An AI consultant in the UAE who sells an answer, not a project
Most people searching for an AI consultant do not need one. They need a written answer, quickly, to three questions: is the thing they have in mind worth building, does a product already do it, and what would it actually cost. This service is that answer: a short paid assessment of one workflow, ending in a written recommendation the client owns and can take to any vendor. The first call is free and takes thirty minutes. Scoping a build you have already decided on is also free and ends in a fixed price. The assessment is for the question a scoping call cannot answer: whether to build at all.
The problem an assessment is built to catch
The usual shape of AI consulting is a discovery phase that discovers a need for the consultant. That is an incentive rather than a conspiracy: an engagement that recommends nothing has nothing to invoice next. The output is a strategy deck, a roadmap with three horizons on it, and a pilot that never reaches anybody's real work.
The industry's own analysts forecast the failure rate. In July 2024 Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, citing poor data quality, inadequate risk controls, escalating costs and unclear business value. Rita Sallam, Distinguished VP Analyst, was quoted as saying that "after last year's hype, executives are impatient to see returns on GenAI investments, yet organizations are struggling to prove and realize value." The same release put business-model-transforming deployments at between USD 5 million and USD 20 million, which is the top of the range rather than the floor, and a different buyer to the one reading this page.
The abandonment figure is the relevant one, because it is a number about decisions rather than about technology. Something got approved that should not have been, or got built before anyone could say what correct looked like. An assessment exists to catch that before the money moves.
What the assessment covers
One workflow, examined properly, written up.
What to send
The workflow you want to change, described the way you would describe it to a new joiner. The systems it runs through, named: HubSpot, Salesforce, Zoho CRM, Xero, QuickBooks, Stripe, Slack, monday.com, Asana, whatever it actually is. Roughly how often it runs and roughly how long it takes today. Production credentials are not required, and are not requested during an assessment.
What happens to it
The first test is whether the task has a definition of correct that somebody can write down, because a task nobody can grade is a task nothing can be tested against. Then the systems: whether an API exists, what it is rate limited to, and whether the data the workflow depends on sits in a system or in somebody's head. Then whether an off-the-shelf product already covers it. Last, a count of the exceptions, which is what turns a two week build into an eight week one.
What comes back
A written recommendation, in prose, that a non-technical director can read and a technical one can argue with. It names the recommended option and the reasons, the constraints found and where they were found, the failure modes worth planning for, and an indicative cost band with the arithmetic shown rather than asserted. The document belongs to the client: it carries no non-compete and no exclusivity clause, so it can be taken to another agency, an internal team, or a product vendor for a response.
The three things the recommendation can say
Two of the three end the engagement. That is what the fee buys.
"Do not build this"
The common reasons. The task is rare, so the build costs more than the years of manual effort it saves. Nobody can state what a correct output looks like, so there is nothing to test against and no way to know when it breaks. No one will own the exceptions, so the queue of unresolved cases sits unread until the workflow quietly reverts. Or the decision inside the workflow is one a person has to make, in which case automating around it moves the work rather than removing it.
"Buy the product that already does this"
A large share of requested builds are a feature of something the client could subscribe to on Monday. Where the workflow is genuinely standard, an off-the-shelf tool wins on price, on time to value, and on who fixes it at 2am; the recommendation names the product and says why. The case for building is never that the alternative is weak. Zapier and Make run code, loops, branching and agents perfectly well. It is ownership, testability and failure handling, and it only holds when those are worth paying for.
"Build it, and here is the shape"
Where the answer is build, the recommendation states what to build, which band it falls into, and what has to be true before it starts. That is the point at which the conversation moves to AI agent development, a conversation the document is designed to support with any supplier, not just this one.
What it costs
The assessment is fixed price, quoted in AED before it starts, and is a small fraction of a build. Where it leads to a build with Telora Labs, the assessment fee comes off the build price.
The build it might lead to is the number most people are looking for: AED 10,000 to 50,000, fixed in the proposal before work starts.
| Band | Scope | Price (AED) | Indicative timeline |
|---|---|---|---|
| Single-workflow agent | One trigger, one or two systems, human escalation path | 10,000 to 20,000 | 2 to 3 weeks |
| Multi-system agent | Three or more systems, branching logic, error handling | 20,000 to 35,000 | 3 to 5 weeks |
| Complex agent | Custom logic, compliance constraints, self-hosted model, or heavy integration surface | 35,000 to 50,000 | 5 to 8 weeks |
Model and hosting costs are billed by those vendors and sit outside the fee; the arithmetic is broken down in what an AI agent costs in Dubai. Start there if the question was really a budget question.
Send the workflow, get an answer on whether it is worth building
The workflow, the systems it runs through, and roughly how often it happens. The assessment fee is quoted before anything starts, and "do not build it" is a valid answer to have paid for.
hello@telora.dev