Official Rules — ELEV8 RACE — AI-Powered Marathon

1. What the ELEV8 AI-POWERED RACE is

ELEV8 is a two-month online marathon run across our ecosystem of companies. Teams build an AI-powered solution that creates real business value — increasing productivity, removing a bottleneck, or solving a concrete business problem. The scope is deliberately open: the problem can sit anywhere in the business.

Eight finalist teams travel to Limassol in late October to pitch on stage in front of a jury of company CEOs, followed by an awards gala. Three prizes are awarded.

The programme is organised and run by GR8_TECH. Every stage of the race — lectures, workshops, mentoring, pit stop, submission and finals — is announced through the landing page and the dedicated Slack channel.

2. Key dates

All dates below follow the ELEV8 communication calendar. Times are EEST (Cyprus).

Date	Milestone	What you need to do
5 Aug	Race announced across all companies	Read the announcement, start forming a team
6 Aug	Rules published + Slack channel opens	Join #elev8-race
5–15 Aug	Registration open	Submit your team registration form
10–17 Aug	15-minute pre-selection calls (rolling)	Book and attend your slot
15 Aug	Registration closes	Hard deadline — no late entries
17 Aug	Teams reveal — starting grid published	See who you are up against
18 Aug	Mentors assigned	Meet your mentor within 5 working days
13 Aug – 17 Sep	Lecture & workshop series	Attend live or watch the recordings
7 Sep	Pit stop — mid-term check-in	Post a one-paragraph progress update
25 Sep	Submission portal opens	Start uploading
30 Sep, 23:59	Submissions close	Hard deadline — portal locks automatically
2–5 Oct	Blind scoring by the selection committee	Nothing — stay available for clarifications
6 Oct	All teams notified personally	Check your email / DM
8 Oct	Public shortlist of 8 finalists announced	Finalists confirm travel details
22 Oct	Finalists arrive in Limassol, welcome kit	Travel
23 Oct	Pitch Day + Awards Gala	Pitch

3. Who can take part

3.1 Eligibility

Open to all employees of the participating companies, regardless of role, seniority or location.
Managers and C-level staff may participate as team members. A person may not both compete and sit on the selection committee or the jury.
Mentors, lecturers and the organising team may not compete.
Participation is voluntary and takes place on personal time. It is not a work assignment and no work-hours allowance is granted.

3.2 Team composition

Requirement	Rule
Team size	3 to 5 people. Teams of 2 or 6+ are not accepted.
Non-technical member	At least 1 team member must be a non-technical specialist (e.g. product, marketing, finance, ops, HR, support, legal, design).
Travel readiness	At least 2 team members must be able and willing to travel to Cyprus and present the team's result if shortlisted.
Cross-company teams	Allowed and encouraged. A team may mix people from any of the participating companies.
Solo entries	Allowed at registration. Solo entrants are placed in the matching pool and must join a compliant team before the 15 Aug close, or they are withdrawn.
One team per person	Each person may belong to exactly one team.
Team captain	Each team names one captain as the single point of contact for the organisers.

3.3 Roster changes

The roster is locked when the starting grid is published. After that, a member may be replaced only with the organisers' written approval, and only if the team still meets every requirement in 3.2. A team that drops below 3 members withdraws from the race.

3.4 Language and conduct

English is the working language of the race: registration, submission, pitch and all official communication.
Prototypes must not use real customer personal data, and must not contain confidential production data without prior written approval.
Existing internal projects already on a company roadmap may not be re-submitted as race entries. Building on top of one is fine if the new contribution is clearly separable.
Any cost of third-party AI tooling is the team's own unless the organisers provide credits.

4. Registration

Registration runs 5–15 August via the form on the ELEV8 landing page. To register, a team captain submits:

Team name and captain contact details.
A short idea description — 2–3 sentences covering the problem, who has it, and how AI is involved. This is the basis for the pre-selection call, not a graded submission.

Registration is only complete once the pre-selection call has been passed. Teams that register but do not attend a call by 15 August are not admitted.

5. Pre-selection call — the entry gate

Every registered team attends one 15-minute video call, held on a rolling basis between 10 and 15 August. Calls are run by the GR8_TECH core team together with an external mentor. Slots are booked through the link in the registration confirmation.

