AI for HR
The repetitive half of HR, handled automatically.
What is AI HR?
AI for HR covers the administrative side of the function: screening and ranking applications against stated criteria, answering staff questions from your own handbook, and driving onboarding paperwork to completion, while leaving hiring and disciplinary judgements with people.
HR time disappears into two places. Reading applications, most of which will not proceed, and answering the same staff questions about leave balances, pay dates and policy wording.
Both automate well because both are high volume and rule driven. Applications are screened against criteria you set and ranked with reasons attached. Staff questions are answered from your own handbook rather than from a general knowledge base.
What stays with people is every decision that affects somebody employment. The system sorts, summarises and answers. It does not reject candidates or resolve grievances.
What we refuse to automate in hiring
Duara implements HR automation from Nairobi, most often for businesses receiving more applications than anybody can realistically read.
We refuse to build automatic rejection, and we say so before the scoping call rather than after. Screening that discards candidates without human review is both a fairness problem and a legal exposure, and the time saved is small compared with the ranking that we will build.
Who this is for
- You receive more applications than you can read
- HR answers the same questions every week
- Onboarding steps get missed
- Leave requests are tracked in email
Industries we serve
- Retail chains
- Manufacturing
- Hospitality groups
- Logistics operators
- Healthcare providers
- Professional services firms
Problems this solves
These are the four situations where HR automation returns real hours rather than simply adding a system.
- Applications nobody has time to read. A single advert produces hundreds of responses, most of which are skimmed at best and never acknowledged.
- The same staff questions weekly. Leave balances, pay dates and policy wording consume a disproportionate share of the HR week.
- Onboarding steps quietly missed. Accounts, equipment, documents and induction happen inconsistently because nobody is tracking the full list.
- Leave managed in email. Requests, approvals and balances live in inboxes, so nobody can answer a balance question without searching.
What we deliver
Applications screened and ranked with reasons, staff questions answered from your handbook, onboarding tracked to completion, and every employment decision left with a person.
Applications screened and ranked against your stated criteria
Reasons attached to every ranking so it can be challenged
Staff questions answered from your own handbook
Onboarding checklists triggered and tracked to completion
Leave requests routed through your approval rules
A record of what was asked and what was answered
Features
- Ranking, never rejection. Candidates are ordered with reasons shown. Nothing is discarded automatically, and we do not build that even on request.
- Answers from your handbook only. Staff questions are answered from your own policy documents, so nobody receives a generic answer that contradicts your terms.
- Sensitive topics escalate immediately. Grievances, pay disputes and anything disciplinary go straight to a person without an attempt at an answer.
- Onboarding driven to completion. The system chases outstanding steps rather than presenting a checklist somebody has to remember to open.
- Criteria stated explicitly. Screening criteria are written down and visible, so the basis for ranking can be examined and argued with.
Benefits
- Every applicant gets read. Ranking means the whole pile is assessed on the same basis rather than the first thirty getting attention.
- HR stops being interrupted. Routine questions are absorbed, which returns the function to work that needs judgement.
- New starters begin properly. Nothing is missed on day one, which affects how quickly somebody becomes useful.
- Candidates hear back. Even unsuccessful applicants receive a reply, which costs nothing and protects your reputation as an employer.
Technologies we use
- Structured screening. Criteria encoded explicitly so ranking is reproducible and can be audited rather than being a black box.
- Handbook retrieval. Answers drawn only from your policy documents, with the source shown so staff can verify.
- HRIS integration. Connection to your existing HR or payroll system so leave balances and records stay authoritative.
- Escalation rules. Topic detection routing anything sensitive to a person before any answer is generated.
- Onboarding workflow. Task sequences with owners and reminders, tracked until every step is complete.
Fairness constraints shape the technology here more than capability does. We deliberately build less than is possible.
How the process works, step by step
HR automation typically reaches live use in four to five weeks, with screening running in parallel with manual review at first.
-
Step 1
Free volume review
We establish application volumes and the questions HR answers most often, which sets the scope.
-
Step 2
Criteria and policy capture
We write down the screening criteria explicitly and gather the handbook the answers will come from.
