabled. Higher ethics policy.
Behavioral assessment · Big Five framework

Personality vs. role profile

Anti-distortion checks passed. Social desirability score: low (good).

Full assessment
Drive Collaboration Stress tolerance Decision making Consistency
Drive 9.2
Collaboration 9.1
Stress tolerance 8.7
Decision making 8.9
Consistency 7.6
Dashed line · target profile for role
H·
Higher
Demo Workspace
ML
M. Lindqvist
Back to Sarah Chen
Authenticity engine · Forensic verification

Sarah Chen's documents — the system shows its work.

Every document parsed, cross-referenced against issuing institutions, and analysed for layout and signature anomalies. No black-box judgements — every score has a justification a human can audit.

7 documents · 1 flagged Filter
PhD Thesis · Causal Inference
KTH · 2018 · 247 pp · 99.9%
MSc Computer Science
Uppsala · 2013 · 99.7%
BSc Mathematics
Stockholm Uni. · 2011 · 99.5%
AWS Solutions Architect
Direct API · 100%
Reference · CTO Klarna
Direct verification · Replied
Code samples · GitHub
3 repos · 14k stars · Authenticated
Course cert · Nordic AI Institute
Medium auth · 78% · Flagged
Verification summary
Verified6 of 7
Flagged for review1
Avg. authenticity96.7%
Time to verify3m 14s
Forensic report · Generated 14 minutes ago

PhD Thesis — Causal Inference for High-Dimensional Time Series

KTH Royal Institute of Technology · Defended 11 June 2018 · Supervisor: Prof. M. Berg

99.9
Authentic
OCR & text extraction
3 / 3 passed
Optical character recognition
247 pages extracted · No degraded regions detected
99.7% confidence
Language detection
Primary: English · Secondary: Swedish (abstract, references)
2 languages
Structural parsing
Title page, abstract, 7 chapters, references, appendices identified
All sections
Document forensics
4 / 4 passed
Cryptographic signature
KTH digital seal · Issued 2018-06-14 · Chain valid
SHA-256 verified
Embedded watermarks
3 invisible watermarks detected · All match KTH 2018 cohort
3 / 3 authentic
Layout pattern match
Font, margins, header style consistent with KTH 2018 dissertation template
98.4% match
PDF metadata
Creation date, author, modification history all internally consistent
No anomalies
External cross-references
2 sources · matched
KTH public registry
Direct API · Defended 2018-06-11 · Award confirmed
Match
Google Scholar publication record
14 papers · 2,847 citations · Author identity confirmed
Match
Plagiarism & pattern analysis
3 / 3 passed
External database match
Compared against arXiv, IEEE, ACM, Springer · 12.8M documents
0% match
Internal coherence
Single-author writing pattern across all chapters · No style drift
Consistent
AI-generated text probability
Stylometric analysis · Predates current LLM era
2.3% (low)
Algorithmic audit log entry · 2026-04-30 14:22:07 UTC
SHA-256 · a4f2b9e1c8d7…f3a82e1bc44d
Copy entry hash
H·
Higher
Demo Workspace
ML
M. Lindqvist
Back to Sarah Chen
AI interview · 42 minutes · 12 questions

Interview playback — Sarah Chen.

The AI interviewer adapted its follow-up questions in real time. Every answer was scored against the role's competencies, with direct quotes pulled to justify the rating.

