Fazley RabbiAI can take some of the paperwork out of hiring: drafting job posts, organising applications, summarising CVs against criteria you set and scheduling interviews. What it should not do is decide who is rejected or hired. CV screening tools can quietly repeat the biases in the data they learned from, misread unusual or scanned CVs, and reward people who are good at writing CVs rather than good at the job. Use AI to save admin time, and keep every decision about a person with a person.
The useful parts
For a small or medium business that receives a pile of applications for each opening, AI can help with:
- Writing a clear job post. Turning rough notes into a structured description with duties, requirements and how to apply.
- Organising applications. Pulling names, contact details, experience and location from CVs into one sheet, so nothing is lost in an email inbox or a WhatsApp chat.
- Summarising against your criteria. "Does this CV mention experience with accounting software? Which district does the applicant live in?" The answers still need checking.
- Drafting interview questions based on the actual duties of the role.
- Scheduling and polite updates. Sending interview times and "thank you, we have filled the role" messages, so that no applicant is left waiting indefinitely.
The fairness risks
Bias in AI screening is not a remote, theoretical worry. It follows from how these systems work. An AI that learns what a "good candidate" looks like from past examples can learn the patterns of past decisions, including unfair ones.
Common ways screening becomes unfair:
- Proxies for things that should not matter. University name, home district, gender, age, religion, a career gap for childcare or illness, even the style of a name can influence a score without anyone intending it.
- Format penalties. A CV written in Bangla, a phone photo of a printed CV or an unusual layout may be misread or scored lower. That can unfairly filter out capable people from district towns.
- Keyword games. People who stuff their CVs with the right words get through; people who describe real experience in plain words may not.
- Confident summaries. An AI summary can leave out the most relevant line, or mention a qualification that is not there.
What AI cannot judge
A CV is a story a person tells about themselves. AI can read it, but it cannot see honesty, reliability, how someone treats customers or whether they will work well with your team. It cannot verify certificates or references. And it cannot weigh what you know about your own business, such as whether a candidate can manage the commute or what the shop floor really demands. Those need conversations, practical tasks and a manager's judgement.
A fair process that keeps people in charge
- Decide the criteria first. Write down the must-haves and nice-to-haves before you read any CVs, based on the job, not on a "type" of person.
- Hide what should not count. Where possible, remove photos, age, religion and similar details before review.
- Use AI to organise and summarise, not to reject. Every application should be seen by a person, at least briefly, especially those the AI ranks low.
- Check summaries against the original CV before you rely on them.
- Use a practical task or a structured interview with the same questions for everyone.
- Record why each decision was made in plain words, so you can explain it later.
Privacy and the law
CVs are full of personal data: phone numbers, addresses, family details, sometimes health information. Before you paste them into any AI tool, find out whether the tool stores or learns from what you send, and who can see it. Delete applications you no longer need. Employment and data protection rules apply to hiring, and they change, so ask a lawyer or the relevant authority about your obligations instead of relying on a tool vendor's assurances. The article on AI and data privacy for businesses lists the questions to ask.
Questions to ask before buying an AI hiring tool
- Can it explain why it scored or ranked a candidate the way it did?
- Has it been checked for bias, and can the vendor show how?
- Can it handle Bangla CVs and scanned documents?
- Where is applicant data stored, and for how long?
- Can a person easily override every decision?
Where to start
If you want to use AI in HR, start with lower-risk admin: interview scheduling, polite status messages, or leave and attendance tracking. Keep screening decisions human. The principles in human review in AI workflows matter here more than almost anywhere else. If you would like help designing a fair, practical setup as part of my AI and workflow automation work, get in touch.