Prime Studio · AI-native transformation for gaming operators

Most operators have added AI. Almost none have become AI-native.

Adding AI is a few tools and a pilot that demos well and never reaches production. Becoming AI-native changes the operating model: how work is specified, how decisions are governed, how results are measured, and who answers when a system acts on its own.

Twenty years inside one gaming operator. Two companies built in parallel — both acquired by the group.
Operator-side since 2000: payments, fraud, data, automation.
Everyone here ran these functions inside an operator before advising anyone about them.

The distinction

The difference is not how much AI you bought.

It is whether the company changed shape around it. One of these columns describes a procurement decision. The other describes an operating model. Only the second one survives an audit.

Adding AI

  • Tools bought per team, licences counted as adoption
  • Pilots that demo well and stall before production
  • Output judged by whether it looks right
  • The knowledge lives in whoever ran the pilot
  • No control model, so nothing risky ever ships
  • Value described in anecdotes
  • Locked to whichever vendor sold the pilot

Becoming AI-native

  • Work specified before it is built, and the spec is the artefact
  • Named human authority at every consequential decision
  • Output judged against evaluations, with evidence retained
  • Knowledge versioned and owned by the company
  • Controls designed first, so production is reachable
  • Value tied to a metric someone already reports monthly
  • Provider-agnostic — the knowledge, rules and governance stay yours

Where we start

Four domains. One chain.

Money arrives, it is protected, it is verified, and it is balanced. We start at the money chain, and the reason is not that AI is interesting there.

The loss is countable

Approval rate, chargebacks, hours to close, cost per contact. Nobody has to be convinced the problem exists — it is already in a report someone reads every month.

The budget already exists

Regulatory pressure on KYC and fraud means a line is already assigned. We are not asking anyone to create a budget, only to spend an approved one better.

It is where we are deepest

Payments, fraud, banking, KYC and service automation are covered twice over by people who owned them inside an operator. We do not sell the domains we merely visited.

The instrument

Four domains, six dimensions, twenty-four scored cells.

The same six questions asked of every domain, so the results are comparable against each other — and against the next operator we measure. Every score carries the evidence behind it. No evidence, no score.

Scale L0 — L4 24 cells Three weeks 12–16 interviews

The scale

Five levels, from people doing it by hand to agents working inside controls.

L0
Manual

People execute. Nothing is instrumented. The knowledge lives in someone's head.

L1
Instrumented

The process is measured, but still runs at human speed.

L2
Automated

Rules and scripts execute the defined paths. People handle the exceptions.

L3
Assisted

Models recommend. People decide, and remain accountable for the decision.

L4
Governed autonomy

Agents execute inside controls, with human authority checkpoints, audit trails and evaluations.

The part most diagnostics leave out

L4 is not the goal everywhere.

Part of the value of the diagnostic is naming where L2 is sufficient, and where reaching for autonomy is regulatory risk dressed up as innovation. A diagnostic that recommends the maximum in every cell is not a diagnostic. It is a sales deck.

The column that decides the rest

Every one of these programmes fails in the same place.

Six dimensions get scored. Five of them are technical. The sixth decides whether the other five survive contact with the organisation.

A transformation rarely fails because the model was wrong. It fails because the person who owned the process was never told what happens to their job, and quietly made sure the new thing did not work. That is not resistance to technology. It is a rational answer to a question nobody answered.

So we score it like everything else: who owns this, what happens the day they resign, how much of the logic exists only in their head, and who is being asked to hand over the part of their work that made them valuable.

We have been on both sides of that conversation — what the system can do, and the fear of being replaced by it. A plan that only answers the first one does not get executed, however good it is.

What you receive

Four deliverables, and the last one is a decision.

01

Maturity scorecard

All twenty-four cells scored, each one with the evidence that supports it — the process observed, the data pulled, the person who confirmed it.

02

Opportunity map

Every opportunity carries the metric it moves, the estimated value, the effort, the control model it requires and the proposed owner. Ranked by value over effort, not by how interesting it is.

03

Three-move recommendation

What to do in ninety days, in six months, and in eighteen. Explicit about what not to do yet, and why waiting is the right call.

04

Executive readout

Ninety minutes with the committee. Not a PDF sent by email. The last slide is the proposal for the next engagement, with scope, price and a start date — so the decision happens in the room.

Week one

Domain interviews and data extraction. Twelve to sixteen conversations with the people who actually run the process.

Week two

Analysis and scoring, with every finding validated against the process owner before it is written down.

Week three

Synthesis, quantification of the opportunity map, and the executive readout.

Who does the work

Operators first, advisors second.

Every person on this team ran the function they assess — through acquisitions, regulatory pressure, live money and legacy systems that could not be switched off. That is a different qualification from having read about it.

Payments · Fraud

Processing, integrations, routing and fraud prevention end to end, fiat and crypto, inside a live operator.

Banking · KYC · Automation

Banking rails, identity verification, and the automation of customer service and fraud decisioning. Data and applied AI.

Operations · People

Process design, organisational structure, team building and talent development — who implements it after we leave.

Architecture · Reconciliation

Systems architecture, Spec-Driven Development, agentic implementation, and the synthesis that turns findings into a plan.

Gustavo Granados
Founder

Twenty years inside one gaming operator, ending on its Operations Committee. Two companies built in parallel — the group acquired both. Operator-side since 2000.

Verify on LinkedIn →

What comes after

The assessment is the entry point, not the product.

Each step is contracted on the findings of the one before it. Nothing starts over, and nothing is sold before it has been measured.

Money & Risk AssessmentStart here · three weeks

Payments, fraud, KYC and reconciliation scored against the maturity model, with a prioritised opportunity map and the business case behind each item.

Transformation RoadmapNext

A sequenced twelve to twenty-four month plan tying every initiative to an operational metric, a control model and a named owner.

Agentic ImplementationBuild

Production agent architectures on Spec-Driven Development, with human authority checkpoints, audit trails and evaluation harnesses.

Engineering EnablementCapability

Multi-cycle programmes for senior engineering teams, so the capability stays in the building after the engagement ends.

Fractional AdvisoryOngoing

Continuing counsel to leadership and operating committees, at the cadence the transformation actually requires.

The brief

Read the scope before you talk to us.

The brief is the actual scope document for the Money & Risk Assessment — not a teaser for it. If the method does not convince you on paper, a call will not fix that.

  • The full four-by-six matrix, with all twenty-four questions
  • The five-level maturity model and how each level is evidenced
  • The three-week schedule and what we need from your side
  • How the opportunity map is quantified and ranked
  • What we will tell you not to do, and why

One email with the brief attached. No sequence, no newsletter, no sharing your address. We do not take product pitches, staffing proposals or agency outreach at this address.