HR and recruitment
Sheet 8/12 · Medium · 3–8 weeksTwo hundred CVs for one vacancy, the same leave question every week, and onboarding by word of mouth.
- A vacancy on a WhatsApp group produces hundreds of CVs as photos and PDFs; someone opens each one.
- Shortlisting depends on who reads the pile; reasons for rejection are not recorded.
- Staff ask the same questions — leave days, payslip queries, NSSA, overtime — and HR answers each by hand.
- New staff learn the job from whoever sits next to them; policies exist in a drawer.
- Contracts and warning letters are drafted from old ones with the wrong names left in.
- Read every CV into the same structure (qualifications, experience, location, salary expectation) so a person shortlists from a table, not a pile.
- Apply *your* written criteria consistently and show why each candidate met or missed them.
- Answer policy questions from your own handbook, over WhatsApp, with a link to the clause.
- Draft contracts, letters and onboarding checklists from templates with the right names in.
- Judge character, fit or honesty — it reads what was written, including what was invented.
- Decide who to hire or dismiss; Zimbabwean labour law and simple fairness require a person to own that.
- Avoid bias on its own; criteria you set can encode it, and you must check outcomes by group.
- Answer questions your handbook does not answer — it will say so, if it is set up honestly.
| Effort band | Medium · 3–8 weeks |
|---|---|
| Data you need | Written criteria per role; the handbook; templates; a leave/payroll export |
| Integration points | Email/WhatsApp intake; HR or payroll system (optional) |
| People | HR owner; hiring managers who set criteria; a reviewer for every decision |
| Measure | Time to shortlist; policy queries answered by the assistant; onboarding completion |
| Zimbabwe constraint | Applicant and employee data raise your SI 155 licence tier; labour law keeps decisions human |
| Code | Shape | Fit | Why |
|---|---|---|---|
| S-02 | Pilot | Primary | One vacancy, real CVs, your criteria; compare the shortlist with the one a person makes. |
| Tool | Applicant tracking or HR platform with built-in screening and a policy assistant | Primary | For most firms a subscription tool covers structured CVs and handbook Q&A without a build. |
| S-05 | Training | Secondary | HR staff drafting with templates and a policy on candidate data. |
| S-03 | Implementation | Secondary | A WhatsApp policy assistant on your handbook, connected to leave balances, for larger workforces. |
| S-04 | Integration | Later | Payroll and leave systems, when the assistant must answer 'how many days do I have left'. |
- Where are applicants' CVs stored and processed, in which country, and for how long after the vacancy closes?
- Show the screening criteria as we would write them, and show a candidate's result with the reasons.
- How do we check the shortlist for bias by gender, age, location or school?
- Does the policy assistant answer only from our handbook, and what does it say when the handbook is silent?
- Which decisions are blocked from automation (rejections, warnings, terminations)?
- Does adding applicant data change our data-controller licence tier, and will you help us document it?
The full bank of 60 questions, grouped by risk, is at /vendor-questions/.
Recruitment changes your licence tier
Zimbabwe’s data-controller licensing regulations (SI 155 of 2024) place controllers in tiers by the number of data subjects they process, with fees from US$50 to US$2,000 a year and a licence that must be renewed annually. A business with 40 staff that runs three recruitment rounds a year may hold personal data on well over a thousand people once applicants are counted. Before you automate CV intake, decide the retention period (delete after the vacancy closes plus a stated period), and ask the vendor to document where applicant data is processed. It is a small step that keeps the tier — and the exposure — under control.
What a screening table looks like
| Candidate | Qualification required | Years relevant | Based in | Salary expectation | Meets criteria | Reason |
|---|---|---|---|---|---|---|
| A | Yes (Diploma) | 4 | Harare | US$650 | Yes | All criteria met |
| B | No | 6 | Bulawayo | US$900 | No | Qualification not stated |
| C | Yes (Degree) | 1 | Harare | US$500 | Review | Under experience threshold by 1 year |
Illustrative. The point is the last two columns: every decision has a reason a person can read and overrule. If a vendor’s tool produces a “score” without reasons, it is not usable in a dispute.
Policy assistant: only from the handbook
The right setup answers “how many days’ leave am I entitled to?” with the clause and the number from your handbook, and answers “can I take leave during stock-take?” with “the handbook does not cover this — ask HR”. A model that answers the second question confidently has been set up to guess. Test it with five questions the handbook does not answer.