-
Step 3
Parallel screening
The system ranks while your team also reviews manually, so you can see whether the ranking agrees with your judgement.
-
Step 4
Staff question rollout
The handbook assistant goes live for staff, with sensitive topics escalating from day one.
-
Step 5
Onboarding automation
Checklists and reminders are switched on once screening and questions are settled.
Quality assurance and how we work
The commitments that apply when a system touches employment decisions.
- We will not build automatic rejection. This is refused at scoping, not negotiated later. Discarding candidates without human review is a fairness and legal exposure we decline to create.
- Every ranking shows its reasoning. You can see why a candidate sits where they do, which means the basis can be challenged and corrected.
- Criteria are explicit and reviewable. Screening rules are written down rather than inferred, so nobody has to guess what the system is weighting.
- Sensitive matters never get an automated answer. Grievances and disputes escalate before any response is generated, without exception.
- Staff data stays inside your systems. Employment records remain in your HRIS and are never used to train public AI models.
Frequently asked questions about AI HR
Can it reject candidates automatically?
No, and we will not build it. Discarding applications without human review is a fairness and legal exposure we decline to create, and we say so at scoping.
So what does screening actually do?
It ranks every application against criteria you state, with the reasoning shown, so the whole pile is assessed on the same basis.
How do we know the ranking is fair?
Criteria are explicit and every ranking shows its reasoning, so the basis can be examined and argued with rather than taken on trust.
Where do staff answers come from?
Your own handbook and policy documents only, with the source shown, so nobody receives a generic answer that contradicts your terms.
What about grievances or pay disputes?
Those escalate to a person immediately, before any answer is generated. That limit is built in rather than left to judgement.
Does it integrate with our HR platform?
Yes, to the common HRIS and payroll systems, so leave balances and records stay authoritative rather than duplicated.
Does it process leave through our own rules?
Yes, routed through your existing approval rules, with balances read from your HR system rather than a separate copy.
How long before it is handling HR queries?
Four to five weeks, including a period where screening runs in parallel with your manual review.
Why run screening in parallel first?
So you can see whether the ranking agrees with your own judgement before relying on it. If it disagrees, the criteria need work.
Is staff data kept private?
Employment records stay in your systems and are never used to train public AI models.
Does it acknowledge unsuccessful applicants?
Yes, which costs nothing and protects your standing as an employer. Most businesses currently do not.
Can it handle applications in Swahili?
Yes, and mixed language applications, which are common for operational roles.
What if a candidate challenges a decision?
The audit trail shows what the system contributed and what the criteria were. The decision itself was made by a person.
Can it interview candidates?
No. We do not build automated interviewing, because the assessment quality does not justify the fairness risk.
Will it help with onboarding?
Yes, by triggering and chasing the full sequence of steps rather than presenting a list somebody has to remember to open.
Does it work for high volume seasonal hiring?
That is where it returns most, because ranking hundreds of applications on a consistent basis is exactly the bottleneck.
What criteria should we use?
Whatever you can state and defend. If a criterion cannot be written down clearly, it probably should not be driving decisions.
Can it detect CV exaggeration?
It can flag inconsistencies for a person to look at. It does not conclude anything about honesty, because that judgement carries consequences.
Does HR still need to read applications?
The shortlist, yes. The value is that the ranking is produced consistently across all of them rather than the first few.
What are the running costs for HR automation?
It depends on application and query volume. The monthly projection uses your own figures and is provided before commitment.
Glossary of terms
HR terms as they apply to this work.
- Screening
- Assessing applications against stated criteria to produce a ranked order, without discarding anybody.
- HRIS
- The human resources information system holding your employee records, leave balances and payroll data.
- Onboarding
- The sequence of steps required to bring a new starter into the business properly.
- Escalation
- Routing a matter to a person, used here for anything sensitive or disciplinary.
- Selection criteria
- The explicit requirements a candidate is assessed against, written down before screening begins.
- Audit trail
- The record of what the system did and why, needed if a decision is later questioned.
Related services in AI Implementation
Talk to us about AI HR
If you are receiving more applications than you can read, bring the last advert and its response volume. That is usually the clearest case for this.
Start your project