Questions All competencies
1
Walk me through your last six months of work.
Opener · context
00:00 – 04:18
2
Most technically demanding decision you've owned.
Technical · ownership
04:19 – 09:42
3
Time you misjudged something important.
Self-awareness
09:43 – 14:11
4
Leading a team through disagreement on a technical call.
Leadership · listening
14:12 – 19:08
5
How you'd explain ML model risk to a non-technical board.
Communication · cross-fn
19:09 – 23:47
6
When the data tells a story stakeholders don't want to hear.
Integrity · communication
23:48 – 28:22
7
Trade-offs between model performance and explainability.
Technical · judgement
28:23 – 32:41
8
Working under conflicting priorities from two senior leads.
Stress · diplomacy
32:42 – 36:09
9
A piece of feedback that changed how you work.
Growth
36:10 – 38:54
10
What you'd do in your first 90 days here.
Role-specific · plan
38:55 – 41:02
11
Questions for the hiring manager.
Closer
41:03 – 41:48
12
Anything you wish I'd asked.
Closer · open
41:49 – 42:00
Question 4 of 12 · Behavioural · Leadership & listening
"Tell me about a time you led a team through a technical decision where the team disagreed with your initial recommendation."
14:12 – 19:08
The AI followed up because Sarah's first answer was abstract. It pushed for a concrete situational example: "Could you take me through a specific instance — what the disagreement was, what you did, and what changed?"
"We were choosing between two architectures for a customer churn model — a simpler logistic baseline versus a gradient-boosted ensemble. I initially pushed for the ensemble. Two of my engineers pushed back: they argued the baseline would ship in two weeks versus eight, and we needed signal in production fast to prove the project's value. I asked them to walk me through their reasoning, then I changed my mind. We shipped the baseline first, learned the actual feature importance was very different from what we'd assumed, then built the ensemble against real production signal. The final model was better than what I'd originally proposed because it was informed by data we wouldn't have had otherwise. The two engineers who pushed back ran most of the second build."
Concrete situation cited Demonstrates intellectual humility Strong outcome reasoning Credits team explicitly
Competency mapping for this answer
Listening 9.4 / 10
Decision-making 9.1 / 10
Outcome ownership 8.9 / 10
Team credit attribution 9.6 / 10
Verbal signals · this question
Speech clarity 9.4
Structured reasoning 9.1
Concrete-vs-abstract ratio 8.7
Paralinguistic readout
Speaking pace 147 wpm
Steady · within natural range
Avg. pause length 0.7s
Considered · not hesitant
Filler words / min 2.1
Low
Permitted under SE jurisdiction. Speaking pace and pause length only — facial-recognition analysis is disabled by Higher policy across all regions.
For your follow-up
The AI suggests asking how Sarah would handle a case where the engineers' alternative didn't work out — testing whether her openness to dissent depends on outcomes.
H·
Higher
Demo Workspace
ML
M. Lindqvist
Back to Sarah Chen
Skills sandbox + Behavioural framework

Assessment results — Sarah Chen.

Live-coded against a real production scenario, then a behavioural profile across 184 questions designed to detect inconsistency. Both scored against the role profile, both auditable.

Sandbox task · 90 minutes · Live-coded
Build a streaming feature pipeline that flags anomalous transactions within 50 ms latency, given partial Kafka schema and three sample fraud patterns.
Python 3.12 PyFlink Submitted in 73 min 2 revisions
fraud_pipeline.py · selected excerpt
L 47 – 68
47def build_anomaly_pipeline(env: StreamExecutionEnvironment) -> DataStream:
48 # keyed by account so out-of-order events stay correctly partitioned
49 transactions = read_kafka_source(env, topic="tx.raw")
50 return (transactions
51 .key_by(lambda t: t.account_id)
52 .window(SlidingEventTimeWindows.of(Time.seconds(30), Time.seconds(5)))
53 .process(RollingZScore(threshold=3.5))
54 .filter(lambda r: r.is_anomalous)
55 .map(enrich_with_account_history))
56
57class RollingZScore(ProcessWindowFunction):
58 # incremental aggregation avoids buffering full window in state
59 def process(self, key, ctx, elements):
60 running = self.get_state("rolling")
61 for tx in elements:
62 mu, sigma = running.update(tx.amount)
63 z = (tx.amount - mu) / max(sigma, 1e-6)
64 yield AnomalyResult(tx, z, is_anomalous=abs(z) > self.threshold)
Test cases
12 / 13 passed
Synthetic stream · 10k events
42 ms p95
PASS
Burst pattern · card-testing fraud
38 ms p95
PASS
Drift pattern · slow-build fraud
44 ms p95
PASS
Out-of-order events
46 ms p95
PASS
Late arrival recovery
49 ms p95
PASS
Adversarial · velocity-bypass pattern
FAIL
Sustained load · 60 min · 8k events/s
47 ms p95
PASS
Note for reviewer: The single failure is a known hard case — Sarah flagged it in a comment and proposed a follow-up that would handle it.
Correctness
92%
Top 4% of submissions
Latency
44ms
Target: <50 ms
Code quality
9.0
Style, testability, comments
Behavioural assessment · 184 questions · Big Five framework

