AI in talent assessments
AI in talent assessments: RPG Group's experience building Ascent from zero
RPG Group and AltUni Labs built RPG Ascent from the ground up as an AI-first talent property, designed to engage high-potential talent earlier in the campus calendar than the recruitment season allows. A five-phase, time-boxed simulation built around RPG's own competency framework produced the shortlist, which went into a two-day live virtual immersion with RPG's senior leadership. Registrations landed 156% above target in the inaugural edition. InsideIIM and AltUni Labs delivered the property end to end, covering branding, microsite, outreach, the AI stack and the immersion itself.
156%
Above registration target in the inaugural edition
5 phases
One timer, one competency framework
2 days
Live virtual immersion with RPG's senior leadership
Built from zero
Platform, branding, campaign, assessment and immersion in a single season
Problem: By the time the season opens, candidate preferences are already formed
RPG's brief was not to hire faster. It was to be present earlier.
By the point campus recruitment formally opens, candidates have already decided which employers they want. A job posting arriving in a crowded season competes for attention against every other posting in the same week. Three things followed from that.
- Familiarity has to be built upstream. Preference forms months before Day Zero, which means engagement has to happen before the calendar gets crowded, not during it.
- Candidates enter properties; they respond to postings. The engagement had to be something a candidate would want to take part in on its own terms, not another application form.
- A first edition has no track record to trade on. With no prior cohort, no alumni and no benchmark, the property had to earn its registrations on the strength of the proposition alone.
Solution: An AI-first talent property built around RPG's competency framework
InsideIIM and AltUni Labs designed and delivered the whole property. The AI sat at the centre of the selection layer rather than being bolted onto it, which meant the assessment and the scoring framework were designed together rather than sequentially.
The five-phase simulation
Five phases, one 45-minute timer, built on SignalAI.
Pacing is itself a signal. With a single timer running across all five phases, how a candidate allocates their time becomes part of the evidence rather than an administrative detail.
Situational judgement, in-basket and strategy. Business acumen tested across a connected multi-stage challenge rather than isolated question sets, with reasoning captured at every choice, so candidates were evaluated on their thinking, not only on their selection.
How the shortlist was built
Profile screening ran through PotentialAI against RPG's competency framework. Each profile returned:
- A score and a tier classification
- Tier reasoning in plain language
- Competency-level evaluations with evidence quoted directly from the resume
- A recommendation for the hiring manager, including cases where a candidate was strong but mismatched to the role level, stated explicitly rather than scored down silently
The two-day virtual immersion
The shortlist went into a full two-day virtual experience with RPG's senior leadership, which our teams ran end to end: hosting platform, breakout rooms, live communications, student operations and intake.
Sessions ran across a masterclass, the journey of a student who had joined RPG as a summer-internship hire, what to expect from RPG, and a live simulation. RPG's leadership showed up for their sessions; everything around them was run.
What InsideIIM and AltUni Labs delivered
- Branding and microsite. Logo, creatives and content on a dedicated microsite hosted on InsideIIM.
- Customer success and outreach. Driving registrations and coordinating students and the RPG team through the cycle.
- Tech and AI. The simulation and the screening framework, built from scratch to RPG's brief.
- Social media outreach. Reels, shorts and carousels to build momentum around the property.
- Live immersion. Hosted, moderated and run end to end.
Impact
- Registrations landed 156% above target in the very first edition
- A high-quality AI-led shortlist progressed into the two-day leadership immersion
- One consistent, objective evaluation standard held while the experience scaled
- RPG Ascent established as a differentiated talent engagement platform from day one
- Talent engaged months ahead of the recruitment season, arriving at the hiring window already familiar with RPG
Before AI and after AI
Engagement timing
- Conventional approach
- Job postings during a crowded recruitment season
- RPG Ascent
- A property candidates enter months before Day Zero
Screening basis
- Conventional approach
- Resume review and keyword filtering
- RPG Ascent
- Profile evaluation against RPG's competency framework, with evidence quoted from the resume
Assessment
- Conventional approach
- Isolated question sets
- RPG Ascent
- Five connected phases under a single timer, with pacing as evidence
What gets measured
- Conventional approach
- The answer selected
- RPG Ascent
- The reasoning behind every choice
Hiring manager output
- Conventional approach
- A ranked list
- RPG Ascent
- Score, tier, plain-language tier reasoning, competency evaluations and a recommendation
Mismatched candidates
- Conventional approach
- Quietly scored down
- RPG Ascent
- Flagged explicitly as strong but mismatched to the level
Final round
- Conventional approach
- A standard interview loop
- RPG Ascent
- Two-day virtual immersion with senior leadership, run end to end
| Area | Conventional approach | RPG Ascent |
|---|---|---|
| Engagement timing | Job postings during a crowded recruitment season | A property candidates enter months before Day Zero |
| Screening basis | Resume review and keyword filtering | Profile evaluation against RPG's competency framework, with evidence quoted from the resume |
| Assessment | Isolated question sets | Five connected phases under a single timer, with pacing as evidence |
| What gets measured | The answer selected | The reasoning behind every choice |
| Hiring manager output | A ranked list | Score, tier, plain-language tier reasoning, competency evaluations and a recommendation |
| Mismatched candidates | Quietly scored down | Flagged explicitly as strong but mismatched to the level |
| Final round | A standard interview loop | Two-day virtual immersion with senior leadership, run end to end |
Key takeaways for talent teams
Common challenges
- Candidate preference forming before the recruitment season opens
- No track record or benchmark to build a first-edition target against
- Resumes that reveal credentials but not judgement
- AI-scored shortlists hiring managers cannot interrogate
- Strong candidates mismatched to a role level, filtered out without explanation
- Leadership time available for the final round, but no team to run everything around it
Practical guidance
- Design the property AI-first rather than adding AI to an existing process; the assessment and the scoring framework should be built together.
- Set first-edition targets off campus penetration and comparable properties in the same sector and cohort, not off instinct.
- Use a single timer across multiple phases if pacing and prioritisation are competencies you care about.
- Capture reasoning at each decision point, so the shortlist arrives with an explanation attached.
- Make the mismatch case explicit. A candidate who is strong but wrong for the level is useful information, not a low score.
- Engage months ahead of the hiring window. Candidates who have spent five phases inside your business context and two days with your leadership arrive already decided about you.
Ready to build an AI-first talent property?
We design and execute campus talent properties end to end, combining outreach, branding, AI-powered assessments and interviews on a single stack, delivered with speed, precision and measurable outcomes at scale.
