Different hiring criteria? TalentGPT now understands your priorities

Hello, this is the TalentSeeker Operations Team.

In this January update, we focused on a problem where each company (team) has different hiring criteria—but those criteria were not being reflected clearly in recommendation results.

Specifically:
“The recommendation priority is ranked differently than the criteria our team has in mind.”

The core of this update is simple.
TalentGPT now reflects priorities more accurately based on the conversation context, and if needed, users can change the priority order themselves to re-rank results.

Problem: “A is the most important criterion for us, but TalentGPT doesn’t seem to think so.”

Every hiring team has a different definition of a “good candidate.”

• Some teams value hands-on, real job execution experience the most.
• Others prioritize proven experience actually using specific skills/technologies.
• And some teams check domain experience first.

Previously, TalentGPT would aggregate criteria and sort results based on internal ranking logic. As a result, the criteria a user cared about most were not always reflected strongly enough at the top of the list.

That led to confusion such as:
“Everything matches—so why is this person ranked higher?”

Voice of the customer: “I want to see this criterion first.”

These are the kinds of requests customers actually left in TalentGPT:

• “Find candidates with this criterion as the top priority.”
• “Even if the years of experience are a bit short, I want to see people with this experience first.”
• “The tech stack matches—why aren’t candidates with domain experience showing up at the top?”
• “Candidates who match the criterion I care most about aren’t visible in the results.”
• “Looking at the results, the priority order seems different from what I expected.”

From these questions, we came to one clear conclusion:
Before “better recommendations,” what we needed first was “recommendations that understand my criteria.”

Our hypothesis: Trust starts with alignment on ranking criteria

For this improvement, we set two hypotheses:

1. What matters in talent recommendations is not the single “correct answer” (the one best candidate), but reaching alignment with the user on “what to look at first for this role.”
2. For AI recommendations to earn trust in real recruiting work, it’s more important than “showing more results” to ensure users can control the priority (weights) or adjust them immediately.

In other words, recommendations shouldn’t be “what the AI decides.” They should function as a tool that accurately follows the user’s hiring criteria.

Update: Candidate Priority Controls

With this January update, we added Candidate Priority Controls to TalentGPT. There are two key parts.

1) Priorities are automatically reflected from the conversation

Based on signals such as emphasized wording, repeated criteria, and intent like “this matters more,” TalentGPT interprets the conversation context and automatically adjusts which criteria to prioritize for this search.

For example:

• “We can be flexible on seniority, but domain experience is the most important.”
• “The tech stack is similar, but I want to see candidates with real operations/launch experience first.”

In these cases, those criteria are more strongly reflected in top-ranked results.

2) If needed, users can change priorities manually

If the priority order suggested by the AI doesn’t fit, or if hiring needs change, users can reorder criteria directly to re-rank candidates.

Priority criteria included in this update

The criteria that can be used to adjust ranking priority are:

1. Role Fit: Alignment between the role and what the candidate has actually done in practice
2. Core Role Competencies: Not just keywords, but real execution capability for the skills needed on the job
3. Seniority Level: Level based on both years and scope of responsibility
4. Domain Experience: Practical experience in a specific industry or organizational environment
5. Background Information: Supporting context such as education, major, and training history

TalentGPT has moved beyond “balancing all criteria equally” to providing ranking that prioritizes:
“What matters most for this specific hire right now.”

Workflow: What has changed?

1. Enter your hiring request in TalentGPT.
2. TalentGPT interprets the conversation and applies priority criteria to ranking.
3. If the results differ from what you expected, adjust priorities manually.
4. Review candidates faster by starting with the re-ranked top results.

This may look like a feature addition, but in practice it reduces:

• Time spent interpreting “Why is this person ranked higher?”
• The cost of repeated scrolling and re-searching
• The burden of explaining ranking results internally (ranking rationale)

Additional updates (Fix & Bug)

Along with feature improvements, we also resolved several issues:

• Fixed an issue where company information was not properly reflected in search.
• Fixed an issue where related companies were missing from TalentGPT search results.
• Improved candidate education/background information search (major, school, languages, etc.).
• Streamlined the JD advice feature by making guidance more concise and summary-focused.
• Restricted responses to non-HR requests such as “What’s the capital of Australia?”, “Today’s fortune,” or “A photo of Santa delivering gifts.”

What we learned: Hiring doesn’t have a single “right answer”—it has criteria

This update reinforced something important:
There is no single correct answer in hiring. And more important than “AI making good decisions” is ensuring it accurately follows the user’s hiring criteria.

TalentSeeker will continue to evolve—not as a tool that “sets the criteria,” but as one that quickly reflects the criteria users actually want.

What’s next

Candidate Priority Controls are only the beginning.

Going forward, we plan to improve the system so it can reflect more context—such as whether speed matters more right now, or whether precision matters more right now—depending on the hiring situation.

If you have feedback like “this criterion should matter more” or “this ranking felt off,” please share it anytime.
Your feedback becomes the basis for our next update.

Try the update in TalentGPT now.
If you’d like support for team rollout or best practices, feel free to reach out for a demo or inquiry.

Thank you.
TalentSeeker Operations Team