How to choose an AI partner in the GCC
Most AI partners can demo. Far fewer can put a working system inside your operation and stand behind the number it moves. This is the short list of questions that separates the two, written for a leader in Dubai, Abu Dhabi, or Riyadh who has been pitched more AI than they can use and wants to know who will actually deliver.
Start with proof in production, not a demo
A demo shows what a tool can do in ideal conditions. Production shows what it does on a Tuesday, against real inquiries, when the data is messy and the edge cases arrive. The gap between the two is where most AI projects die. MIT's State of AI in Business 2025 study found that about 95 percent of enterprise generative AI pilots never reach measurable impact on the P&L, and the cause was not the models. It was the failure to embed them in how the business actually runs.
So the first question is not what can you show me, it is what have you put into production and what number did it move. Ask to see a system that is live, carrying real volume, owned by the client. A partner who can only show a sandbox is selling you the pilot that the data says has a one-in-twenty chance of mattering.
Ask who owns the code, the data, and the prompts
Ownership decides whether you are buying an asset or renting a dependency. When the engagement ends, do you hold the code, the data, and the prompts, with a documented plan for who runs the system next, or does it all leave with the vendor. The honest answer should be that you own what was built. If the structure makes you a permanent tenant of someone else's platform, price that in, because the switching cost is the real cost.
Tie this to service levels in writing. Who responds when an agent misfires at 9pm, how fast, and what is the path to a fix. A serious partner will commit to this plainly. A vendor who deflects on ownership and support is telling you what the relationship will feel like the first time something breaks.
Demand strategy before software
The right partner starts from your number, not their product. Before anyone talks architecture, they should be able to name the single metric the work is meant to move: response time to a first inquiry, the share of inquiries that convert, the cost of an operational backlog. If the conversation opens with a platform and a licence tier, the diagnosis is being skipped, and the diagnosis is the part that decides whether the build works.
Strategy first also means scope discipline. A partner worth hiring will narrow the first build to one problem worth solving, ship it, and prove it, rather than sell a sprawling transformation that takes a year to show anything. Intelligence applied to the wrong process is expensive theatre. The diagnosis is what keeps it off the wrong process.
Check who actually does the work
In many firms the people in the room for the pitch are not the people who build the system. Senior names win the engagement, juniors deliver it, and quality drifts. Ask directly who will be on your build, how senior, and whether the principals stay involved past the kickoff. In a boutique the answer is usually the founders and a small senior team, which is a feature, not a limitation. Fewer hands, more accountability, no layer of account management between you and the people writing the code.
This matters more in the GCC, where the market is young and a lot of supply is resellers fronting someone else's platform. Press on local delivery: can they sit with your team in Dubai or Riyadh, do they understand the regulatory and language reality, and is there a named person who owns the outcome.
Hire the partner willing to tell you no
The strongest signal is a partner who will tell you where AI does not help. If every question gets a yes and every process is a candidate, you are hearing a sales pitch, not an assessment. A clear no, given early, is worth more than a hopeful pilot that burns a quarter and a budget before it admits the same thing.
Test it in the first meeting. Describe a process you assume is ripe for automation and watch whether they qualify it or simply agree. The partner who pushes back, scopes smaller, and ties the work to a number you already watch is the one most likely to be in the surviving five percent.
Common questions
- What should I look for in an AI partner in the GCC?
- Proof in production over demos, clear ownership of the code, data, and prompts, a strategy-first approach that starts from a number you already watch, senior people doing the actual build, and a partner willing to tell you where AI does not help. The market has many resellers fronting other platforms, so press on local delivery and who owns the outcome.
- What questions should I ask an AI vendor before hiring them?
- What have you put into production and what number did it move. Who owns the code and data when we are done. Which metric will this build move. Who, specifically, will do the work and how senior are they. And where would you tell us not to use AI. The answers separate a build partner from a slideware vendor.
- Is a boutique AI firm better than a large consultancy?
- For most operational AI work, a boutique gives you principals on the build instead of a senior name who sells and juniors who deliver. Fewer hands, more accountability, and a direct line to the people writing the code. The trade is scale: for a single high-value problem put into production, that trade usually favours the boutique.
- Why do so many AI projects fail to deliver?
- MIT's State of AI in Business 2025 study found about 95 percent of enterprise generative AI pilots never reach measurable P&L impact, and traced it to an integration gap rather than the models. The projects that work are bought from specialist partners and embedded in the daily workflow, not run as a demo on the side.
- Should we own the AI system or license it from the partner?
- Ownership turns the build into an asset rather than a permanent dependency. Ask whether you hold the code, data, and prompts at the end, with a documented plan for who runs it next. If the structure makes you a tenant of the vendor's platform, price in the switching cost, because that is the real cost of the relationship.
- How do we know if an AI partner can actually deliver locally in the GCC?
- Ask to see a live system carrying real volume for a client in the region, confirm they can sit with your team in Dubai, Abu Dhabi, or Riyadh, and check they understand the regulatory and language reality. A named person who owns the outcome, rather than an account manager between you and the builders, is the signal that delivery is real.