Making a success of AI project

There’s no question that artificial intelligence (AI) has already moved from being a technology rooted in science fiction to one of the most powerful tools businesses can use to transform how they operate. From driving down costs through the automation of repetitive tasks to the creation of new services and products that enable companies to reach new customers, few innovations have such a wide range of potential applications.

“Businesses can no longer avoid AI; it’s here to stay. Today, not incorporating the benefits of AI into your
operations is like not having a website,” says Marta Markiewicz, head of data science at Objectivity, digital transformation specialists.

For example, retailers that don’t integrate advanced recommendation engines into their online platforms will be at a disadvantage when competing with rivals driving forward with AI-backed solutions. Great recommendations engines significantly improve the experience of a website visitor and drive up sales.

But the relatively recent advent of business-friendly AI tools and a general lack of understanding of how to utilise AI products is hindering this technology from becoming a truly revolutionary solution. To effectively execute an AI project, corporate leaders need to have a specific goal or improvement in mind, rather than simply embracing AI because it is seen as a “must” for progressive companies.

“We’ve found it useful to divide the different types of AI projects into three main categories. The first class of projects involves the use of off-the-shelf AI products, moving on to bespoke AI solutions further down the spectrum, and finally high-risk, high-reward projects that require a great deal of resources and a long-term commitment to achieve,” says Ms Markiewicz.

A company may consider the use of AI as a service component in a traditional type of business project, as this would demonstrate the power of this burgeoning technology in a relatively low-risk approach. Yet it’s unlikely that such a limited use of AI would have a profound impact on how a firm operates, with examples of this method being the use of handwriting or voice recognition in typical projects.

The next step along the AI spectrum sees businesses using bespoke solutions to make an improvement that would be impossible without the use of AI. “This could include, for example, creating an AI-based solution to recognise products on a supermarket shelf to check they are in the correct position and priced correctly or to create a digital twin that can accurately forecast revenues,” says Ms Markiewicz.

Objectivity know first hand the benefits of bespoke AI solutions as they have been using this approach to improve understanding of their future revenues. By using AI to predict their resource needs effectively, Objectivity have been able to ensure their IT recruitment plans are in line with accurate forecasts and not rely solely on guesswork and less accurate estimates.

“It’s not some side feature; it’s become a key feature that has revolutionised part of our business and made it easier for us to plan. We’ve seen a great improvement in margins because now we have very few people who are not working on customer projects as we haven’t recruited too early, so it’s an enormous improvement in financial performance,” says Ms Markiewicz.

There is an element of risk in creating bespoke AI solutions as, although patterns of AI usage are well known and understood, applying these models in the specific context of your business has yet to be done. But the rewards can be significant and, as Objectivity has shown, have the power to change how you do business.

Starting the AI project with full immersion in our clients’ business is a very important part of our process

Leaders who want to transform their business fundamentally can embark on an AI project that starts from scratch and attempt to create new algorithms with the potential to advance what is possible to achieve with AI. This high-risk approach requires completely new models to be developed and is usually reserved for only the largest companies and universities that have a large pool of resources to tap.

No matter what category of project you choose, there will be challenges and pitfalls that have the capability to derail your AI solution. By using the wrong model or data, businesses can fail to solve the right problem within the company and result in no clear positive result.

Spending time before starting an AI venture to consider if it is the right approach for your business can help reduce the risk of wasting resources on an inappropriate project that does little to improve the bottom line.

At this initial stage in the overall process, Objectivity can work with clients to consider if there is another project or approach that would be more relevant for the specific challenge the business is dealing with and ensuring the risk-to-reward ratio is as attractive as possible.

The Augmented CRISP approach created by Objectivity helps to simplify AI projects and makes certain the right problem is chosen to solve, gathers the right team to make the collaboration a success and establishes the criteria that address the risks the project may face.

“Starting the AI project with full immersion in our clients business is a very important part of our process. It allows us to understand not just the business and the hypotheses, but also the culture of the organisation, the strengths of their teams, what they will bring to the party. And, of course, the data they hold. These elements are vital to improve the chances of the project being a success. Without a proper understanding of all these factors, some organisations can be tempted to start data-crunching without a real comprehension of what they want to achieve and what is possible,” says Ms Markiewicz.

Possessing high-quality data that is reliable and available in sufficient quantity is clearly of central importance for any AI project as it will be extremely difficult to define what exactly success looks like if the data used is not classified.

“Only when you are sure you understand the data and when you are sure the terms of reference are clear for both Objectivity and yourself can we then move to proof of concept (PoC). PoCs must be chosen carefully and are designed to remove risk by tackling the hardest parts first, in isolation. It can be very tempting for clients to build the low risk parts first, because it feels like progress. But it’s not progress. PoCs help everyone to iterate faster, save money and increase the chances of success,” she says.

It’s critically important to collaborate with a partner that has practical experiences of delivering AI solutions for both themselves and their clients.

“Successful projects need a diverse team of specialists. For the last six years, we’ve been developing our people, methods and processes to maximise the value we create in AI projects for ourselves and our clients. There’s a reason why we work closely with Microsoft, and we’re a valued data and AI partner. I think it’s because we’re good at this stuff and we’re passionate at creating value for our clients,” Ms Markiewicz concludes.

For more information please visit