Personality vs. role profile

The shaded shape is Sarah. The dashed shape is the target profile defined for this role. Closer to the target on every axis is better.

Drive Collaboration Stress tolerance Decision making Consistency
9.2
Drive
9.1
Collaboration
8.7
Stress tol.
8.9
Decisions
7.6
Consistency
Anti-distortion analysis

Did Sarah perform for the algorithm?

184 questions · 23 paired inconsistency checks · Identical-meaning items presented far apart in the test.

Reliability 94%
Q12 ↔ Q47 · Rule adherence
Said she always follows protocol. Later, in a scenario where breaking protocol would save a customer, she said she'd still raise it through the right channel before acting. Consistent.
Δ 0.1
Q23 ↔ Q89 · Stress tolerance
Self-rated as "calm under pressure" early. In a forced-choice scenario about a missed launch, she chose the response acknowledging measurable frustration before correction. Honest, not performative.
Δ 0.3
Q34 ↔ Q102 · Collaboration preference
Strong-collaboration response in both items, separated by 68 questions. No drift toward "expected" answer.
Δ 0.0
Q56 ↔ Q141 · Risk appetite
Minor inconsistency on tolerance for product-side risk. Within statistical noise — flagged for transparency, not concern.
Δ 0.7
Q78 ↔ Q163 · Direct feedback comfort
Both items show preference for direct feedback over diplomatic phrasing. Consistent.
Δ 0.2
Plain-language summary: Sarah answered honestly. She did not select the most flattering option when the test gave her room to do so, and her contradictions across paired items are within the noise we expect from any human respondent. The system cannot detect every form of impression management, but on the dimensions it tests, this profile reads as authentic.
H·
Higher
Demo Workspace
ML
M. Lindqvist
Trust infrastructure · Compliance & ethics

Compliance.

Every algorithmic decision logged. Every variable used in scoring justified. Every protected characteristic locked at the model level — not as an admin setting. Defensible with regulators, with your union, with your candidates.

Quarterly audit ready Q1 2026 review complete. 47 hires across 6 markets · 0 flags raised · adverse-impact ratios within 4/5ths rule on every protected dimension.
Open audit pack
Decisions logged · this quarter
2.4M
Every score, filter, follow-up, and reranking — not just final outcomes.
Protected variables in active scoring
0
Locked at the model level. Not a configurable setting.
GDPR Art. 15 mean response time
4.2d
Against a 30-day legal requirement. 100% SLA met.
Union consultations on record
12
Quarterly bias audits reviewed jointly. Last: 14 March 2026.
Algorithmic audit log
Append-only. Hash-chained. Streamed to your SIEM in real time.
Live · 47 events / min
All events
Verification
Fairness checks
Access & rights
Blocked
Time (UTC)
Actor
Event
Entity
Hash
14:22:09
Auth. engine
Document scored 78% (medium) · Flagged for human review
Sarah Chen → NAI cert
8b1e44d2…
14:22:07
Auth. engine
Document scored 99.9% (high) · Verified
Sarah Chen → PhD thesis
a4f2b9e1…
14:18:32
Bias monitor
Periodic parity check · All groups within tolerance
Q1 2026 cohort
2d8c91a3…
14:14:01
Hiring panel
Decision review confirmed · Awaiting human sign-off
Sr. AI & Data Scientist
7e3a82b1…
14:09:47
Interview engine
Adaptive follow-up triggered · Q4 leadership probe
Sarah Chen
6d1e4596…
14:04:18
GDPR module
Art. 15 access request fulfilled · Full data export delivered
Candidate ID 4471
9b2c33ef…
13:58:22
Variable lock
Attempted access blocked · Field marked as proxy variable
candidate.postcode
3f8e1224…
13:51:55
Pulse engine
k-anonymity threshold enforced · Result suppressed (n=4)
Eng. Q1 — Team B
5a9e22d4…
13:46:09
Recruitment
Role profile published · Reviewed by hiring manager & HR
Sr. AI & Data Scientist
1c4d8807…
13:38:14
GDPR module
Art. 17 erasure completed · Personal data scrubbed across 14 systems
Candidate ID 4382
4e7b1a92…
Showing 10 of 2,478,213 entries this quarter Load more →
Variable lock · what the AI is allowed to consider

