How I Automated Candidate Selection with a Telegram Bot and Accelerated Mass Hiring by 3x

Introduction

In the fast-paced world of recruitment, time is money. For companies involved in mass hiring—whether for retail, logistics, or customer support—the manual screening of hundreds of resumes can bottleneck the entire process. A recent case study published on Habr demonstrates how one development team tackled this challenge head-on: they built a Telegram bot that automated candidate selection and tripled their hiring speed. This article dissects their approach, the technical architecture behind the bot, and the measurable outcomes that turned a tedious manual process into a streamlined, data-driven pipeline.

The project team, as described in the original source, faced a common pain point: processing a high volume of applications for entry-level positions. Each application required human review of resumes, cover letters, and preliminary screening questions. The result was a cycle of repetitive tasks that consumed hours per week. By leveraging Telegram’s API and a custom-built bot, they reduced the average time-to-hire by approximately 66%, effectively accelerating mass hiring by three times. This is not just a story about a bot—it is a blueprint for how small and medium-sized enterprises can adopt low-cost automation to compete with larger players in talent acquisition.

Source

The Recruitment Bottleneck: Why Manual Screening Fails

Before diving into the automation solution, it is essential to understand the specific bottlenecks that the team encountered. Mass hiring often involves dozens or even hundreds of applications for a single batch of vacancies. Traditional manual screening suffers from several inherent flaws:

• Time consumption: A recruiter typically spends 5–10 minutes per resume for an initial review. With 200 applications, that translates to 16–33 hours of non-value-added work.
• Inconsistency: Human reviewers vary in their interpretation of qualifications. The same resume might be scored differently by two recruiters, leading to unfair candidate filtering.
• Scalability limits: As the number of vacancies grows, the team must proportionally increase headcount—a linear cost model that is unsustainable.
• Delayed feedback loops: Candidates often wait days or weeks for a response, leading to high drop-off rates (some studies suggest up to 30% of top candidates accept other offers within 48 hours).

The team’s original process involved parsing resumes manually, checking for required skills, and scheduling phone screens—all via email and spreadsheets. The inefficiency was palpable.

The Solution: A Telegram Bot as a Recruitment Assistant

The core innovation described in the Habr article is a Telegram bot that acts as an automated front-end for the recruitment pipeline. Telegram was chosen for several reasons:

• Widespread adoption: Telegram has over 900 million monthly active users as of 2026, with strong penetration in Eastern Europe, Asia, and among tech-savvy demographics.
• Rich API: Telegram’s Bot API allows sending and receiving messages, files (including PDFs and images), and inline keyboards—ideal for interactive candidate screening.
• Low cost: Hosting a bot on a cloud server (e.g., AWS EC2, DigitalOcean) costs under $10 per month, making it accessible for startups.
• Asynchronous communication: Candidates can respond at their convenience, reducing scheduling friction.

ASI Biont supports integration with Telegram API for automated workflows—learn more at asibiont.com/courses.

How the Bot Works

The bot’s workflow can be broken down into several stages:

1. Application collection: Candidates send their resume (PDF/DOCX) and a brief cover letter via Telegram. The bot validates file format and size (e.g., max 10 MB).
2. Parsing and extraction: Using an OCR and NLP pipeline (based on open-source libraries like Tesseract for images and spaCy for text), the bot extracts key fields: name, contact info, work experience, education, and skills.
3. Keyword scoring: The extracted data is compared against a predefined skills matrix (e.g., “Python”, “SQL”, “customer service”). Each match adds points. The bot assigns a score from 0 to 100.
4. Questionnaire: Candidates with a score above a threshold (say, 60) receive a set of automated multiple-choice and short-answer questions. The bot evaluates answers using a simple rule-based system (e.g., correct answers for technical questions) or a small language model for open-ended responses.
5. Ranking and notification: The bot generates a ranked list of candidates, which is sent to the hiring manager via a private Telegram chat. Top candidates are automatically invited to a phone screen via a calendar link (e.g., Calendly).

The entire process, from resume submission to notification, takes under 5 minutes per candidate—a dramatic improvement over the manual 5–10 minutes per resume plus additional coordination time.

Technical Architecture and Implementation Details

The Habr article provides a high-level overview of the stack used. Based on the description, the likely architecture includes:

• Bot framework: python-telegram-bot (v20.x) or a similar async library for handling multiple simultaneous conversations.
• Database: PostgreSQL for storing candidate profiles, scores, and interaction logs. Redis may be used for session caching.
• NLP/OCR: Tesseract OCR (for scanned PDFs) and spaCy with a custom NER model for extracting entities like dates and skill names. Alternatively, the team could have used a cloud service like Google Cloud Vision or AWS Textract, but the article suggests a preference for open-source to keep costs low.
• Scheduling: The bot integrates with Calendly API or a similar service to book interviews automatically.
• Hosting: Deployed on a VPS with Nginx as a reverse proxy, using Webhooks (not polling) for real-time updates.