The call is a gate, not a competition. Its only purpose is to confirm the team can meaningfully compete. Three checks are applied:

#	Check	What we are looking for
1	Business value	The idea addresses an identifiable business problem or productivity gap, with a plausible beneficiary and a plausible measure of improvement. “Interesting technology looking for a problem” does not pass.
2	Genuine AI component	AI is central to how the solution works, not a label on top of conventional automation. The team can describe, at a high level, what the model or agent actually does.
3	Team composition	3–5 members, at least 1 non-technical specialist, at least 2 able to travel to Cyprus and present.

5.1 Outcomes

Outcome	Meaning
PASS	The team is admitted and appears on the starting grid.
CONDITIONAL PASS	One check is not yet met but is fixable — for example the team lacks a non-technical member, or the business value needs sharpening. The team has until 17 August to remedy it and confirm by email. One follow-up call may be booked if needed. Unremedied conditional passes lapse.

Feedback is given verbally on the call and confirmed in writing within one working day. The organisers' decision on admission is final.

6. Running the race

6.1 Support provided

A mentor assigned to each team from 18 August, available for the duration of the race.
A lecture and workshop series running 13 August – 17 September: AI trends, vibe coding and rapid prototyping, LLM fundamentals, prompt engineering, AI in iGaming, prototype-to-product, and pitching and storytelling. All sessions are recorded and published.
A dedicated Slack channel for questions, matchmaking and announcements, plus the landing page as the single source of truth.

6.2 Pit stop — 7 September

Every team posts a one-paragraph progress update by end of day on 7 September: what you are building, where you are, and what is blocking you. The pit stop is mandatory but not scored. Teams that miss it are contacted directly; a second miss is treated as withdrawal.

7. Final submission

The portal opens 25 September and closes 30 September at 23:59 EEST. The deadline is hard — the portal locks automatically and late entries are not scored.

A complete submission has three parts. All three are mandatory; an incomplete submission is not forwarded to the committee.

#	Deliverable	Specification
1.1	Demo video	Maximum 3 minutes. MP4, 1080p, max 500 MB. In English or with English subtitles. Must show the prototype actually working — screen recording, not slides. Anything after the 3-minute mark is not watched.
1.2	Working prototype	A live link, repository, or sandbox environment with access instructions and any credentials the committee needs. Access must remain valid until 10 October. If the prototype cannot be shared externally, provide a guided walkthrough recording in addition to the demo video.
2	Business case one-pager	One page, PDF, using the template on the landing page. Covers: the problem and who has it; the solution and the AI approach; quantified impact (time saved, cost, revenue, error reduction); what it would take to roll out across the company; and what the team would do next with 3 more months.

Submissions are anonymised before scoring

Do not put team names, member names, company names or logos in the one-pager filename or content. The organisers assign each submission a Team Code and strip identifying details before the package goes to the committee.

Demo videos cannot always be fully anonymised. Committee members who recognise a team must declare it — see Section 8.3.

8. Shortlisting the 8 finalists

8.1 The selection committee

The committee is made up of 3 representatives nominated by each participating company. All submissions are scored blind, using Team Codes only. Every committee member scores every submission.

8.2 Scoring criteria

Each criterion is scored 1–10. The team's score is the weighted average across all committee members.

Criterion	What a 10 looks like	Weight
Business value & impact	A real, named business problem with a credible and quantified improvement. The value is obvious to someone who does not work in that function.	40%
AI depth & appropriateness	AI is central and the approach is well chosen for the problem. The team understands why this method and not a simpler one, and is honest about its limits.	30%
Scalability & adoption	Could realistically be rolled out beyond the team's own function or company. Integration, ownership and change-management are thought through.	30%

Score band	Interpretation
9–10	Outstanding — would advance on this criterion alone
7–8	Strong — clearly above the field
5–6	Solid — meets expectations, nothing exceptional
3–4	Weak — significant gaps
1–2	Does not address the criterion

8.3 Integrity of the process

Committee members score independently and do not discuss submissions before all scores are in.
If a committee member recognises a team, they declare it. Their score for that team is excluded and the team's average is calculated from the remaining scorers, so no team is advantaged or disadvantaged by recognition.
Scores are collected by the organisers and revealed to the committee only after the scoring window closes.
Any member who has coached, mentored or otherwise materially helped a team must recuse themselves from scoring that team.