No protected characteristic ever enters a score.

Discrimination Act §1 · GDPR Art. 9 special-category data · EU AI Act Art. 10. Every blocked attempt is logged.

Permitted 6 categories
Role-relevant skills
Scored from sandbox tasks & competency mapping
Documented work experience
Verified employer histories & CV claims
Verified credentials
Authenticity engine results, not the school's prestige
Behavioral profile
Big Five, mapped to role demands
Communication assessment
Structure, clarity, listening — content not accent
Cognitive task performance
Standardised, role-calibrated
Locked 10 categories
Gender / sex
GDPR Art. 9 special category
Age / date of birth
Used only for legal eligibility, never scoring
Ethnicity / race
GDPR Art. 9 · not collected
Religion / belief
GDPR Art. 9 · not collected
National origin
Work-eligibility flag only · binary
Parental / family status
Discrimination Act §4
Disability status
Disclosed only for accommodation
Postcode / area proxy
Treated as socioeconomic proxy · blocked
Photo-derived features
No facial recognition · system-wide policy
Voice demographic inference
Pace & pause only · accent ignored
GDPR data subject rights

Every right exercisable in one click — by the candidate.

Self-service portal. No ticket required. Mean response time tracked publicly.

Art. 15
Right of access
47 requests this quarter · all served
4.2d mean
Art. 16
Right to rectification
12 requests · resolved within 48h SLA
1.8d mean
Art. 17
Right to erasure
8 requests · 1 lawful exception logged with reasoning
2.1d mean
Art. 20
Right to data portability
23 requests · machine-readable export served
100% met
Art. 22
No fully automated hiring
AI scoring is decision-support · human reviewer required
0 auto
Per the Swedish DPA: Mean response time is published quarterly. Any request taking longer than 14 days requires written justification, also auditable from this log.
Bias monitoring · Q1 2026

Adverse-impact ratios across the funnel.

Voluntary, anonymous self-ID at the end of the candidate journey. Aggregated only — never tied back to individuals.

Application → Category A/B classification All groups in tolerance
Lowest:highest = 0.86 · 4/5ths rule threshold = 0.80
Category A → Interview invitation All groups in tolerance
Lowest:highest = 0.91 · within tolerance band
Interview → Final offer All groups in tolerance
Lowest:highest = 0.88 · sample size below threshold for some sub-groups, marked transparently
Methodology & group definitions: Read the methodology →
Union dialogue · Sweden

Negotiated framework with worker representatives.

The union sits in the algorithmic governance committee. Quarterly bias audits are reviewed jointly. This is not consultation theater — they have a vote.

Unionen · Akademikerna · IF Metall
Co-signed framework agreement · Last consultation 14 March 2026
Active
AI is decision-support, never decision-maker. No employee is hired, fired, promoted, or denied a development opportunity by the system alone.
Right to human review. Any candidate or employee can request a named human to review any AI score, with stated reasoning, within 5 working days.
Joint quarterly bias audit. Union-nominated statistician reviews fairness metrics with full data access. Findings publishable.
Pulse anonymity threshold ≥ 5 respondents. No survey result returned for groups smaller than 5. No identification of dissenters.
Worker rep on the algorithmic governance committee. Voting member. Reviews any change to scoring weights or new variable inclusion before deployment.
Right to opt out of AI assessment. Candidates may request a fully human-led process. Stated and visible at application time.
Framework v3.1 · 14 March 2026 View consultation log →
H·
Higher
Demo Workspace
ML
M. Lindqvist
Back to Sarah Chen
Candidate experience preview

What Sarah sees on her phone — the same data, none of the asymmetry.