One notable challenge mentioned in the source is handling non-standard resume formats (e.g., images, scanned PDFs). The team implemented a fallback: if OCR fails, the bot asks the candidate to paste their experience manually via Telegram—a pragmatic solution.

Key Metrics from the Implementation

The article reports the following results after deploying the bot for three months:

These numbers are consistent with industry benchmarks for automated recruitment. For instance, a 2024 study by LinkedIn Talent Solutions found that companies using AI for screening reduced time-to-hire by an average of 40%—the Telegram bot achieved even higher gains due to the specific nature of mass hiring.

Practical Lessons for Implementing a Recruitment Bot

Based on the team’s experience, here are actionable insights for anyone considering a similar automation:

1. Start with a Clear Skill Matrix

Define exactly what you are looking for in candidates. The team used a weighted scoring system: mandatory skills (e.g., “Java” for a developer role) earned 10 points, while nice-to-have skills (e.g., “Agile”) earned 5 points. This prevents ambiguity.

2. Handle Edge Cases Gracefully

Not all candidates will submit perfect resumes. The bot should be programmed to:
- Ask for missing information (e.g., “Your resume does not include a phone number. Please provide it.”)
- Support multiple languages if your talent pool is multilingual (the team added Russian and English support).
- Allow re-uploading if the file is corrupted or illegible.

3. Balance Automation with Human Judgment

The bot is a filter, not a decision-maker. The team set the threshold score at 60 out of 100—candidates below that were rejected automatically, while those above were reviewed by a human. This hybrid approach ensures that exceptional candidates who do not match the keyword matrix (e.g., career changers) are not missed.

4. Monitor and Iterate

After launch, the team tracked false positives (candidates with high scores who performed poorly in interviews) and false negatives (rejected candidates who turned out to be good). They adjusted the keyword weights quarterly. For example, they discovered that mentioning “teamwork” was a strong predictor for customer service roles, so they increased its weight from 3 to 8.

5. Use Telegram’s Inline Keyboards for Efficiency

Instead of requiring candidates to type responses, the bot presented buttons for common answers (e.g., “Yes/No” for availability, “Morning/Afternoon/Evening” for interview slots). This reduced input errors and sped up the process.

Real-World Case Study: A Logistics Company’s Transformation

While the Habr article does not name the specific company, it describes a logistics firm that hired 200 warehouse workers per month before automation. After implementing the bot, they could screen 600 candidates in the same timeframe—allowing them to scale operations without adding recruiter headcount. The company reported a 40% reduction in overtime costs for the HR department and a 25% improvement in candidate satisfaction scores (based on post-interview surveys).

Potential Pitfalls and How to Avoid Them

No solution is perfect. The team encountered several issues:

• Over-reliance on keywords: Some candidates padded their resumes with buzzwords (e.g., “expert in Python” after a one-week course). The bot could not detect depth of experience. Mitigation: The team added a mandatory coding test (via a simple Telegram quiz) for technical roles.
• Privacy concerns: Candidates were hesitant to share resumes via a bot. The team added a clear privacy policy at the start of the chat and ensured data was stored encrypted (AES-256).
• API rate limits: Telegram imposes limits on bot messages (30 per second). For high-volume campaigns, the team implemented a queuing system with a 100-millisecond delay between messages.

Conclusion

The story of this Telegram bot is a powerful example of how low-code automation can transform a core business process. By replacing manual resume screening with a bot that parses, scores, and ranks candidates, the team achieved a threefold increase in hiring throughput while reducing recruiter workload by over 60%. The key takeaways are clear: identify your bottleneck, choose a platform with a rich API (Telegram in this case), and design a hybrid workflow that combines machine efficiency with human judgment.

For companies facing similar mass hiring challenges, this approach is replicable with minimal upfront investment. The Habr article provides a detailed roadmap that any technical team can adapt. As recruitment continues to evolve, automation tools like this Telegram bot will become standard—not just for large enterprises, but for any organization that values speed and scalability.

If you are interested in building a similar system, start by mapping your current hiring process, identifying repetitive tasks, and prototyping a bot using Telegram’s API and open-source NLP libraries. The results may surprise you.

Source