Five critical success factors for AI implementation | NTT DATA

Fri, 22 September 2023

Five critical success factors for AI implementation

How to get the most value from data and AI

Despite the ongoing AI revolution, numerous organisations continue to grapple with integrating AI initiatives into their broader business strategies in an effective way. Business leaders are manifestly aware of the need to unlock the potential of generative AI and harness its benefits. Given that the journey will not be straightforward, knowing the best approach to take is often the first challenge.

Generative AI offers businesses the potential to digest and summarise information and validate existing knowledge. For the data scientist, it serves as a tool to empower practitioners, enhancing efficiency and effectiveness, allowing them to concentrate on delivering tangible value. However, its effectiveness relies on having prior knowledge of what specific questions to ask the AI and how to interpret its responses. Our partners such as Microsoft and Salesforce, are incorporating generative AI into their product suites for these kinds of tasks, but we see that tailored solutions for specific client challenges are still in the experimental phase.

As I reflect on NTT DATA’s experience of real-world data science solutions, there are a range of critical success factors for organisations to get the best value from data and AI capabilities.

1. Solve real business problems

When it comes to implementing AI solutions, it's crucial to focus on solving a business problem that genuinely requires attention. While it may be tempting to embark on a standalone project to explore new technologies, this approach rarely leads to long lasting, impactful outcomes.

Start small and scale – by focusing on the business challenges which have large impact and generate substantial value over a short period of time, you will be able to promote more interest and senior buy-in for the proposed solution. This will lead to more successful change initiatives and uptake, as well as significant senior investment.

We recently delivered a risk prediction model for a client which had tangible operational value, reducing the time spent manually identifying errors and increasing the productivity of the team. By focusing on this specific challenge, we generated interest from across the organisation, including senior leadership, who wanted to democratise the solution and enable us to explain in more detail how the model worked to the wider business.

While staying informed about the latest developments in AI through horizon scanning is important, it should not be the starting point for ‘solution-ising’. Instead, decision-makers can rely on experts who understand how to apply these approaches in the service of a business challenge.

2. Maximise value through business change

When implementing machine learning and AI solutions, it's essential to consider the implications for business change. Simply deploying a model is not enough; it is necessary to ensure that the outputs are embedded in the business process to deliver the intended value. A key factor in the success of these solutions is education and understanding.

Even when you operationalise and deploy the technology, colleagues will be hesitant to use it unless they comprehend its capabilities and trust its outputs. This emphasises the importance of focusing on the use of models which are easily explained. In some cases, we find ourselves referred to as ‘the people doing AI’ in conversations around the office. This perception creates an air of mystery and complexity, leading some to feel that the technology is beyond their comprehension – or worse, a direct threat. To overcome this, it is crucial to demystify AI and educate users, fostering trust and encouraging widespread adoption within the organisation. How this is approached will vary from business to business and from team to team so a tailored strategy is required.

3. Enable experimentation and prioritise data engineering

Industrialising data science products is a critical aspect of any successful AI implementation. It’s essential to have the infrastructure in place that allows models to be easily evolved and re-deployed.

In some organisations this ‘ML Ops’ capability is missing and we need to roll this out as part of the solution. Likewise, good data science is built on a foundation of strong data engineering. AI and ML heavily rely on large volumes of high-quality data, and organisations often face challenges in terms of data availability, access permissions, and data quality. This results in valuable time being spent upfront on addressing these issues rather than allowing data practitioners to focus on their area of expertise. By addressing these challenges and prioritising data engineering, organisations can lay a strong foundation for successful AI implementation and maximise the value of their data-driven initiatives.

4. Create synergy between data scientists and the business

It’s not enough for data scientists to possess the technical expertise to develop sophisticated algorithms and models. They need to understand day-to-day processes and issues that their colleagues experience. Close collaboration between data science and business teams ensures that AI solutions are developed with a deeper understanding of the business context and goals.

By working together, data scientists can tailor their models to address specific business needs, resulting in more relevant and impactful insights. In turn, business teams can provide valuable domain knowledge and feedback that guides the direction of AI projects, ensuring that they align with strategic priorities. Ultimately, the collaboration between data science and business allows for the development of AI solutions that not only deliver accurate predictions and insights but also leads to value-based outcomes for the organisation.

We recently worked with an organisation who hadn’t collaborated closely enough. Their data science team hadn’t engaged thoroughly  with the business to understand the specific issues and requirements, which ultimately led to an overly technical solution that did not assist  decision making.

5. Always consider the ethical implications

No discussion of this topic would be complete without considering the ethical implications of AI implementation. Considerations such as data privacy, algorithmic bias, and transparency, play a vital role in ensuring the responsible and fair use of AI technologies. Fortunately, leading organisations have their own ethical frameworks – as does the UK government – what is crucial is that these are acknowledged as part of the process.

Take your next step to integrate AI

AI solutions are never stand-alone, instead a collaborative approach of multiple disciplines is needed. Selecting a partner with a proven track record, operating at the forefront of data science innovation, ensures that that all these critical success factors are considered to pave the way for the optimal integration of AI into your business strategies and unlock its transformative power. 

If you’re looking for the best approach for embracing AI's potential and enhancing your organisation's approach to problem-solving and value delivery, book an appointment with me today:


How can we help you

Get in touch