8.4 Ranking and ties

The 8 highest weighted averages are shortlisted. If a tie affects the 8th place, it is broken in this order: (1) higher Business value & impact score; (2) higher AI depth & appropriateness score; (3) lower spread across scorers, i.e. the more consistently rated team advances. If still tied, the organising committee decides.

8.5 Notification

All teams are notified personally on 6 October — finalists with next steps and travel logistics, non-finalists with written feedback drawn from committee comments. The public shortlist is announced on 7 October. Individual scores are never published; teams may request their own aggregate scores and comments.

9. Pitch Day — Limassol, 23 October

9.1 Format

Item	Detail
Venue	TBA
Time per team	5 minutes pitch + 5 minutes jury Q&A. Hard stop at 5 minutes — the timer is visible and the microphone is cut.
Order	Drawn by lot and published the evening before.
On stage	At least 2 team members must present. All travelling members may be on stage.
Slides	Optional, 16:9, submitted by 21 October 23:59 EEST. Live demos are welcome but you are responsible for having an offline fallback.
Language	English.
Audience	All finalists, mentors, jury and invited guests. Teams watch each other pitch.

The pitch may go beyond what was submitted on 1 October — teams are free to keep building. The jury is briefed on the submitted package but judges what it sees on the day.

9.2 What a strong pitch covers

The problem, and why it is worth solving.
What you built, shown rather than described.
The impact, with numbers.
What it takes to put this into real use across Our Holding.
What you would do next.

10. Jury and prizes

10.1 The jury

The jury is composed of the CEOs of the participating companies. The organisers do not vote.

The jury decides by deliberation rather than by numeric scoring. After all 8 pitches, the jury convenes in private and agrees on the three winners. Committee shortlist scores do not carry forward — all 8 finalists start pitch day equal.

The jury's discussion is guided by the same dimensions used for shortlisting — business value and impact, depth and appropriateness of the AI, and scalability across the group — together with the quality of the pitch itself and the team's readiness to take the work forward. These are guidance, not a scoring sheet.

The jury's decision is final and is not subject to appeal.

10.2 Prizes

Place	Award
1st place	Grand prize — value and composition to be confirmed (see Section 12)
2nd place	Runner-up prize — to be confirmed
3rd place	Third prize — to be confirmed

Prizes are announced and awarded at the gala dinner on the evening of 23 October. Prizes are awarded to the team and shared equally among its members unless the team agrees otherwise in writing.

11. Data protection and responsible AI use

This section is not a formality. We operate in a regulated industry, and a single careless prompt can do more damage than a failed prototype. Read it before you start building — it constrains what you may put into a model and what you may build.

11.1 Your AI environment

Every admitted participant receives access to a shared Claude Team environment, including Claude Code and Claude Design, provisioned against their work email address.
Access is granted after the pre-selection call and the Section 11.8 confirmation, and is withdrawn after 31 October.
Use your own work account. Do not share credentials, do not let a non-participant work through your seat, and do not do race work in a personal Claude account or on personal devices outside company policy.
The Team environment is the default place to do the work. It sits under the company's commercial agreement, and content in it is not used to train models. That is a reason to work inside it, not a reason to relax anything below.

11.2 What must never be entered into any AI tool

The following categories are a hard stop. They may not be typed, pasted, uploaded, screenshotted or connected into any AI tool — including the Claude Team environment — at any point in the race.

Category	Includes, but is not limited to
Client and player records	Names, contact details, account or player IDs, KYC documents, identity documents, gameplay history, transaction and deposit history, support tickets.
Personal information	Data about any identifiable person — colleagues, candidates, players, partners. CVs, performance reviews, salary and compensation data, health or absence data, home addresses.
Financial data	Revenue, margin, P&L, forecasts, unreleased results, payment provider details, banking and settlement data, pricing agreements.
Credentials and secrets	Passwords, API keys, access tokens, connection strings, private certificates, internal URLs that expose infrastructure.
Confidential business material	Unreleased product plans, M&A material, contracts, legal advice, board or investor material, anything marked confidential or internal-only.
Regulatory and licensing material	Licence correspondence, regulator submissions, audit findings, AML and fraud case files, responsible-gambling case records.
Third-party confidential data	Anything a partner, vendor, supplier or client shared with us under an NDA or equivalent obligation.

