iGB@ICE 2026: Dr Eyal Loz on AI, simplicity, and player psychology in slot development
Dr Eyal Loz, chief product officer at RubyPlay, sat down with presenter Katie Goldfinch at the iGB@ICE Studio to talk straight about what actually works in slot development. His core message: simple concepts, shipped often, with strong quality control beat over-designed systems that try to do everything.
AI is speeding up production, but it's not the source of creative judgment. As Loz put it, "We should really make the effort not to let AI take our voice away."
Simplicity beats over-engineering
High-performing games usually come from a clear core loop, not a tangle of mechanics. Removing extras sharpens the emotional hit players feel and makes the game easier to learn, repeat, and love.
- Define the core loop in one sentence. If you can't, it's too complex.
- Cut features that don't increase anticipation, clarity, or reward.
- Ship small, testable updates. Let real player data decide what stays.
- Instrument everything: session length, hit frequency, volatility perception, churn triggers.
Volume, iteration, and quality control
Loz argues success comes from consistent release cadence and ruthless iteration, not from chasing a single "perfect" product. More shots on goal, with strong QA and live-ops discipline, outperforms sporadic big bets.
- Prototype weekly. Kill weak ideas early with clear gates.
- Use feature flags and server-configurable parameters for fast tuning.
- Automate regression tests for math models, RTP ranges, and edge-case payouts.
- Run post-release reviews within 72 hours to adjust volatility, bonus frequency, and messaging.
AI as an accelerator - not a replacement
Generative tools can crank out art, animation, and variants faster, especially for event- or brand-driven content. But the team's taste and decision-making still determine whether players care.
- Keep tight art direction docs; lock color palettes, composition rules, and character guidelines.
- Human-in-the-loop checkpoints for style consistency, readability, and emotional tone.
- Track asset throughput and review time saved, not just "number of assets produced."
- Maintain dataset hygiene and legal checks for licensed IP and likenesses.
Data over opinions
"Watch what players do, not what they say." Behavioral data beats self-reported preference. Prioritize signals that map to emotion and momentum inside the session.
- Time to first thrill (first mini-win, first near-miss, first bonus tease)
- Clarity events (tutorial completion, UI hint engagement, help-page hits)
- Frustration moments (rage-quits after dead spins, rapid bet-size drops)
- Return cadence (D1/D7 retention tied to feature discovery, not just RTP)
For a solid primer on behavioral vs attitudinal data in UX, see this overview from NN/g Behavioral vs. Attitudinal.
For teams building slots in 2026
- Run discovery→prototype→live test in 2-4 week loops. Keep backlog grooming brutal.
- Plan an event-based content calendar (sport finals, holidays, pop culture drops) and prebuild reskins with flexible math.
- Segment by motivation (thrill-seekers vs collectors) and tune volatility curves per cohort.
- Adopt an experimentation platform for A/B/C tests on paytable tweaks, bonus cadence, and UI prompts.
- Document kill criteria. If a feature misses its metric by X% after Y users, pull it.
Watch the session
Catch more conversations from ICE Barcelona 2026 on the iGB YouTube channel here.
If you're refining AI workflows for art and animation pipelines, this curated list of tools can help AI tools for generative art.
Bottom line: Keep the game loop simple, ship often, let data call the shots, and use AI to go faster without losing your creative voice.
Your membership also unlocks:
