5 questions to answer before CIOs launch their first AI projects

Many organisations underestimate the complexity of AI projects, which leads to delayed project launches, or failure to deliver business value, reports Gartner.

Comments

Not understanding what AI can be best used for is a top reason for failure

Many organisations underestimate the complexity of AI projects. This can lead to delayed project launches, or failure to deliver business value, reports Gartner.

Organisations need to understand that they don’t have to always build the solution in-house, which is a common reason for failure, reports Gartner in a new Gartner research for CIOs who are deploying AI projects for the first time.

“AI technologies are complex, yet still immature, and there is a lack of AI skills in most mainstream organisations,” according to the report authors, Sandy Shen, Erick Brethenoux, Nick Ingelbrecht and Denise Ganly.

They state CIOs should be aware of the alternative options for reliable and performant AI solutions, such as those supplied by software vendors or custom-built by service providers.

These, they add, are the preferred options to in-house development unless the project is of strategic importance.

Have you included business-domain, IT and analytics experts?

They note how respondents to a recent Gartner Research Circle study cited a number of organisational and technical challenges such as lack of skills, data quality issues and finding the right use cases for their AI projects.

They advise CIOs in charge of these programmes to ask five questions that can help identify potential failures as well as improve performance for future initiatives.

Identify use cases: Have you clearly defined the business problem?
Define business success: Can you define success from a practical business standpoint?
Organise the team: Have you included business-domain, IT and analytics experts?
Procure the solution: Have you investigated all sourcing options and ensure ongoing maintenance?
Adapt business processes: Have you operationalised the solution to get the most out of technology?

Organisations often start AI projects out of competitive pressures, fearing they will be left behind if they do not invest in the technologies.

But, as the Gartner analysts point out, not understanding what AI can be best used for is a top reason for failure.

Gartner recommends CIOs to work with business users to understand their daily tasks and challenges. They can then identify opportunities where AI can play a key role. This should be an ongoing rather than a one-time exercise.

CIOs can then establish a list of use cases so they can work on several projects in parallel and apply lessons learned. They are also advised to give priority to projects that are technically easy to develop especially when the team is new to AI technologies.

Defining business success, the second question, requires the expected outcome to be quantified, according to the Gartner analysts.

This can be difficult as the outcomes of AI projects are typically less certain than traditional IT projects. Gartner says organisations need to be open to an experimental and heuristic approach to AI projects.

Key business metrics should be shared with senior executives, AI project leaders, data scientists, IT and operations team, and be prominent in the dashboard.

Gartner says having the right people assigned to the project is critical to success. This means building a cross-departmental, multifunctional group that includes business-domain experts, IT staff and data scientists.

“This team needs to be empowered to make decisions and to take actions when needed. It can be an actual team with direct reporting lines, or of a taskforce with ‘dotted line’ reporting to various parties,” says Gartner.

Plan for the discontinuities that AI systems introduce within business processes, and continuously adapt decision models

At the same time, organisations must also decide whether to buy, outsource or build the AI solutions. Most likely, organisations will use a hybrid of solutions, but Gartner says they should consider the impact of each approach.

On operationalising a solution, Gartner says this goes beyond efficiently allocating the appropriate IT resources.

“It also means anticipating — in a systematic way — the ongoing evolution of AI elements, planning for the discontinuities that those systems introduce within business processes, and continuously adapting decision models.”

Operationalisation or industrialisation of AI techniques is “less glamorous” than the development of these techniques, notes Gartner.

But neglecting this phase will prevent organisations from fully getting the business value of their AI initiatives, and compound lack of confidence in AI techniques.

“Successful organisations have adopted various technologies to manage operational decision services in a reliable, repeatable, scalable and secure way,” the report concludes.