Every AI score the system generated about her, with a plain-language explanation. Every right she has under GDPR, one tap away. A "talk to a human" button that's never more than a swipe from any screen.

9:41
Hello,
Sarah
Sr. AI & Data Scientist · Higher & Co.
Your application
Stage 3 of 4 · Interview reviewed
Hiring panel reviewing now · expected response within 48h
Your AI assessment scores
Tap "Why?" on any score for the full reasoning.
Overall match Why this score? →
92%
Technical sandbox Why this score? →
9.0
Interview signal Why this score? →
9.4
Behavioral fit Why this score? →
8.9
Your interviewer notes
Hiring manager left a personal message
"Genuinely impressed by how you handled the disagreement question. Looking forward to talking again." — Erik L.
Average wait: 4 min · Mon–Fri 08:00–18:00 CET
Your rights · GDPR self-service
Download my data
Art. 15 · Access
Request human review
Art. 22 · Human-in-loop
Correct my data
Art. 16 · Rectification
Delete my data
Art. 17 · Erasure
Recent activity
2h ago
Your interview transcript was reviewed by Erik Lindqvist (CTO).
Yesterday · 16:42
All 7 of your documents passed authenticity verification.
2 days ago · 10:18
You completed the AI interview (42 minutes, 12 questions).
3 days ago · 09:00
Application submitted.
01
Every AI score the candidate sees in the dashboard, the candidate sees too.
No secret algorithm conversation possible — Sarah can read every score this system generated about her, the same instant a recruiter can. Information asymmetry is where mistrust lives. We removed it.
02
Plain-language reasoning behind every score.
"Why this score?" opens a panel that explains which answers, which moments, which behaviours drove the rating. Not a model card — a sentence. The kind of explanation a candidate could read out to a lawyer.
03
"Talk to a human" is never more than a tap away.
From every screen. Average wait time published. Real humans, named, accountable. AI is decision-support — and when the candidate wants a person, the path is one button, not a maze.
04
Every GDPR right is a button, not a form.
Access, rectification, erasure, human review — four taps, not four forms emailed back to a generic data protection inbox. Mean response times are tracked publicly. Candidates know how long it'll take before they ask.
05
Status updates push themselves.
No more silent voids between application and rejection. Every meaningful event in Sarah's process — a review, a verification, a panel comment — appears in her timeline. Whether or not she gets the role, she'll have known where she stood.
H·
Higher
Demo Workspace
ML
M. Lindqvist
Continuous development · Q1 2026 cycle

Development.

Every employee has a 35-minute conversation with the AI each quarter. Higher synthesises their answers and gives every manager a one-page brief — in plain language, before the 1:1.

Ready for a conversation Maria Andersson has earned a promotion conversation. Her Q1 brief is ready for you to review before the 1:1.
Open brief
MA
Maria Andersson
Senior Data Engineer
Erik Lindqvist
CTO
AB
Anna Bergström
Product Designer
JH
Johan Holm
Engineering Manager
LW
Lin Wei
ML Researcher
+18
Show more
23 employees this quarter
Maria's self-assessment · 23 March 2026 · 38 minutes
"I want to be the person leading platform decisions, not just executing them. The technical work doesn't scare me anymore — speaking up in front of the leadership team still does."
Synthesised from a 38-minute AI dialogue. Maria reviewed and approved this summary before publication.
Her three priorities
1 Move from execution to technical leadership on the streaming platform.
2 Get comfortable presenting architecture decisions to non-technical stakeholders.
3 Mentor at least one junior engineer — she sees this as proof of growth.
AI dialogue · excerpt

How the system found the actual constraint.

