How We Built Recruit 360 AI Assistant

Six Agents working together on google cloud.

We Gave Recruiters a Colleague Who Has Read Every Single Record. Here’s How We Built Recruit360 on Google Cloud.

Ask any recruiter what slows them down, and it’s rarely a shortage of data. It’s the opposite.

A mid-sized recruitment platform sits on thousands of candidate profiles, visa records, billing entries and placement histories. Somewhere in there is the perfect candidate for the role that opened this morning. Somewhere else is a visa rejection that nobody has noticed yet.

Finding either means scrolling, filtering, exporting to a spreadsheet, and more often than anyone likes to admit – pinging a technical colleague to write a query. By the time the answer arrives, the moment has passed.

That was the brief we took on with Recruit360. Not “add AI to the platform.” Rather: let a recruiter ask a question in plain English and get the right answer in seconds.

The Problem In Three Sentences

• Too much data. Thousands of records that no human can scan by hand.
• Answers take hours. Every question meant manual searching and spreadsheet gymnastics.
• Things slip through. Rejected visas and urgent cases surfaced only when it was already too late.

The cost wasn’t just time. It was missed placements, missed deadlines, and recruiters spending their day on retrieval instead of relationships.

The Solution: One Assistant, Six Specialised Agents

Instead of a single monolithic chatbot, we designed the Recruit360 AI Assistant as an orchestrated team of six AI agents, each one an expert at a specific job:

Agent	What it does
Ask your data	Turns a plain-english question into an exact database answer (text-to-SQL)
Smart Matching	Finds candidates by meaning, not keywords using embedding and vector search
Predict Placements	A machine-learning model that ranks who is most likely to be placed
Visa Fix-It	Reads a visa rejection and produces a step-by-step plan to re-apply
Urgency Watch	Automatically flags the most urgent and at-risk candidates
Auto Profiles	Reads a resume or ID document and fills in the candidate profile

A recruiter can type something like: “Which candidates are most likely to be placed, and who has a rejected visa?” and get both answers, instantly, from live data.

How It Works Under The Hood

We built the entire stack on Google Cloud:

• The recruiter asks a question in natural language.
• LangChain interprets the intent and routes it to the right agent.
• Vertex AI (Gemini) understands the request and writes the precise query.
• BigQuery runs it against real data, with BQML powering the placement-prediction model.
• The answer comes back – exact, explainable, and in seconds.

Alongside these, Document AI handles résumé and ID extraction, and Data Studio gives leadership a live view of the pipeline.

The design principle throughout was simple: the recruiter should never need to know any of this. They ask; it answers.

Before And After

Before	After
Scroll through thousands of records	Ask one question -> instant answer
Hours building reports by hand	AI sorts, matches and predicts in seconds
Miss visa rejections and urgent cases	Urgent and at-risk candidates flagged automatically
Every question needs a technical person	Any recruiter gets answers themselves

That last row matters most to us. The real shift isn’t speed it’s who gets to ask. When every recruiter can query the entire platform in their own words, the whole team levels up, not just the person who knows SQL.

What’s Next

We’re now moving Recruit360 into production:

• Live data through Cloud SQL with Datastream sync
• WhatsApp access on the official Meta API, with the agents hosted on Cloud Run – so recruiters can ask questions from their phone, mid-conversation with a client
• A full GCP web UI with the assistant built in, plus continued testing against real résumés and real placement outcomes

Why We’re Sharing This

At Avanciers, we believe the most useful AI isn’t the flashiest. It’s the kind that quietly removes the friction between a person and the answer they need — and does it reliably, on infrastructure that scales.

Recruit360 is one example. Recruitment happens to be the domain, but the pattern – natural-language access to complex operational data, delivered by specialised agents on Google Cloud applies just as well to logistics, healthcare admin, finance operations, or any team drowning in records.

If your team spends more time finding information than acting on it, we’d love to talk.