Preparing a grant-funded AI project in Türkiye: a readiness checklist

Before applying for public support for an AI (artificial intelligence) project in Türkiye, make sure the project itself is ready: the decision it will change, the data, the success metric, the team and the documentation. Data Department, an applied machine learning consultancy in Istanbul, compiled this checklist for companies considering programmes from TÜBİTAK (the Scientific and Technological Research Council of Türkiye) and KOSGEB (the Small and Medium Enterprises Development Organisation).

Which public programmes are relevant to AI projects?

The table includes only facts we could verify on official or reputable sources. Funding amounts, rates and dates change from call to call, so they are left out; always check the current call text and implementation rules before applying.

The TÜBİTAK 1711 call has three roles. The customer organisation is a company based in Türkiye, of any sector or size, that needs an AI-based solution and commits to applying the results in Türkiye. On the technology provider side there must be at least one SME and at least one university research laboratory or public research centre. Applications without a consortium are not evaluated. The 2026 call closed on 2 October 2026, after an extended application period (TÜBİTAK announcement, in Turkish).

Which decision should be clear before you apply?

A funding application should describe the business problem, not the technology. The first step is to write in one sentence the decision that the AI output will change: which order to produce and when, how much of each part to stock, or which batch carries a high quality risk.

Once the decision is clear, the data, the method and the measure of success follow. If no decision can be named, the project may qualify for support and still create little value for the company. To compare candidates, use the scoring table in how to choose your first AI project.

Is the data actually usable?

The project plan should rest on a checked data inventory, not on the assumption that data exists. Before applying, write down which systems hold the data, how far back it goes, whether it is recorded at the level of the decision (item, machine, customer) and who will grant access. ERP (enterprise resource planning) and MES (manufacturing execution system) data are often enough, but field definitions and missing records directly affect the timeline.

If the project uses personal data, settle the obligations under KVKK (Türkiye's Personal Data Protection Law) and where the data will be processed before work starts.

Which metric will show success?

A measurable target makes the application easier to assess and makes the company's own continue-or-stop decision easier at the end. Model accuracy alone is not enough. Choose a KPI (key performance indicator) the finance or operations team already tracks, such as inventory holding cost, scrap rate, lead time or waiting time.

Measuring today's baseline is part of this step. If you do not record how the decision is made now and what it produces, you cannot show improvement at the end.

How should the team and responsibilities be split?

Publicly supported projects often involve several organisations working to one plan. In the TÜBİTAK 1711 call, consortium members sign a cooperation agreement before applying, and that agreement sets the ownership of intellectual property from the project (TÜBİTAK). Roles should therefore be agreed before the application is written.

On the company side, at least three people should be named: the manager who owns the decision, the IT (information technology) contact who provides data access, and the person who will run the system after the project.

Readiness checklist

Before drafting the application, each item below should have a written answer:

• [ ] Decision: the decision that the AI output will change is defined in one sentence.
• [ ] Decision owner: the manager who will act on the output is named and has approved the project.
• [ ] Data inventory: required fields, source systems, history length and data quality are checked.
• [ ] Data access and KVKK: access permissions and personal-data obligations are clear.
• [ ] Success metric: an existing KPI is chosen and its current value is measured.
• [ ] Scope: the project can produce a measurable result within the period the programme allows.
• [ ] Roles: each organisation's and person's responsibility is written down.
• [ ] Work plan: activities, periods and the activity each cost belongs to are defined.
• [ ] Intellectual property: ownership of the project outputs is agreed.
• [ ] Going live: where the model will run, how it connects to ERP or MES and who monitors it are planned.

What happens after the support period ends?

Support covers a fixed period; the value of a model appears after it. The TÜBİTAK 1711 call expects the customer organisation to apply the results in Türkiye. Plan going live, monitoring and team training as separate activities in the project plan.

We explain why models stop being used after the first version, and how to prevent it, in why AI projects stall after the first model. The free 30-minute AI opportunity review is one way to test which decision suits a funding application.

This article is a general preparation guide, not legal or financial advice. For eligibility, support rates and deadlines, the current call text of the relevant institution applies.

Frequently asked questions