Where Should You Start With AI? First, See Where You Stand Today

When we bring up AI with companies, we hear three sentences again and again:

• "We don't know where to start."
• "Our data must not leave the company."
• "Departments are already using tools on their own, and we have no control."

They sound different, but they point to the same thing: before anyone talks about what to do, nobody knows where the company stands today.

---

Assess First, Recommend Later

AI conversations usually start with the solution: which tool, which model, which project? But the same solution plays out very differently in two companies. In one, customer contracts don't allow data to leave the company. In the other, employees already use chat tools every day on personal accounts, and nobody knows what goes into them.

You can't give those two companies the same recommendation. A recommendation can only be right once you know the current position.

Below we describe that position in six dimensions. Each one ends with a question to ask yourself. If you don't know the answer, that's a finding too. We come back to this further down.

---

Your Current Position in Six Dimensions

1. Usage

AI is most likely already in use at your company. The real question is how: through approved tools the company provides, or through employees' personal accounts?

The second question is the level of use. Is it limited to individual tasks like writing, translation and summaries? Has it become part of a department's process? Or is it running in an application connected to the company's own data?

"We don't use it at all" is no better a place to be than uncontrolled use. In both cases, the company has no experience it can build on.

Ask yourself: Which AI tools are your employees using today, and on whose account?

2. Governance

Two things matter here. First, is there a written rule or guideline on AI use, and do employees know about it? Second, can you see what information goes into these tools? That takes company accounts and logs.

If the only rules are verbal, you can't see what leaves the company. There seems to be a rule, but nobody knows whether it's followed.

Ask yourself: If an employee pasted a customer quote into a chat tool last week, would you know?

3. Clarity of Constraints

Behind "our data must not leave the company" there is usually one of three sources:

• Law. Data protection law, such as GDPR in the EU or KVKK in Turkey. These cover personal data; a price list or a technical drawing is outside their scope.
• Customer contracts. Confidentiality commitments, and the audit and certification requirements your customers place on you.
• Internal policy. A preference with no written basis, of the "that's how we do it here" kind.

Which one applies directly determines what's possible. Often the problem isn't the constraint itself but not knowing where it comes from. Have the confidentiality clauses in your customer contracts ever been reviewed with AI use in mind?

Ask yourself: Where is the rule that stops data from leaving the company actually written down?

4. Data Classification

Having sensitive data is neither good nor bad. Every manufacturer has prices, costs, contracts and technical drawings. What matters is whether the company has defined which data is sensitive.

A practical way to check is this question: how does an employee know whether a piece of information can go into an external tool? From a written rule or a label on the data? By asking their manager? Or at their own discretion?

Ask yourself: How does a new hire find out which files are confidential?

5. Access to Company Data

AI usually adds value to a company by reaching that company's own data. So it matters where the data sits: ERP, production and quality systems, CRM, file servers, SharePoint or Google Drive, email, Excel files... In most companies, data is spread across several of these.

Two more questions follow. Are these sources connected, or are they merged by hand in Excel? When a manager asks a question that needs data, do they see the answer on screen themselves, or does someone have to spend days preparing a report?

Even if you don't have an ERP, this dimension gives you a meaningful result. It's about the company's data, not a particular piece of software.

Ask yourself: How many different places does your company's business data sit in?

6. Ownership

The last dimension is the simplest: does the topic have an owner? Is there an appointed person with a defined role and authority? Is someone looking after it informally? Or is it on the agenda with nobody responsible for it?

Budget belongs to the same question: is there a budget or management approval for AI work, or does everything rely on individual initiative?

Ask yourself: If an AI project were approved tomorrow, who would own it?

---

"Don't Know" Is an Answer Too

You'll answer "don't know" to some of these questions. That isn't a failure; it's a finding. It shows which areas are outside the company's field of view.

"Don't know" and "no" are not the same. "We have no written rule" is one situation. "I don't know whether we have a written rule" is another. Keeping them apart is what makes the result readable.

---

Run This Assessment in 4 Minutes

Instead of working through the six dimensions one by one, you can use our AI Readiness Assessment (link at the end of this post). It has 16 questions across six dimensions and takes about 4 minutes.

We made some deliberate choices when designing it:

• The result is on screen right away. You don't need to give your email address to see it. Receiving the report by email is optional.
• Every score shows its source. Open any dimension to see which answer you gave to which question, and how many points it earned.
• No AI in the scoring. That may sound odd for an AI assessment. But an assessment has to be repeatable and explainable. Because it uses fixed rules, the same answers always give the same result.
• A low score is not a failure. The result is a score from 0 to 100 and one of four levels: Starting, Aware, Structured, Managed. The levels are stages, not grades.
• It makes no recommendations. The tool only shows where you are today. Where to start, and under which conditions your data can be used, is the subject of a separate piece of work.
• No industry average. We don't show a figure comparing you with others. There isn't enough real data for that yet, and we don't make it up.

The assessment is available in English and Turkish. If you'd like to go through your result together, you can book a 30-minute call from the result screen.