The test to apply

If you would not paste it into a public forum, do not paste it into a prompt.

If you are unsure whether something falls into the list above, the answer is no until Security says otherwise. Asking costs you an hour; guessing wrong costs the company far more.

11.3 Build on synthetic data

Prototypes must run on synthetic data — data you generate yourself that mimics the structure of the real thing without being derived from it. Claude is good at generating realistic synthetic datasets; ask it to.
Anonymised or pseudonymised production data is not an accepted workaround. Re-identification risk is real, and that assessment is not something to make under a hackathon deadline.
Exports, screenshots, database dumps and copy-paste from production systems all count as production data, however small the sample.
If your idea genuinely cannot be demonstrated on synthetic data, raise it with Security before you build, not after. That conversation may reshape your idea — far better in August than on 1 October.

11.4 Using AI tools other than Claude

The shared Claude Team environment is the default and preferred tool for the race.
Other tools — a different model provider, a self-hosted open-source model, a specialised API — may be used, but only with prior written approval from Security. Request it through the organisers and allow 3 working days.
Approval covers a named tool for a named purpose. It relaxes nothing in 11.2: the hard-stop list applies to every tool, everywhere, always.
Any third-party tool used in a submitted prototype must be disclosed in the submission. Undisclosed use is treated as a breach, not an oversight.
Licence or usage costs for approved third-party tools are the team's own unless the organisers say otherwise.

11.5 How your solution itself must behave

What you feed the model is half of it. What you build has to be defensible too. Every submission is expected to have thought through:

Principle	What we expect to see
Human oversight	A person remains accountable for consequential decisions. Nothing affecting a customer, an employee or a payment runs fully automatically without a human check.
Transparency	Anyone interacting with the solution can tell they are dealing with AI, and can reach a human.
Fairness	If the solution ranks, scores, filters or recommends people, the team can say what could go wrong and who could be disadvantaged.
Accuracy and hallucination	The team knows where the model can be confidently wrong, and has designed for it rather than hoped against it.
Data minimisation	If this went live, what personal data would it actually need? The answer should be the minimum, not the maximum available.
Auditability	You can reconstruct why the system produced a given output.

The business case one-pager includes a two-line risks and safeguards statement covering this. The jury will ask about it on stage.

11.6 Out of scope

The following are not admitted as race entries, regardless of technical merit:

Anything designed to circumvent regulatory, AML, KYC or responsible-gambling controls.
Profiling players in order to target those showing signs of harm or vulnerability.
Covert monitoring, scoring or surveillance of employees.
Consequential decisions about people — hiring, discipline, credit, account closure — with no human in the loop.
Scraping or acquiring third-party data in breach of that party's terms.

11.7 If something goes wrong

Mistakes happen, and the useful response is speed. If sensitive data reaches an AI tool — yours or a teammate's — tell the organisers and Security the same day. Do not delete the conversation and hope.

Incidents are assessed case by case by the organisers together with Security, and the response is proportionate to what actually happened: remediation and guidance at one end, disqualification at the other. Self-reporting is treated as good practice. Concealment is what escalates.

Nothing here replaces your existing obligations under your employment agreement and the company's IT and data-protection policies. Those apply throughout the race and are enforced through the normal channels.

11.8 What every participant confirms at registration

Each team member ticks the following individually. No confirmation, no environment access.

I have read Section 11 and understand what I may not put into an AI tool.
I will build and demonstrate on synthetic data only.
I will request approval from Security before using any AI tool other than the provided Claude environment, and I will disclose any third-party tool in our submission.
I will not share my environment access with anyone outside my team.
I will report any incident to the organisers and Security the same day.

12. Intellectual property, confidentiality and data

Participants remain bound by their existing employment agreements, confidentiality obligations and company IT and data-protection policies for the duration of the race. Nothing in these rules overrides them.

Ownership of the solutions created during the race is being confirmed with Legal and HR and will be added here before registration opens on 5 August. Registration will require explicit acceptance of the final clause.

Participants agree that their name, image, team materials and pitch may be used by Our Holding for internal and external communication about the race.