To estimate the price of an IT project, and AI project in particular, is a really tough business. Why? You will find the answer in this(link) article. However, before starting the project you must know - the order of magnitude of the cost and the risks associated with the project should be evaluated. In this article, we will try to throw light on these issues.

If you don't have the experience, or you don't have a partner or employee on board who is familiar with the topic, you should most likely get answers by talking to a technology consultant like Numlabs. However, the following questions are worth preparing for:

1. Define the business problem well.

It sounds trivial, but - software is not an end in itself. Nobody needs software, everyone needs a solution to their business problem, such as cutting costs by automating repetitive work. When you start discussions, you should know what problem you want to solve for your clients and have a roughly outlined vision. A good software house should advise you on the technological details, ultimately they are the experts.

2. What is the role of AI in the project?

Once you know what you want to do and you’ve decided that there is a place for AI in your project you should consider whether the functionality that uses it is crucial or maybe it is just “the cherry on top”? In each of these cases the approach would be slightly different. Letting the contractor know will ensure that you are guided accordingly.

3. Are you sure you need AI?

You have already identified the places where you would need AI. However, perhaps it can be done more cheaply, or even better, using classic methods. You should work out this topic with a trusted consultant.

4. Identify what AI department you will need.

AI is a broad term as well as of technology and business. According as on what project you want to produce you will need experts in other "pieces" of AI.If you're going to analyse text it will be an NLP specialist, if the analysis involves images or video - then vision experts. Perhaps it's about predicting sales or finding unusual customer actions in an app? That's a job for a data scientist.

5. Do you need any hardware in addition to software?

Perhaps you want to analyze video images and automatically detect disturbing events, or with AR glasses support your work by automatically to read out of measuring instruments. But have you thought about choosing the right camera? The lens? Lighting conditions in which they will work?

6. Have you heard about similar solutions?

Are you able to point out solutions that work similarly? Yes? There is a good chance that you will also be able to reproduce a similar program and it will work well, taking into account marketing of course. However, be careful - "the devil is in the details". A good company should help you sift out unrealistic promises from the capabilities of current technology.

7. What data do you dispose of?

Consider if and what data you are collecting. Artificial intelligence algorithms need that informations to learn what you request of them. There needs to be enough data, it needs to be of good enough quality, and of course it needs to be capable of having the necessary amount of knowledge to meet your objectives.

8. How much data will the AI process?

Typically AI solutions are designed to analyze data. Video footage from cameras, documents flowing into the system or detection of disturbing signals from measured machines. Depending on how much data there is, how fast it flows, how quickly it is to be processed, and how long it is to be stored, the cost of the entire project may grow.

9. Moravec’s paradox

First of all, I refer you to articles devoted to this, for example on Wikipedia. Briefly - however, tasks that are easy and obvious for humans are not necessarily the same for AI or robots, and vice versa. An AI that beats chess champions was created much earlier than one that can recognize faces as well as a human, which is so obvious to us that we don't even think about it.

10. Check the rates of AI programmers

To estimate the price of your project, it's worthwhile for you to familiarise yourself with the rates of AI developers. This will give you an idea of how much you would have to pay by hiring an employee, a freelancer (rates may be a bit higher), or a software house (which will be more expensive, ultimately having to pay the employee and make the money themselves) for one or more months of the project.