Designing the AI relationship nurturing tool for a recruiting agency from concept through to the launch of version 1
Clear Match Talent
UX Lead & Technical Product Strategist partnering with engineering to define product features, AI logic, and a clean database structure to support future AI capabilities.
The problem
Even the best recruiters were losing touch with warm leads. Strong relationships were going cold simply because they weren’t being nurtured consistently.
“I have hundreds of warm contacts, but I only reach out when there’s a job. It’s hard to keep things warm in between.”
Agency Recruiter
That became our lightbulb moment.
The product vision
To build an AI-powered relationship management tool integrated with HubSpot—one that helps recruiters remember, reconnect, and rekindle relationships without losing their human touch.
The end goal
To increase revenue through enhanced engagement deal conversion rates.
Target Users: Agency Recruiters
What they do:
Manage extensive contact networks, needing to prioritize daily activities for maximum revenue impact.
Juggle hundreds of candidate and client relationships while working under pressure to close deals quickly. They struggle to identify which opportunities deserve immediate attention among competing priorities.
How this AI tool helps them
CMT AI acts as their daily command center—automatically surfacing the highest-probability revenue opportunities so recruiters can focus their limited time where it matters most.
Users’ pain points I want to address
My discussions with several recruiters revealed these critical pain points I aim to solve.
Hard to keep hundreds of relationships warm
Difficult to keep track of all the contacts, which includes candidates and clients
When there’s new job opening, the outreach to the client feels cold, making it harder to secure the deal.
I boiled it down to one problem statement:
How do we nurture relationships in a way that leads to more wins?
How do we address this problem statement?
First we need to address how to objectively quantify relationship engagement
Secondly we need to figure out how to prioritize which business opportunities to pursue
We introduced a formula:
Engagement Score × Alert Rating = Task Criticality
How does this help recruiters again?
This helps recruiters prioritize outreach where they already have warm connections—maximizing the chance of a successful outcome.
Engagement Score
× Alert Rating = Task Criticality
1 - Dissecting the first part of the equation: engagement score
I started by asking users what they were struggling with specifically in managing hundreds of contacts; what do they wish they can quantify?
I found out they were experiencing:
Challenges keeping track of candidates’ career movements
Insincere outreach when only reaching out to clients when a job opening is listed
Difficulty keeping track of meaningful conversations
What recruiters are saying
So how exactly do we quantify engagement levels?
To address gaps in how users measured relationship health, I developed scoring methods and tested them against the six most common contact profiles identified by users, ensuring accuracy and usefulness.
Scoring metrics examples:
Level of connection (5 = I’d grab beer with, 1 = avoid)
Profile type (Lead role = 2 or technical = 1)
Total past partnerships
Total past jobs with successful close together
Response rate (average days for a response)
Test result: Automation will be challenging initially
Test results showed the scoring system would be highly valuable, but challenging to apply accurately across hundreds of candidates initially. A manual approach would be required early on before automation becomes feasible.
Engagement Score x
Alert Rating
= Task Criticality
2 - Dissecting the second part of the equation: alert rating
How to rate alerts?
Users identified various alerts (industry news, job openings etc) prompting nurturing actions (emails, calls).
I decided to complete a card sorting activity since there is a large set of general alerts. I wanted to understand their logical categories and identify natural data sources. My goals are to validate whether certain alerts relate to business development versus personal notifications (like birthday reminders), and to discover priority or urgency levels that emerge organically from user groupings.
Card sorting activity validated my assumptions
Card sorting activity showed two distinct groups of alerts
Business dev alerts (directly tied to revenue): job opening, M&A activity, company lay off
Personal alerts (relationship maintenance that improve conversion rates): birthday, holiday, looking for work
Conclusion from card sorting: need to focus on one alert
After showing users the volume of potential alerts, we realized the need to narrow the scope for V1. We chose to focus on job opening alerts.
The card sorting activity validated my assumption, that alerts fall into two distinct groups: those directly tied to revenue-generating activities and those focused on relationship nurturing with indirect revenue impact.
We’ve defined the equation, but what’s feeding the data into it?
As a product lead, I worked with the engineering team to ensure the database schema could support features like engagement scoring, alert prioritization, and task automation—without overcomplicating the experience.
My databased creation process:
Designed a database schema supporting engagement scoring and task automation
Audited existing HubSpot data to identify reusable fields and new requirements
Mapped manual vs. automated data inputs to streamline user experience
The existing systems map to three new databases that will feed the algorithm and support future AI capabilities.
Prototype showing product vision with AI capabilities
(input description)
(input description)
(input description)
Usability testing - What I evaluated
I asked users to open the dashboard as they would on a typical workday and try to find the information they’d use to nurture their relationships.
I wanted to find out:
Can they easily identify which relationships to prioritize?
Is the dashboard content clear, relevant, and easy to scan?
Is the interaction timeline easy to locate and helpful for guiding next steps?
Promising Test Results
Increase number of meaningful touch points by 100% each week
4 hours saved each week bby reducing time spent chasing cold or low-potential lead
East to scan and navigate - dashboard praised for its clarity and visual appeal
While the interaction timeline was seen as helpful, users felt it’s not essential at this stage of the workflow.
Key Takeaways
This project reinforced the critical importance of grounding algorithm design in real user workflows. By collaborating closely with agency recruiters, I learned that productivity tools succeed when they seamlessly integrate into existing habits rather than requiring behavioral change.
The most valuable insight was discovering how relationship strength and opportunity urgency could be mathematically combined to drive daily prioritization—transforming subjective decision-making into data-driven action.
Moving forward, I'll continue prioritizing user research in algorithm development, ensuring that complex backend calculations translate into intuitive, actionable insights that directly impact business outcomes.
