How much does artificial intelligence (AI) cost in 2018?

Artificial intellligence

There has been so much talk about Machine Learning and Artificial Intelligence lately, as it has become obvious – they are drastically changing the world. Due to how promising these technologies are and the amount of benefits they already deliver, many companies are willing to use them for their business transformation.

Although Machine Learning and Artificial Intelligence are related, they are not the same thing, and it’s worth understanding the differences between the two. Artificial Intelligence (AI) is the concept of computer systems that are capable of performing tasks that usually only human can do, such as decision-making, visual perception, speech recognition and so on. Machine Learning (ML), in turn, is a field of computational science that incorporates numerous technologies. By processing large quantities of submitted data, ML reveals connections between various factors, and therefore builds new analytical models. Though ML is a subset of AI, technically speaking it’s more accurate to think of it as its current state-of-the-art.

So how much does it cost to utilize artificial intelligence? Lets’ discover!


Future Is Now

ML certainly has a lot to offer. The newest technology becomes pervasive in our lives, as it starts to be widely adopted by many companies across different industries.

With its promise of automating routine tasks as well as offering creative insights, industries in every sector from insurance to healthcare are reaping the benefits. Online selling platforms leverage ML algorithms to suggest their customers what they might want to buy next basing on their online behaviour. It has been seized as an opportunity by marketers – machine learning chatbots prove to be effective at generating more leads than before and at providing enhanced customer experience.

Businesses are gaining competitive advantage with AI-driven solutions, which now are used in making accurate predictions, and generating business insights. The advanced voice and image recognition capabilities have been adopted by easy-to-use mobile applications.

Image modelling applicationAzati’s Image Modeling Application that allows to immediately change wallpapers on the photo taken from a smartphone


Project Phases

The atrificial intelligence project price depends, first of all, on the work being done to develop a product. The development work is usually split into several phases. Having a general idea of the project phases may help you make a rough estimation of cost. The following application roadmap has been found useful for developing systems based on ML algorithms, and is adopted by Azati.

1. Discovery & Analysis Phase

The purpose of this phase is to conduct a feasibility study and set up business and project objectives.

The work on a project starts with analyzing the customer’s formal business processes, data assets and current metrics. At this stage the project team defines success factors (expected metrics improvements), applicable technological stack, timeline and budget and reflect all that in the corresponding documentation.

At the end, parties find out whether or not AI concept is possible, and if it is, define the scope of work needed for the next move, namely prototype development.

If all required data, processes and metrics are available in the required format, the phase takes 5-7 working days in average. Typically we do it for free.

2. Prototype Implementation and Evaluation Phase

Next step is devoted to the implementation of an AI-based prototype. That step extremely important to calculate the costs of AI development.

A prototype is a business model created in order to test feasibility and proof of concept. It can be a limited, text or drawing-based mock-up, or a more sophisticated code-based prototype, depending on the project complexity and tools (screen generators, application simulation programs, or design tools) it was created with. Prototypes are shown and discussed with the clients.

Prototyping is a great technique that allows software professionals to validate requirements and design choices. Prototypes are quick and cheap to produce, and flexible to adjust. The risks and costs associated with software implementations are dramatically reduced, as the requirements are well-discussed early on, before development begins.

We are striving to make this step as inexpensive as possible, typically prototype costs are about $25’000 including estimation for the MVP or another next stage required.

3. Minimum Viable Product (MVP)

The main difference of the prototype and MVP is that the later is a viable product, meanwhile the former is more of a software visual representation.

An MVP is a real product with a set of basic and functional features developed with bearing the findings from the prototype in mind. An MVP works with the customer’s actual data and is exposed to a small group of real customers as the minimal first version of the ultimate product solution. Their feedback is very relevant, as it is way less expensive to modify the system at this stage, than when it is fully developed. Average MVP costs vary between $35’000-$100’000, depending on the total project size and complexity.

4. Product Release

At the last phase, the product with complete set of predefined features is developed and then launched to the market. The preceding steps put lot of emphasis on the requirements elicitation and validation, so at this step the end product is made with minimal risks. The cost of this phase is usually estimated during previous stages. Yeah, we may try to skip that step to reduce ai development price, but it will significantly decrease project quality.


ML-specific factors affecting the final ai development price

The process of developing an ML-based system has a number of distinctive features which determine the final artificial intelligence development costs.

Data Issues

The development of reliable ML-system depends not only on excellent coding – its is often the quality and quantity of the training data that plays a pivotal role.

First of all, large representative data sets are required in order to reasonably capture the relationships that may exist both between input features and between input features and output features. If there isn’t enough data, there are options like collecting more data or using external data sources. Another solution is using data augmentation methods to artificially increase the sample size.

One more requirement is that the data must be easy to work with –  it must be well-organized, stored in the proper format in one operational data store (data warehouse). Though sometimes it’s not the case, so some preparational activities (e.g. ETL processes) are required.

Next cost-effective factor is whether or not the data is structured. It is easier (consequently cheaper) to work with well-structured data. In some cases, structured data is subject to data cleansing, tidying, data type conversion, working with missing, extreme and unexpected values, dealing with outliers and obvious errors and so on (the full list of data transformation methods is not covered in this article).

Although, in practice a great number of companies have unstructured (e.g. free-form text notes), or semi-structured (e.g. XML, email) data. There is a whole class of ML-algorithms created to make use of this kind of data, and typically these projects’ cost is a bit more.

Regardless of the type and condition of the company’s data assets, there are ways to make the value of it.

Required Algorithm’s Performance

The sufficient algorithm’s performance is another key cost-effective factor, as oftentimes a high-quality algorithm requires a round of tuning sessions, which increase the final costs.

It’s worth noting, that performance sufficiency varies according to the client’s business objective and the cost of wrong prediction. A broker would take advantage with the system, that produces 55% of correct predictions, for it already enables them to earn money. But a 90,9% accurate system aimed at defining test results on a disease with treatment being lethal to false-positive patients, is by far not satisfactory.

Why do we need to care about data processing speed and algorithm’s performance to understand how much does it cost to utilize AI? The answer is simple: all modern ai applications are based on data and uses it for the learning process. So if your data is processing slowly – it will take more time to teach the neural network. In that way, the time to teach neural network will dramatically increase according to the number of data needed for the neural network to learn.


So, how much in numbers?

It’s a common misconception that leveraging such an elaborate technology as ML must cost a fortune, but it’s not true nowadays.

If just a few years ago only the web-giants such as Google, Microsoft and Facebook could afford to build ML-powered software, now a wide number of companies can also do this. Thanks to the emergence of various tools, libraries and frameworks for building ML-based software the ML technology is becoming more available for businesses.

Prices are calculated for each case in particular. For a fraud detection insurance-case the price ranges between ~ 100k- 300k$. Numbers greatly differ in practice – it all depends on the project scope and complexity, customer and system requirements and other factors mentioned before.


Would you like to shift to a new AI-powered paradigm?
Call us at +1 (973) 597-1000 or fill out the form below for a free consultation.


Your Name (required)

Your Email (required)