Ting Comparisons
Build vs buy AI: which is right for a Saudi company?
The decision, in short
Buy when a product already does your task on your data, in your users' Arabic, inside the systems you run; build when the task is specific to how you operate or the data has to stay under your control. Either way, write the requirement first so both paths are judged against the same cases.
Reviewed by Badr Al-RashedUpdated
Faster to first use
More control and fit
- Buy a productConfigure it; the vendor owns the roadmap
- Buy and integrateA product plus custom connectors to your systems
- Build a custom systemSpecified, built and owned on your terms
Write the requirement first, then judge every option against the same cases.
The comparison, criterion by criterion
Every row has a check you can run before you decide. None of them needs a vendor's word for it.
| Criterion | Buy a product | Build a custom system |
|---|---|---|
| Time to valueHow to checkRun the product on a sample of your own cases before you sign; ask a builder for a dated plan to the first live user. | EdgeFast if the product fits your process as it is; slow if the process must change to fit the product. | Slower to start: specification, build, integration and testing come before the first live user. |
| Data and control[1]How to checkList where every component processes and stores data, the model provider included, and who can access it. The Personal Data Protection Law applies to any processing of personal data in the Kingdom (Article 2). | The vendor's hosting, sub-processors and model provider. You control what the contract gives you. | EdgeYou choose the hosting and access model, and you carry the work of securing it. |
| Integration with ERP, CRM and WhatsAppHow to checkAsk for a working connection to your actual system, not a logo on a slide. | Limited to the connectors the product ships. Anything else is custom work anyway. | EdgeBuilt to your systems' APIs. Each integration is scoped, tested and maintained. |
| Arabic qualityHow to checkScore your own Arabic material, dialect and scans included, separately from English. | Set by the vendor's model and tuning. You can test it but rarely change it. | EdgeChosen and tested on your material. You can switch models when a better one appears. |
| Ongoing ownershipHow to checkName who fixes a wrong answer, and how that fix reaches the live system. | EdgeThe vendor owns the roadmap and carries maintenance; you own configuration and content. | You own the system, its tests and its maintenance, or contract someone to. |
| Cost driversHow to checkModel the cost at your real volume over the life of the system, not the first invoice. | EdgeNo build budget, and the fee is known before you sign. Per-user or per-conversation fees grow with use, plus switching costs if you leave. | Specification and build up front, then hosting, model usage, monitoring and maintenance. |
| ExitHow to checkAsk what you take with you: data, configuration, prompts, test sets, code. | Often a data export only. Prompts and configuration may stay behind. | Everything, if the contract says so. |
| Best when | A product already does the task on your data and the contract covers exit | The task is specific to how you operate, or the data must stay under your control |
An edge belongs to that row alone. It does not choose for you; the Best when row does.
Choose by situation
Buy when
- A product already does the task end to end, on your data, in your users' Arabic.
- The process is standard and you are willing to adapt to the product.
- It connects to your systems through connectors that exist today.
Build when
- The task depends on how your operation works: your documents, your approvals, your exceptions.
- The data has to stay on infrastructure you control, or be processed in a specific place.
- It must write into your ERP, CRM or WhatsApp flows in ways no product supports.
Most real systems are hybrids: you buy the model and the hosting, and build the workflow, integrations and tests around them. NIST's Generative AI Profile flags poorly traced third-party components as a risk of its own, so a hybrid still needs every component and supplier listed. Sometimes the answer is neither: a rule in a system you already own does the job.
General information, not legal advice. Each regulatory point summarises the official source cited beside it, as checked on the date shown under Sources; confirm how it applies to your company with your legal adviser.
Questions to ask any vendor
- Can you run it on a sample of our own cases before we sign, and give us the outputs?
- Where is each component hosted, which model providers are involved, and where is our data processed?
- What happens to our data, configuration and prompts if we leave?
- How do you measure quality after launch, and what do we get to see?
- Which parts of our process would we have to change to fit your product?
- At our volume, what does the cost depend on, and what makes it grow?
Where Ting fits, and when you don't need Ting
Where Ting fits
- If you lean towards building, the AI Solution Blueprint specifies the system so that Ting, your team or another vendor can build it, and ends in a defined implementation quote.
- If you are buying, the same written requirement and acceptance tests let you score competing products against identical cases.
- See the AI Solution Blueprint
When you don't need Ting
- A product clearly does the job on your data, your team can configure it, and the contract covers exit. Buy it.
- The task is a fixed rule in a system you already own. Configure the rule; it isn't an AI project.
Sources
Primary sources, checked on the date shown.
