We built an AI skills platform that maps employees' real skills from their work and recommends internal career moves, with a logged reason for each. Attrition fell 22%, and 45% of roles were filled internally.
Client: Citadel AutobiographyIndustry: HR TechService: AI & Automation
Overview
Client
Citadel Autobiography
Distributed workforce of 5,000+ people across 12 regions
Engagement snapshot
The challenge
The client found internal candidates through static CVs and yearly performance reviews. Those snapshots missed what people had learned since, and 60% of technical employees felt their skills were underused, so many looked for jobs elsewhere.
Skill records were dropdown lists updated once a year. Internal moves often depended on which managers knew whom.
Long-serving engineers left because they could not see a way to grow, taking key system design knowledge with them.
The company paid agency fees for roles where current staff already had an 80% skill match.
Tech stacks changed so quickly that HR profiles were out of date within 6 months of a hire.
The solution
AI agents read project commits, Jira data and feedback logs to build a current picture of each person's skills, beyond what their CV says.
We built a skills graph in Neo4j that maps related skills, such as Python to data engineering, to predict which internal moves are likely to work.
Each career recommendation is saved with a match-reason code, which supports HR compliance and helps guard against algorithmic bias.
Key features
AI agents read project commits, Jira data and feedback logs to build up-to-date skill profiles that refresh as people work.
Business impact
Matches suggested in 1.8 seconds
A skills graph predicts successful internal moves, with demographic data removed and a reason for every match.
Business impact
Internal fill rate up from 12% to 45%
The system watches Slack, GitHub and Jira activity and updates employee profiles as people work.
Business impact
Zero manual updates
Demographic data is removed before the first round of matching, which keeps candidate shortlists diverse.
Business impact
100% DEI compliance
Agents tell employees about internal roles that fit their path before they start looking on LinkedIn.
Business impact
22% drop in attrition
Before and after
Before
Fixed list
Rigid dropdowns could not capture cross-functional work or new technologies.
After
Living skill map
Links between skills change as the graph learns from the work people actually do.
Before
Manager bias
Internal hiring depended on managers' favourites and who happened to be visible.
After
Matching on evidence
Data-based matching with full audit trails, so moves are based on merit.
Before
Annual review
Skill data was updated once a year, so it was always out of date.
After
Live sync
Profiles update automatically when a project finishes or a peer confirms a skill.
Results
22% Turnover Drop
Employees with a clear path to their next role stayed longer, especially in high-risk engineering teams.
Before: HighAfter: 22% Lower
45% Internal Fill Rate
Better visibility let the business fill complex roles internally, saving millions in agency fees.
Before: 12%After: 45%
1.8s Query Speed
Tuned graph queries give managers matches almost instantly when they staff a new project.
Before: < 5sAfter: 1.8s
Testimonial
“Our internal fill rate was 12%. It is now 45%, and attrition is down 22%. Skill profiles refresh as people work, demographic data is removed before matching, and each suggested move comes with a reason code.”
Joyce GutierrezDirector of Internal Mobility, Citadel AutobiographyKey results
Trust and control
Personal information is masked in all project data collection, so feedback stays confidential.
PrivacyTraceable reason trails let third-party auditors review recommendations for bias.
AuditableCloud-based graph clusters keep 99.9% uptime during global traffic spikes.
High availabilityThe client owns 100% of the skill map and custom career-path logic.
100% ownedFaster delivery
Starting from tested components saved about 10 weeks, leaving more time for the client's own skill map and matching rules.
Pulled activity from Slack, GitHub and Jira into the skills graph through tested connectors.
Gave the AI agents a tested way to pick out skills from commits, tickets and feedback.
Showed HR how many suggestions led to internal moves, by team and region.
Compared internal fills with the agency fees they replaced.
Internal candidates were found through static CVs and yearly reviews, so managers could not see current skills. Of the company's technical employees, 60% felt their skills were underused and looked outside. Meanwhile, the company paid agency fees for roles that had an 80% skill match with people it already employed.
AI agents read project commits, Jira data and feedback logs to build a current skill profile for each employee. A Neo4j graph maps how skills connect, such as Python to data engineering, and suggests realistic next roles. Employees hear about matching openings early, and managers get matches in under 2 seconds.
Demographic data is removed before the first round of matching, so candidates are ranked on skills. Every recommendation is logged with a match-reason code, and these trails can be reviewed by third-party auditors. Personal information in project data and feedback is masked to keep it confidential.
If you are paying agencies for skills your own people have, we can help you build a live skills map that shows who is ready for their next move.