Scaling Job Postings via Platform Intelligence
Reducing manual operations through AI-assisted workflows
InfoEdge Hackathon (48 hours)
AI-Assisted Workflows
Context
Increasing job postings from recruiters with active subscriptions was a core business OKR for NaukriGulf, directly tied to active database growth and renewal likelihood.
Despite multiple initiatives, a significant portion of job recovery still depended on manual client servicing workflows—creating a bottleneck that limited progress against this goal.
My role was to identify structural inefficiencies in this process, reframe the problem beyond operational execution, and explore whether a system-level approach could unlock scale.
Constraints & Operating Reality
The problem existed under several non-negotiable constraints:
💳
Recruiters were cautious about consuming paid credits
👥
Client servicing bandwidth was limited and already stretched
🔍
Hiring intent existed outside the platform, but detection was slow and manual
Any viable solution had to:
✔️
Reduce human dependency
✔️
Preserve recruiter trust and control
✔️
Integrate into existing workflows rather than replace them
Ground Truth
To understand the problem beyond metrics, I reviewed the client servicing workflow directly.
Observed Reality
🕵️
Career sites and LinkedIn were monitored manually
🔄
Jobs found externally were mapped back to NaukriGulf to avoid duplication
📧
Recovery required multiple emails / scripted persuasion / free job credits
Despite this effort
❌
90% recruiter declined
❓
Approval authority was often unclear
⏳️
Time to convert stretched into weeks
Core Insights
The constraint was not demand.
Recruiter intent, candidate supply, and historical behavior already existed — but the system relied on humans to connect these signals.
This reframed the problem from sales execution to platform intelligence.
Strategic Decisions
🤖
Shift ownership from humans to the platform
Rationale: Manual pattern recognition does not scale and delays response to hiring intent
Trade-off: Accept partial coverage in exchange for higher confidence and speed
☑️
Validate value before
prompting action
Rationale: Recruiter trust erodes when nudges are not backed by supply confidence
Trade-off: Fewer prompts, but significantly higher relevance
👥
Treat client servicing as a primary user
Rationale: Any solution that increased their cognitive or operational load would fail
Trade-off: Reduced configurability in favor of operational simplicity
System Design Response
We designed a multi-agent AI system, where each agent addressed a specific failure point in the existing process.
🔎
Job Discovery Agent
Detects jobs posted on career sites and competitor platforms that are missing on NaukriGulf.
🧠
Client Insight Agent
Analyzes recruiter history to infer posting preference, credit availability, and likelihood of action.
🛡️
Supply Validation Agent
Confirms candidate availability to avoid low-value recommendations.
🤖
Decision & Pitching Agent
Synthesizes insights to decide:
✔️
Whether to prompt at all
✔️
What action to recommend
✔️
Why the recommendation is relevant
Reflection
This work reinforced a consistent pattern in my approach:
Scalable impact comes from re-architecting how intent, insight, and action connect — not from adding features.
Impact
For recruiters, this meant:
✔️
Reduce job detection latency from weeks to hours
✔️
Increase utilization of existing paid credits
✔️
Lower dependence on manual client servicing effort
✔️
Improve likelihood of recruiter reactivation and renewal
Scaling Job Postings via Platform Intelligence
Reducing manual operations through AI-assisted workflows
InfoEdge Hackathon (48 hours)
AI-Assisted Workflows
Context
Increasing job postings from recruiters with active subscriptions was a core business OKR for NaukriGulf, directly tied to active database growth and renewal likelihood.
Despite multiple initiatives, a significant portion of job recovery still depended on manual client servicing workflows—creating a bottleneck that limited progress against this goal.
My role was to identify structural inefficiencies in this process, reframe the problem beyond operational execution, and explore whether a system-level approach could unlock scale.
Constraints & Operating Reality
The problem existed under several non-negotiable constraints:
💳
Recruiters were cautious about consuming paid credits
Client servicing bandwidth was limited and already stretched
👥
🔎
Hiring intent existed outside the platform, but detection was slow and manual
Any viable solution had to:
✔️
Reduce human dependency
✔️
Preserve recruiter trust and control
✔️
Integrate into existing workflows rather than replace them
Ground Truth
To understand the problem beyond metrics, I reviewed the client servicing workflow directly.
Observed Reality
🕵️
Career sites and LinkedIn were monitored manually
🔄
Jobs found externally were mapped back to NaukriGulf to avoid duplication
📧
Recovery required multiple emails / scripted persuasion / free job credits
Despite this effort
❌
90% recruiter declined
❓
Approval authority was often unclear
⏳️
Time to convert stretched into weeks
Core Insights
The constraint was not demand.
Recruiter intent, candidate supply, and historical behavior already existed — but the system relied on humans to connect these signals.
This reframed the problem from sales execution to platform intelligence.
Strategic
Decisions
🤖
Shift ownership from humans to the platform
Rationale: Manual pattern recognition does not scale and delays response to hiring intent
Trade-off: Accept partial coverage in exchange for higher confidence and speed
☑️
Validate value before
prompting action
Rationale: Recruiter trust erodes when nudges are not backed by supply confidence
Trade-off: Fewer prompts, but significantly higher relevance
👥
Treat client servicing as a primary user
Rationale: Any solution that increased their cognitive or operational load would fail
Trade-off: Reduced configurability in favor of operational simplicity
System Design Response
We designed a multi-agent AI system, where each agent addressed a specific failure point in the existing process.
🔎
Job Discovery Agent
Detects jobs posted on career sites and competitor platforms that are missing on NaukriGulf.
🧠
Client Insight Agent
Analyzes recruiter history to infer posting preference, credit availability, and likelihood of action.
🛡️
Supply Validation Agent
Confirms candidate availability to avoid low-value recommendations.
🤖
Decision & Pitching Agent
Synthesizes insights to decide:
✔️
Whether to prompt at all
✔️
What action to recommend
✔️
Why the recommendation is relevant
Experience Principles
The experience was deliberately designed around four principles:
☘️
Explainability over automation
Recruiters understand why an action is recommended.
☘️
Control over enforcement
The system recommends; the recruiter decides.
☘️
Low friction, not zero friction
Actions are simplified but never forced.
☘️
In-flow delivery
Recommendations surface within existing recruiter workflows and outreach channels without altering operating models.
Impact
Although developed as a concept, the system demonstrated potential to:
✔️
Reduce job detection latency from weeks to hours
✔️
Increase utilization of existing paid credits
✔️
Lower dependence on manual client servicing effort
✔️
Improve likelihood of recruiter reactivation and renewal
Reflection
This work reinforced a consistent pattern in my approach:
Scalable impact comes from
re-architecting how intent, insight, and action connect — not from adding features.