Welcome to the MediaGamma blog

The gap is widening between those with a handle on their AI strategy and those only scratching the surface of its potential.

How to decide on your business’ first AI project: 5 steps

Posted by Jeet Bhatt

business AI project

How to decide on your business’ first AI project: 5 steps

Scientists and their discoveries have led to great revelations throughout history and, in recent years, none have proven more impactful than the discovery and development of artificial intelligence (AI). First coined by John McCarthy as early as 1955, delivering on the promise of AI has been a goal of scientists ever since.

Although AI isn’t yet the stuff of science-fiction, it’s already having a meaningful and profound impact on businesses and the modern economy. 47% of digitally mature companies now implement AI in some form, and 54% of those claim that doing so has significantly boosted productivity.

Still, over 50% of companies are reluctant to get started in this forecasted $118.6 billion industry, with a 43% expressing significant concerns about doing so. In large part, this is due to confusion surrounding implementation, security, and other such potential setbacks. But, as the market continues to shift towards real-time data analysis, deep learning, and more, failing to at least understand whether AI could work for you is no longer an option.

That leaves you facing one significant question — what should your first AI project be, and what can you do to reduce the risks that have kept you at bay until now? That is exactly what we will help you answer here. 


Step 1: Zone in on your business goals 

Outcomes and business goals need to sit at the heart of any IT investment — that’s the only way to keep the project on track and make sure it delivers something meaningful. AI projects are no exception to this rule. 

The very first thing you should do when looking to embark on an AI project is to write out a list of the main problems in your business you would like to resolve, and the main transformational outcomes you would like to achieve. Then start looking at the viability of common AI tools to deliver on those outcomes. For example, if customer service is on your list, you could easily look at chatbots and other AI-augmented messaging tools on the market. 

Then, you need to assess roadblocks and challenges that may get in the way of your AI project. Likely, that will centre around budget and your ability to focus internal resources on building the expertise required to get your AI project off the ground successfully. In summary, before investing any money into an AI project, consider the following: 

  • Your current budget
  • What you’re aiming to achieve 
  • How/if AI can help you to deliver on these goals
  • Your timeframe for implementation
  • How you intend to track progress
  • How you define and measure a successful outcome


Step 2: Assess your data

The sheer amount of data that businesses are now dealing with is largely behind the AI revolution. In fact, with internet users generating as much as 2.5 quintillion bytes of data each day, many organisations will face little choice but to turn their attention to AI where data handling is concerned. 

As well as driving your AI efforts, though, data and its quality can help you to choose and perfect your first project here. After all, quality datasets were behind even early machine intelligence implementations like IBM’s iconic Deep Blue, which relied entirely on data input for outward success.

Only by assessing the data you have access to can you judge potential outcomes, consumer needs, and even whether you have what it takes to make intelligence pay, with a comprehensive data landscape fundamental for avoiding missing variables and performance setbacks. This is important for judging the suitability of an AI project for your business at all. However, it can also be a mechanism by which you determine what project to prioritise. Follow the data and choose the functions for which you have the best data sets.   

When judging your data quality, be sure to focus on crucial considerations including: 

  • Representation
  • Diversity
  • Balance
  • Exhaustive representation
  • Sufficiency (do you have enough data to achieve your goal?)

Admittedly, analysing data on such a broad scope can be difficult, but this alone might be a decent indicator of your AI needs moving forward. If you struggle to understand the data you have right now, for instance, then analytics automation or assistance from a better-equipped AI consultancy is probably your best chance at a positive future.

Once you better understand the data you have coming in, it’s time to start thinking about how you could use the information. Again, focus on areas where you have quality data. However, this is also an opportunity to think more abstractly about what that data analysis is telling you. 

For example, an ecommerce retailer might notice spikes in shopping cart abandonment — as is true in around ¾ of ecommerce purchase. This reveals a great opportunity for automated personalisation software that targets emails — but only if you have the customer data to feed into that engine. Simply tally those facts, and think hard about what they’re pointing towards, or what they alone could bring to this AI implementation. 


 Step 3: Know what you’re capable of

While your efforts here are likely an attempt to improve your current capabilities, it’s also fundamental that you think about what you can do right now, especially when it comes to tech. Again, this can come down to basics like your current budget, but, you also think about things like your existing communications, data storage, and so on. 

