KarmaBuilt-in AI

The AI thatis built onyour data.

Karma is not a chatbot bolted onto an HR system. It reads across every module — organisation, absence, payroll, recruitment, feedback and learning — because they are one dataset rather than six.

Which teams are over their absence budget?
Three teams are over budget this quarter: Support (112 %), Produktion & Lager (104 %) and Sales (101 %).
Draft · note to 3 managers
Adds a talking point to their next 1:1
Approve
Example data

How it works

Ask in plain language. Karma prepares the work.

Any question about your people, in your own words. No report builder, no filters. Karma pulls the data, drafts the change and waits for a person to say yes.

Try Karma
Karmawithin your access

Choose a question…

    AskIn your own words, about anyone you have access to
    Karma preparesThe data pulled, the change drafted, field by field
    You approveNothing is written until a person says yes
    LoggedThe approval and the change go on the audit log

    AI as the operating layer

    Not a feature. The layer underneath everything.

    Most HR tools add AI to a single screen. In Komma, Karma sits on the same live employee record that every module reads from — so it sees the whole picture and can act across all of it in one draft.

    One record, no copiesHired in Recruitment, they already exist in Organisation, get an onboarding flow, record time, join the pay run and are assessed in a review. Nobody types the same thing twice.
    Context from every moduleBecause nothing lives in a separate system, Karma’s drafts take contracts, schedules, agreements and pay bands into account at the same time.
    No exports, no nightly filesEvery module reads from and writes back to the same data. Nothing drifts apart, and nothing has to be re-synced before Karma can help.

    Human in the loop, by design

    Karma drafts. You approve. Everything is logged.

    Karma pulls the data, drafts the change and shows you exactly what will happen — field by field — before anything is written to your HR system.

    1. ReadsKarma pulls what it needs from the live employee record — within the access of the person asking.
    2. DraftsIt prepares the change across every module the request touches.
    3. ShowsYou see exactly what will change, field by field, before anything is saved.
    4. Waits for youNothing is written until a person approves. The approval, and what was changed, goes on the audit log with their name.
    1. 2Two colleagues offered to take the shift. Karma lines them up; a person picks.
    Karma finds: 54 findings ranked across the platform, with roster cards and reimbursement cases12
    1. 1All 54 cards, ranked across the platform: reimbursement, payroll, roster, recruitment and more.
    2. KarmaEach card states its basis: derived from live data.
    3. KarmaMikkel declined sick, but no sick leave is registered, so payroll does not know yet.

    Karma in Recruitment

    Point at a match and Karma explains it.

    Karma reads the job ad, proposes the skills the role needs and how much each one counts, and scores every candidate against them. Hover a match and the reasoning is there, criterion by criterion.

    • Requirements, skills, experience, motivation and culture fit, each weighted by importance
    • Marked as AI-generated, with a reminder to review it
    • Who moves on is decided by people
    Candidates for Sales Manager with the job match breakdown open on hover: requirements, skills, experience, motivation and culture fit1
    1. KarmaEach criterion scored and explained, weighted by importance.
    2. 191 % job match. The shortlist starts from facts.
    3. KarmaAI-generated, and it says so. People decide.

    Karma on mobile

    In the pocket of the people on shift.

    Most employees will never open the desktop. On the phone Karma notices what is missing, suggests the questions people actually ask, and takes them to the right screen.

    • A missing registration, caught before it reaches payroll
    • Holiday left, the next 1:1 or last month’s payslip — one question away
    • The same rule: Karma prepares, the person confirms
    Karma on mobile: a missing time registration, and suggested questions about holiday, the next 1:1 and the payslip1
    1. KarmaKarma sees a missing registration for Friday and offers to fix it.
    2. 1Suggested questions: holiday left, the next 1:1, the August payslip.

    Karma and your data

    The AI is an assistant, not a decision-maker.

    There are good reasons to be sceptical about AI in HR. Here is what we do about them.

    1. No automated decisionsKarma suggests and ranks. Hiring, pay, warnings and dismissals are decided by people — and that is not a setting you can switch off.
    2. Everything is traceableEach suggestion is logged with what Karma proposed, what was changed, and who approved it.
    3. Your data does not train anyone else’s modelsKarma works on your own dataset. The content is not used to train models that benefit other customers.
    4. Access follows rolesRole-based access and SSO through Microsoft Entra. Karma works within the permissions of the person asking.
    5. European by designHosted in the EU and built for GDPR, with transparency and human oversight in every AI feature.
    6. It can be switched offKarma can be disabled per module, if you would like it in recruitment but not in payroll processing.

    Built for the Nordics

    Made for how Nordic HR actually works.

    Collective agreements, supplements and the payroll engine you already trust. Komma owns everything gross — and hands over a validated pay run.

    Collective agreements built inCollective-agreement rules and supplements are applied to registered time before anything reaches payroll.
    Validated before it leavesPre-payroll validation catches errors while they are still cheap to fix — not after the payslips are out.
    Your payroll engine staysGross-to-net, A-skat, AM-bidrag, eIndkomst and payslips stay with the engine you already use.

    Where Karma is heading

    Two tracks. One rule.

    We are widening Komma across more of HR, and teaching Karma to carry multi-step workflows. What does not change: Karma drafts, people decide. The roadmap describes direction, not delivery dates.

    Track 1 · Wider: more of HR, in one recordEvery module we add reads from and writes back to the same employee record. You get fewer systems — and Karma gets more context to draft from.
    Track 2 · Deeper: from single drafts to whole workflowsToday Karma drafts one change at a time. Next, it prepares a whole flow — hire, onboard, schedule, first pay run — as one plan you review and approve step by step.

    FAQ

    What HR teams ask about Karma.

    Does Karma change anything on its own?

    No. Karma drafts the change and shows you exactly what will happen, field by field. Nothing is saved until a person approves it, and every approval is recorded in the audit log with that person’s name.

    What can Karma do today?

    Karma parses CVs into scored skills, drafts job ads, suggests shift coverage, summarises evaluations and spots absence patterns — with the same approve-first rule everywhere.

    Is our data used to train AI models?

    Karma works on your own dataset. The content is not used to train models that benefit other customers.

    Where is our data stored and processed?

    In the EU. Komma is built for Nordic organisations, with GDPR as the baseline for every AI feature.

    Who can see what Karma sees?

    Karma works within the role-based access of the person asking — if you can’t see salaries, neither can Karma when it works for you. Sign-in runs through Microsoft Entra SSO.

    Can we switch Karma off?

    Yes, per module. You can use it in recruitment but not in payroll processing, for example.

    Does Komma replace our payroll system?

    No. Komma owns everything gross — pay components, collective-agreement rules, time-to-supplement valorisation and pre-payroll validation. The gross-to-net calculation, A-skat, AM-bidrag, eIndkomst and payslips stay with your payroll engine: Zenegy, Danløn, Dataløn, Visma, Lessor or Intect.

    Ask it something on your own data.

    The quickest way to judge Karma is to put a real question to it against your own org structure.