Almost every HR system sold in Pakistan now says “AI” somewhere on its home page. That is not a criticism: the tools have genuinely improved, and the first pass through two hundred CVs for a junior developer role is exactly the kind of reading a model does well. But the word on a banner tells you nothing about what the software will do with your candidates’ data, or which decisions it will make for you. This post reads the claims closely.
We read the AI pages of five HR vendors popular with Pakistani companies on 4 October 2026, and wrote down what each page says, in its own words where possible. Then we describe what the AI in our own HR tool, Orbiz HRMS, does and does not do, with the same care. For a broader comparison, prices included, see the best HR software in Pakistan; this post is about the AI alone.
What the AI claims say, vendor by vendor
| System | What its own pages say the AI does |
|---|---|
| PayPeople | AI agents for recruitment and onboarding, payroll, compliance, performance, learning and engagement; an AI ATS that screens résumés against the job description; AI Buddy, a self-service assistant that takes text or voice requests for leave, payslips and expenses |
| SmartHCM | AI Recruitment: job descriptions written with a built-in AI chatbot, AI shortlisting, AI that conducts interviews and analyses responses and facial expressions, AI tagging of stored CVs; an AI assistant in the mobile app |
| FlowHCM | Describes itself as an AI-based, voice-driven HRMS; its mobile app supports voice functions such as sign-in and sign-out |
| Paismo | AI CoPilot, which watches payroll, performance and attendance for anomalies: duplicate payments, unauthorised raises, ghost employees, chronic absence and buddy punching |
| WebHR | The pages we read (home, Pakistan, applicant tracking) describe automation, such as automatic notifications and generated job images, rather than a named AI feature. Ask them directly |
| Orbiz HRMS | Résumé parsing, AI screening switched on per job, AI-generated assessments a person approves, and AI grading a person can override; plus a connector for Claude and ChatGPT |
Read side by side, the claims fall into four quite different kinds of AI, and they carry different risks. It is worth knowing which kind you are buying.
Four kinds of AI in HR software
1. Reading CVs
Parsing a CV into fields and scoring it against a job is the most common claim, and the most useful. PayPeople, SmartHCM and Orbiz all describe it. The question is not whether it works; it mostly does. It is what happens to the score. A score that orders your reading pile is a time saver. A score that silently rejects people is a decision, and one you would have to defend to a candidate.
2. Interviewing and judging people
SmartHCM’s page goes further than reading: it says its AI conducts interviews and analyses responses and facial expressions. That is a different category. If you consider it, ask how candidates are told, how the analysis is used in the decision, and whether a candidate can ask for a human review. Orbiz does not do this; HRMS reads CV text against a job, and it does not analyse video, voice or faces.
3. An assistant for employees
PayPeople’s AI Buddy and SmartHCM’s mobile assistant answer employees’ everyday questions and take requests: apply for leave on a date, show a payslip. That is genuinely handy, and it lives inside the vendor’s own app. The newer way to get the same thing is to connect the AI assistant the employee already uses, Claude or ChatGPT, to the HR system. More on that below.
4. Watching for anomalies
Paismo’s CoPilot is the clearest example: it looks for payments, raises and attendance patterns that do not fit. That is a sensible use of a model, close to what an auditor does. Ask what it flags, who sees the flag, and whether the person flagged is told.
What the AI in Orbiz HRMS does
HRMS is the only Orbiz tool with AI built in, and there are four pieces. Each one either reads, suggests or drafts. None of them moves, rejects or hires a candidate.
- Résumé parsing. A CV is read into a structured profile: name, current role, summary, years of experience, skills, languages, contact details, work history, education and certifications, with a confidence score. The candidate’s timeline records when parsing started, finished or failed.
- AI screening, per job, off by default. Switch it on for a job and each CV is read against that job’s requirements: a match score, a recommendation from Strong yes to Strong no, matched and missing skills, strengths, concerns and a summary, with the requirements used kept beside the result. The AI screening page covers it in full.
- AI-generated assessments. Pick the job, five to thirty questions and whether to include coding questions, and HRMS drafts an assessment with a suggested time limit and passing score you can override. You preview it and either approve it or discard it before any candidate sees it.
- AI grading. On a completed test, a grader can trigger AI grading: each graded answer gets feedback, the issues found and a confidence score. Someone with grading permission can change any score.
The data question has a plain answer. Screening and parsing are done through OpenAI’s API, the CV text and job description are sent only when your organisation uses those features, and OpenAI is listed on our sub-processors page with that scope. If that is not acceptable for a role, leave screening off for it; the pipeline, interviews and scorecards work the same without it.
The other route: connect the assistant you already use
Separate from the AI built into HRMS, every Orbiz member can connect Claude or ChatGPT to their organisation through Orbiz MCP. In HRMS that gives the assistant 14 actions, 9 that read and 5 that change something: the workflow tasks assigned to you, open jobs and candidates. “What HR tasks are waiting on me?”, “add the candidate from this email to the Backend Engineer job” and “move Ayesha Khan to the interview stage” each work as written.
The assistant acts as the person who connected it, with that person’s permissions, so a recruiter’s assistant sees what the recruiter sees. It cannot reach employee records and documents; interviews, assessments and offer letters; organisation, department and role settings. Nothing can be deleted through it. The same connection reaches pay in Ledgr for each person’s own payslips, which is the question HR hears most in a Pakistani office at the start of every month. The full HRMS list is on the HR assistant page.
This is not unique to Orbiz, and we would not claim it was. Several HR vendors abroad publish connectors of their own; we compared seven in which HR software works with Claude and ChatGPT. Setup is described inside the product, under Settings → Connected AI apps. On the assistant side, Claude’s help pages list custom connectors for the Free, Pro, Max, Team and Enterprise plans, with one connector on Free.
Questions to ask any vendor about its AI
- Which decisions does it make on its own? Ask specifically whether a score can reject, advance or hide a candidate without a person.
- Can it be switched off, and how narrowly? Per job, per feature, or only for the whole account?
- Where does the data go? Which model provider, which country, and is that provider listed as a sub-processor in writing?
- Does it analyse anything other than text? Video, voice and faces are a different conversation with candidates.
- Can a person override it, and is the override recorded? A grade or score you cannot change is a grade you cannot defend.
- For assistants: whose permissions does it use? It should be the employee’s own, not an account that sees everyone.
Who this suits, and who it does not
Orbiz HRMS suits a software house, agency or startup in Pakistan that hires steadily, wants help with the reading but not with the deciding, and would like HR, attendance and payroll in one place. It is a weaker fit if you want AI to run interviews, if you need benefits administration, or if your main need is an assistant inside a mobile app today: the Orbiz mobile app is coming soon, and we do not give dates. For the statutory side of pay, see payroll in Pakistan.
Sources, read 4 October 2026: PayPeople’s home page and its AI ATS and AI Buddy pages; WebHR’s home page, its HR software in Pakistan page and its applicant tracking page; SmartHCM’s home page and AI Recruitment page; FlowHCM’s home page and recruitment page; Paismo’s home page and AI CoPilot page; Claude Help Center, “Get started with custom connectors using remote MCP”. Vendors change their pages; check each one before you decide.
Built for your industry