Maria's first answer was abstract. The AI didn't move on — it pushed for a specific recent moment. That's where the real signal lived.

AI
Higher · turn 14
"You said you want to grow into technical leadership. Can you take me to a specific moment in the last quarter where you felt you were being a leader — and one where you wished you had been?"
M
Maria
"The architecture review last month — Erik asked me to walk the team through the new event schema. I did fine. But two weeks before, in the platform strategy meeting with the CFO, I had thoughts but I didn't share them. I was the only senior engineer in the room and I went quiet."
AI
Higher · turn 15
"What stopped you?"
M
Maria
"I think I was waiting until I was 100% sure I was right. With other engineers I'm fine being half-formed. With executives I'm not. That's the gap."
System note: the constraint isn't presentation skills — it's the comfort threshold for being publicly half-formed in front of senior leaders. The development plan is calibrated to that.
Skills snapshot · last 4 quarters

Where Maria has moved.

Calibrated by peer feedback, observed work output, and project review. Not self-rated.

Distributed systems design
+1.4
Architecture documentation
+0.9
Cross-functional communication
+0.2
Mentoring & feedback
+1.1
Stakeholder presentation
+0.3
Technical risk judgement
+0.7
Spark: Q2 2025 → Q1 2026 All 18 skills →
Suggested development plan · next 90 days

Three concrete actions, not 30.

Calibrated to her stated priorities and current skill curve. Not generic e-learning catalog.

01 Low effort ~2h / week · 6 weeks
Lead two architecture reviews per month — including the CFO's office.
Specifically: ones that include non-technical stakeholders. The point isn't speaking practice, it's getting comfortable being half-formed in front of executives. Erik to introduce, then step out of the room.
02 Medium effort 3 sessions · ext. coach
Three 1:1 sessions with an external executive-presence coach.
Focused specifically on speaking before certainty — not generic public speaking. Pre-vetted coach: Sofia Markström, recommended internally by 4 other engineers who made the same transition.
03 Medium effort Ongoing · 1h / week
Mentor Lin Wei (ML Researcher, Q4 2025 hire).
Lin's stated her own priority is "learning to advocate for unfinished ideas." Mutually useful; Maria gets the felt experience of giving permission to be half-formed, by giving it.
Brief for Maria's manager

One page. No HR jargon. No surprises in the 1:1.

EL
Erik Lindqvist
CTO · Maria's manager
Hi Erik — Maria's growth this quarter has been real. The technical curve is solid; the constraint she identified isn't a skill, it's a comfort threshold — speaking before she's certain, in rooms where the audience is non-technical and senior. She named the platform-strategy meeting with the CFO as the specific moment.

What would help most right now: two things only. Bring her into the architecture conversations that include non-technical stakeholders, and don't translate for her — let her find her own way to explain the trade-offs. And in your 1:1 this Friday, ask her about Lin Wei. She's identified mentoring Lin as proof to herself that she's leading. That mentoring relationship is the lever, not the side project.

One thing to watch: Maria is high-conscientiousness and may try to take all three actions at full intensity. Two will produce more growth than three. Let her drop one without it feeling like failure.
Generated 2 minutes ago · Maria has read this · Reviewed by HR Add to next 1:1 →
H·
Higher
Demo Workspace
ML
M. Lindqvist
Organisational health · Q1 2026 pulse

Pulse.

Three short open-text prompts every six weeks. Higher clusters the language into themes and never reports on any group smaller than five respondents. The price of anonymity is honesty — and we charge it.

Movement worth your attention Engineering sentiment dropped 8 points this quarter — the theme is workload predictability after the December reorg. 47 employees mentioned it, unprompted.
See the theme
Response rate · Q1 2026
87%
217 of 249 employees · up 4pp from Q4. No incentive offered.
Themes surfaced
6
Clustered from 587 open-text responses by NLP. 4 returning themes, 2 new this quarter.
Anonymity threshold
≥5
Hard floor. 8 sub-groups suppressed transparently this quarter.
Last pulse fielded
23 mar
3 prompts · median 4 min to complete · next: 12 May.
Themes this quarter

What people wrote about — clustered, not categorised.

