The Five
G42: Government backed and government scale. If you are a government entity or need regional compute at a national level, they are the answer. If you are a private enterprise, read the next section before proceeding.
Presight: Builds computer vision and smart city infrastructure. Surveillance, traffic, urban analytics. If your use case involves a camera, they are relevant.
Core42: The infrastructure layer. Cloud compute, data centers, sovereign cloud. They are who other AI companies run on. They do not build agents. You would still need a development partner on top.
AI71: Produces open source foundation models, primarily the Falcon series. Strong models. You still need an engineering team to build anything useful on top of them.
Rifoa: Deploys custom AI agents for enterprise workflows. Finance, operations, sales. Abu Dhabi based. No government backing.
The Question Nobody Asks
When an AI company automates your finance workflows, they see your books. Your reconciliation data. Your vendor payment terms. Your cash position.
When they automate your sales pipeline, they see your client list, deal sizes, and conversion rates.
When they automate your HR processes, they see your headcount costs, salary bands, and performance data.
The question is not whether the technology works. The question is: what are their incentives with that data?
A government backed company is building national AI infrastructure. Your operational data, aggregated across hundreds of enterprises, is valuable training material for that infrastructure. You are not being paranoid to wonder about this. You are being a competent CFO.
A private founder led firm has one incentive: deliver the project, get paid, earn referrals. There is no model to train. There is no government reporting requirement. There is no national initiative that benefits from knowing what your margins look like.
Aligned incentives are not a nice to have. They are the foundation of a working relationship.
What Rifoa Does Differently
Model-agnostic: We use whatever works for the task. Claude for reasoning. Codex for code generation. Haiku for high volume simple tasks. The model that performs best, not the one that sounds best on a pitch deck.
On site: We come to your office. We watch how work actually moves. The workarounds your team has quietly built up over years. Documentation misses this. Being there does not.
Direct access: When something breaks at 11pm, you reach the person who built it. Not a support ticket. Not tier one escalation. The person who knows exactly why the agent behaves the way it does.
How a Project Runs
Weeks one and two: We audit your operations. Every workflow mapped. What to remove, what to automate, what to leave.
Weeks three and four: One workflow in production. Proof before expansion.
Month two to three: Iterate on real usage, expand scope, tighten reliability.
Month three onwards: Handover with documented blueprints and team training. Or we stay on retainer if you want us to.
Working agents in production within thirty to sixty days. No strategy decks.
The Honest Summary
- •Choose G42 if you are a government entity or need national scale compute.
- •Choose Presight if your use case involves cameras or smart city infrastructure.
- •Choose Core42 if you are building your own AI stack and need UAE sovereign cloud infrastructure.
- •Choose AI71 if you have ML engineers and want to build on open source models.
- •Choose Rifoa if you want custom agents built for your specific workflows by a team whose only incentive is to make them work.