The right AI consulting firm in Saudi Arabia is the one that would be paid exactly the same if it told you to buy nothing. Everything else in this guide follows from that one test. Below: the five questions, what to get in writing, a worked example, the traps, and how to run it all in one afternoon.
AI is projected to contribute about $135bn to Saudi Arabia’s economy by 2030, the largest gain in the Middle East (PwC, via trade.gov). Money like that attracts everyone. Overnight, every systems integrator, marketing agency, and software reseller in the Kingdom became an “AI consulting firm.” Some of them are excellent. Many are something else wearing the label, and the label is free.
The five gates, in order
Picture the questions as five gates in a row. A firm has to pass through each one to earn the next, and in our experience most do not make it past the second.
flowchart TD
A(["You are evaluating an AI consulting firm"]) --> B{"Do they earn anything if you buy what they recommend?"}
B -->|Yes| X(["That is a sales channel, not an advisor"])
B -->|No| C{"Will the people advising you also build it?"}
C -->|No| Y(["Expect slides, then a hand-off"])
C -->|Yes| D{"Can they show Arabic and PDPL fluency?"}
D -->|Yes| E{"Is the first proposal small, priced, and measurable?"}
E -->|Yes| F(["Shortlist them and ask for their 'no' story"])
Gate one: do they earn anything if you buy what they recommend?
Ask it in exactly those words, and ask for the answer in writing. Many firms in the region earn reseller margins, referral fees, or partner incentives on the platforms they recommend. That does not make them dishonest. It makes their advice structurally biased toward whatever pays them, at exactly the moment you need a clear head. If the answer is yes, you do not have an advisor. You have a salesperson with a clipboard. We have written about why vendor-neutral evaluation pays in detail.
Gate two: will the people advising you also build it?
Strategy decks do not run inference. A firm that advises but does not build has no skin in whether the advice survives contact with your data, your systems, and your team. A firm that builds but does not advise will build whatever you ask for, including the wrong thing. You want the same engineers in both seats.
Gate three: can they show Arabic and PDPL fluency?
Saudi specifics are not a nice-to-have. If your customers or staff work in Arabic, ask the firm what goes wrong with Arabic in large language models and listen for a real answer about dialects, script, and how they evaluate output in Arabic rather than trusting English benchmarks. Then ask how they would keep your project inside the PDPL, which took effect on 14 September 2023 with full compliance required by 14 September 2024 under SDAIA as regulator (Morgan Lewis, 2024). A firm that shrugs at either question will learn on your budget.
Gate four: do they push you to start small?
Gartner predicts at least 30% of generative AI projects will be abandoned after proof of concept by the end of 2025, due to poor data quality, inadequate risk controls, escalating costs, or unclear business value (Gartner, 2024). The honest response to those odds is a small, contained first project with a clear price and a clear payoff, proven before you spend more. So the tell is the size of the first proposal. Anyone pushing a large, all-at-once AI transformation before a single use case has paid for itself is selling, not advising. A good firm will happily tell you what your first project should cost, in numbers.
Gate five: will they tell you not to buy?
McKinsey found that only around 39% of organizations report an enterprise-level EBIT impact from generative AI, and only about 5 to 6% are genuine “AI high performers” (McKinsey, 2024). Which means that for some use cases, the right answer is “not this one, not yet.” Ask the firm for an example of a client it talked out of an AI project. If it has never done that, every road with it leads to an invoice.
What to get in writing before you sign
Answers in a meeting are cheap. These are the documents we would ask for, and what each one tells you.
- A conflict-of-interest statement. One page listing every vendor, platform, or reseller relationship the firm earns from. Gate one, on paper. A firm with nothing to hide will send it the same day.
- A named team with shipped systems. Names, roles, and the last two production systems each person delivered, with a client you can call. The engineers who will do the work, not the partner who pitched.
- A fixed-price, fixed-scope first phase. Four to eight weeks, one use case, and a written definition of what “worked” means before it starts.
- A data-handling annex. Where your data is processed, which models or APIs it passes through, what leaves the Kingdom, and how that squares with the PDPL. If your sector answers to SAMA or the NCA, the annex should say so by name.
- Ownership and exit terms. You own the prompts, the fine-tuned weights, the evaluation sets, and the code, and you can take them elsewhere without a fee. A firm that resists this is planning to be your lock-in.
- The “no” story. A short written account of a client the firm advised not to proceed, and why. The cheapest reference check you will ever run.
A worked example: two proposals for the same problem
Say you run a 60-person distribution company in Riyadh and want an assistant to handle first-line customer questions in Arabic and English. Two firms respond. The numbers below are made up to show the shape, not to price anything.
Firm A proposes an enterprise AI platform: SAR 1.4 million over nine months, covering customer service, sales, and finance, plus licenses of SAR 20,000 a month. It is a certified partner of the platform vendor. The deck is beautiful, the team names are “to be confirmed,” and the first measurable result arrives in month seven.
Firm B proposes a six-week pilot on one support queue: SAR 95,000 fixed, its own two engineers, a plain data annex, and one number to judge it on, the share of tickets resolved without a human in each language. If that share clears the agreed bar, the second phase is priced from what the pilot measured. If it does not, you have spent SAR 95,000 to learn something Firm A would have charged you SAR 1.4 million to discover in month seven.
Firm B has passed gates one, two, and four before you have asked a single question, and Firm A has failed all three. That is what the gates are for: a gut feeling about a polished pitch becomes something you can defend to your board.
Where the choice goes wrong
- The badge trap. Partner logos, award walls, and “top AI firm” listicles measure marketing budgets, not engineering. The credentials that predict your outcome are named people with shipped systems, references you can call, and a written answer to gate one.
- The free assessment. A free “AI readiness assessment” is usually a sales funnel with a clipboard. Pay a modest fee for an independent one and you get an answer instead of a proposal.
- The cheapest bid. The lowest first-phase price often hides the reseller margin or the license fees that follow. Compare the twelve-month cost, not the first invoice.
- The tender that scores the wrong things. If you are a government entity buying through Etimad, the scoring sheet tends to reward page count and certifications. Write gates one, two, and four into the evaluation criteria, or the process will select the best proposal writer instead of the best engineer.
How to start: run the gates in one afternoon
The reliable pattern is boring. Pick one problem that costs you real money today, and write it down in a paragraph. Send the five questions, in writing, to every firm you are considering, with the document list above. Shortlist the two or three that answer all five without flinching. Ask each for a fixed-price first phase on your one problem, with a single number that will decide whether there is a second phase. Run the pilot, measure that number honestly, and only then talk about scale.
A firm that passes all five gates is rare, which is rather the point. Ask what it earns if you buy. Ask who builds. Test the Saudi specifics. Check the size of the first proposal. And listen for the word “no.”
Want to see how we answer our own five questions? SDCG’s AI consulting is independent by design: no resale, no commissions, and the engineers who advise you are the ones who build. Book a free 30-minute review and bring us your hardest question first.