Sentiment scale 0–10 · trend vs last quarter
Theme · Returning · 4 quarters
Compensation transparency
38%
Mentioned
Sentiment
6.2 ↓ 0.4
Sample paraphrased quotes (k=5)
"I don't know how raises are decided here. I'd rather know the rule and disagree with it than guess."
Engineering · 11 respondents in cluster
"Our band ranges aren't published. So when someone gets promoted I have no idea what 'good' looks like."
Product & Design · 7 respondents
Theme · Returning · 3 quarters
Cross-team handoffs
31%
Mentioned
Sentiment
5.8 — 0.0
Sample paraphrased quotes (k=5)
"Eng and Product still operate on different timelines. We agree on the what; the when is always a fight."
Engineering · 9 respondents
"By the time something hits design review, half the trade-offs have already been made."
Product & Design · 6 respondents
Theme · Returning · 2 quarters
Senior leadership visibility
27%
Mentioned
Sentiment
7.4 ↑ 0.9
Sample paraphrased quotes (k=5)
"The recent town halls have been honest in a way I didn't expect. More of that, please."
Cross-functional · 14 respondents
"I actually know what the next 18 months look like now. That wasn't true six months ago."
Engineering · 8 respondents
Theme · New this quarter
Workload sustainability — Platform team
19%
Mentioned
Sentiment
5.5 ↓ 1.2
Sample paraphrased quotes (k=5)
"We've been on-call rotation for a year with no proper coverage backfill. It's catching up to us."
Platform Eng · 8 respondents
"Two people left in Q4 and we never replaced them. The work didn't go away."
Platform Eng · 7 respondents
Sentiment trend · top themes · 4 quarters

What's moving, and what isn't.

9 7 5 3 Q2 '25 Q3 '25 Q4 '25 Q1 '26
Comp transparency
Cross-team handoffs
Leadership visibility
Workload · new
Suppressed sub-groups

What we won't show you — and why.

Eight sub-groups had fewer than 5 respondents this quarter. We tell you they exist; we don't tell you what they said. Their data still rolls into the company-wide aggregate.

Engineering · Team B
n=4
Suppressed
Design · Senior IC
n=3
Suppressed
Data Science · Stockholm
n=4
Suppressed
Finance · Senior
n=2
Suppressed
People Ops
n=4
Suppressed
Negotiated with the union: the k=5 floor is in the framework agreement. Nobody — not the CEO, not HR, not us — can lower it. If you need finer-grained signal in a small team, ask the team directly with their consent.
Suggested manager actions

What the data is asking you to do — concretely.

For Erik Lindqvist · CTO
Address Platform team workload before it becomes attrition.
Sentiment dropped 1.2 points in one quarter — fastest fall on the board. Two unfilled roles since October. Not a culture problem; a math problem. Recommend: protect Q2 hiring slot, redistribute on-call rotation by 18 April.
Schedule briefing →
For the leadership team
Publish band ranges before the next pulse.
Comp transparency has been in the top three themes for four quarters. Sentiment has only fallen. The conversation is happening — without you in it. Recommend: bring a draft to the 6 May leadership offsite, publish before the 12 May pulse.
Add to offsite agenda →
For Anna Bergström · Head of Product
One cross-functional handoff retro per quarter — make it standing.
Cross-team handoff sentiment hasn't moved in 12 months. Same sentiment range, same teams, same words in the open text. The pattern says: ad-hoc fixes haven't worked. Recommend: a 60-minute joint Eng/Product retro on the last Wednesday of each quarter.
Draft the meeting →
1 / 11 Sign in
Higher prototype
Jump anywhere · next · esc back
Setup
Pipeline
Trust & reach
People
Action completed