Even more importantly, this is the ideal time to consider what you’re currently capable of in terms of workload. Given that this is your first AI project, the chances are that you don’t have an in-house team just yet. That’s not a significant issue, but 32% of companies let this single fact hold them back. Make sure you don’t fall into that same trap by instead considering the alternatives. For instance, turning instead to pre-trained APIs like Google Prediction or MLaaS such as Azure ML could see you enjoying basic capabilities through the cloud and other platforms that are ready to go.

Of course, the downside to integrations like these is the fact that they still may be outside of your skill set as it stands. In fact, many businesses overcommit on the assumption of simplicity. If you want guaranteed success on your first AI project, you should look for help with AI consultancies. Even business with extensive in-house teams partner with outside experts to gain specialists skills on-demand. But the value is compounded when you are just getting started.

A well-selected consultancy will be able to help you assess the foundations of your project, and then help execute your plan. Then, while a consultancy takes care of your initial AI focus, you’ll be free to develop in-house capabilities for a hybrid approach that integrates intelligence firmly into your company moving forward.


Step 4: Develop your strategy

At this stage, you have some idea of your requirements, capabilities, and potential pain points. Now, it’s time to develop your all-important strategy. An amalgamation of the above points, this is what will ultimately uncover whether you need AI, as well as helping you to integrate it into your business accordingly. 

The main thing to remember here is that there’s no such thing as a one-size-fits-all AI strategy, and keeping your company in mind every step of the way is fundamental to success. If you jump in with complex AI programming, then you’re asking for confusion, trouble, and ultimately wasted capabilities. 

Instead, let AI research or your on-hand consultancy guide you towards the best simple applications to get you started. Don’t be afraid of pre-trained models to get going, and always consider the most straightforward implementations possible. If you’re interested in machine learning, for instance, a pre-trained keyword software is always going to serve better than a sophisticated speech recognition alternative. Build your strategy around such ideas to make sure that you don’t run before you can walk, even when you have professional help behind you. 

This is also time to think about strategy surrounding your AI approach itself. Determining, once and for all, whether you intend to take care of this yourself, with the help of a consultancy, or using a hybrid model will have a significant impact on your project choice and your chances of success.


Step 5: Always review

Last but by no means least, reviewing your AI project at every stage can help you to determine whether it’s working for your bottom line. Ultimately, integration in any AI application is not a one-time thing. Instead, this is a process of trial and error, and you need to continually revisit the drawing board to stand any chance at successful implementation.

The sad fact is that an estimated 85% of AI projects fail to deliver on their intended outcomes, and a lack of review is often behind such issues. Again, an AI consultancy could drastically simplify this process by overseeing your AI project for you, and alerting you to any potential problems before they unravel your efforts. Starting planning with the help of experts and an AI Discovery Programme is a great way to guarantee that your resources are put to good use.  


With the right training and hybrid focus, your team should then come to understand what exactly it takes to build the layers of a successful AI project, meaning both that your first effort stands to succeed, and that your future implementations don’t let you down. It’s an exciting time in AI. Get planning and good luck! To learn more about starting a successful AI project...

Get in touch

Jeet Bhatt
Jeet Bhatt
Head of AI Consulting Services

Related Posts

Google Privacy Sandbox: What it means for your programmatic ads

Google’s Privacy Sandbox looks like it might be the end of programmatic advertising as we know it, with third-party cookies largely set to become a thing of the past, and reliable automated data collation at risk along with them. The only way to prevent this from significantly unravelling online marketing efforts is to focus on shifting data landscapes, sources, and handling on a business-wide scale.

Is AI the right solution for your business: a 2020 assessment guide

AI adoption is on the rise at a rate of around 25% year-on-year. That’s an increase of approximately 270% in the last four years alone, and it’s a figure that set to continue rising as big data and tech potential continue to grow hand in hand. Still, despite this popularity, AI is far from a panacea for everything, as is evidenced by the fact that only 23% of businesses report using such solutions regularly. More disturbingly, a reported 85% if AI projects fail to deliver on their intended goals.  

How to train your own NLP model transformers

In recent years, search engines have led the way in natural language processing (NLP) and the benefits it stands to bring across business landscapes. Focused around machine learning capabilities, applications like Google’s BERT work to process natural language without human input. Used right, such systems can remove redundant work processes for faster and more efficient handling, analysis, and translation of written or spoken input from various sources.