Five AI mistakes and how to avoid them in your organisation

Dr Leslie Kanthan, CEO and co-founder of TurinTech, identifies five AI mistakes and how your organisation can avoid them

The road to adopting AI isn’t necessarily a quick one. You need to collect the right data, find the right tools for your organisation, and team members need to be trained on how to build and optimise the AI models. However, even if you have found the right AI for your business and it’s been onboarded well, there’s still the risk that you will not be getting what you want or need from it.

If an AI mistake is made, it has varying levels of severity. For example, AI that predicts real estate prices (similar to Zillow’s recent issues) but isn’t set up properly can result in over or under-evaluations. AI that is used to predict cryptocurrency trading trends, but doesn’t get retrained on new data could lead to missed opportunities. However, for a business that is looking to adopt AI to improve processes and make them more efficient, even the smallest mistake can result in lost productivity, money, or reputation.

With this in mind, here are five common AI mistakes and what organisations can do to avoid them:

Focusing on the tech

Although AI is, of course, very technical, when it comes to implementing AI in your business, it has to be linked back to business needs. Without this, it will undoubtedly not have the impact you want or need it to have.

Before onboarding takes place, outline the main aim of your AI project with all key stakeholders and determine what you need to achieve from a business perspective. For example, if you’re adopting AI to predict customer churn, make sure you outline and quantify the ideal outcomes (e.g. improve retention rate by 5%), rather than getting caught up in the technicalities of implementation.

How organisations can drive value from AI on the edge

Mike Ellerton, partner at Go Reply, spoke to Information Age about Reply’s recent research conducted into edge AI, and how organisations can drive value from the technology. Read here

Overlooking architectural fit

Despite temptations to get started with AI, it can be hard to reap the rewards you set out to achieve if you don’t have the right data infrastructure in place which ultimately leads to a whole host of mistakes.

If an organisation wants to get value from data, it must be able to collect, store, and process it before even considering adding AI on top. If they don’t, companies run the risk of implementing analytics that aren’t mature and it could make teams more prone to a whole range of mistakes.

Limiting the team members that can access the data and AI

Another common mistake is limiting the number of individuals or teams that can see the data and insights that the AI is providing. With the capacity to become available for immediate reuse by other models across the organisation, for example, a quality feature store can reduce the duplication of data engineering efforts and allow new ML projects to bootstrap with a library of curated production-ready features.

When onboarding AI, you must identify each of the internal stakeholders who will benefit from the data and insights, and ensure that data can be shared timely and seamlessly.

Training machine learning models to be future-ready

As digital innovation continues to accelerate, we explore how machine learning models can be trained to be future-ready. Read here

Not optimising existing AI models

Data is ever changing. Business expectations and regulatory requirements can also change rapidly. As a result, the performance of models can quickly become degraded, even obsolete. To maintain performance, models need to be continuously optimised. AI optimisation platforms empower businesses to build and enhance their models on-demand, based on specific criteria.
Businesses can therefore roll out AI to various clouds and devices at scale, while also maintaining accuracy and efficiency. As recently outlined by Rick Hao, AI optimisation leads to more efficient code and greener, cheaper, and fairer AI for all. Without AI optimisation, AI can be costly and difficult to scale.

Forgetting to think green

COP26 in Glasgow and the UN’s Sustainability Goals have elevated the conversations and urgency around creating platforms and businesses that prioritise sustainability. Your AI shouldn’t be any different.

Some people are surprised to find out that the average carbon footprint of AI is equivalent to five times the lifetime emissions of an average car. But, it doesn’t have to be this way. By optimising AI, you can lower memory and energy consumption and simultaneously reduce carbon emissions. Not only will this improve the sustainability plans of your AI, but will also support the long term sustainability of businesses in general.

Mistakes can be costly, time consuming, and severely impact a company’s reputation. By putting simple steps in place, organisations can save themselves from these mistakes and reap the rewards of well-implemented AI.

Written by Dr Leslie Kanthan, CEO and co-founder of TurinTech

Editor's Choice

Editor's Choice consists of the best articles written by third parties and selected by our editors. You can contact us at timothy.adler at