5 things to consider to make your AI project a success
According to Gartner analysts, up to 85% of big data projects fail¹. I suspect that for AI projects, especially if you count the ones that never start, the failure rate is even higher.
What are the reasons for the high failure rate with AI projects? Drawing on my technology consulting experience from working at Accenture UK and wider transformation project experience within the highly complex environment of the UK National Health Service (NHS), I wanted to identify the causes and potential solutions.
As highlighted in a previous post, although there is a widespread belief among senior executives that AI can give their companies or help them maintain a competitive advantage, only a few had implemented AI in a meaningful manner.
I noted three reasons that could be contributing:
- AI for the sake of AI — where the approach is driven by the ‘need to do AI’ and not business challenges and objectives
- Automating inefficiency — where AI is used to automate current processes, which may be non-value adding or inefficient
- Getting disillusioned with AI — taking a short-term view on Return on Investment (ROI) expectations, leading to missed opportunities
How can organisations design and initiate meaningful AI projects?
Based on my experience, I believe considering the following 5 things will help:
1. What are the objectives?
It is important to consider what problem you are trying to solve and where you are currently. Especially for an early AI project, where the demonstrating value is important, it is helpful to select objectives and outcomes that can be easily measured (and compared against the investment).
2. What is the solution?
When considering the solution, consider how AI can support the key expected benefits. What is AI ‘good at’ (or not) and what is realistic? For example, AI has a proven track record in improving and automating classification and identifying ‘categories’ and making recommendations.
3. How is the solution incorporated into Business As Usual (BAU)?
It can be easy to develop an impressive AI solution or train a ‘machine learning model’ that is not connected to day to day activities, requires excessive work to set up and ‘run’, and is a ‘one off’ requiring manual updating of the underlying training data. Careful consideration needs to be given to the following. First, have you got a clear understanding of the business processes and how the new AI solution will ‘fit in’ to minimise interruptions and inefficiencies? To facilitate this, end users and operational managers should be engaged early on and a ‘co-development’ approach to solutions followed. Second, what is the data ‘supply chain’ — where did the underlying data come from, will it be easily accessible in the future and will it be available in the same ‘format’? Third, have you identified the training requirements to embed the solution into the business processes and will there be an adequate governance structure to manage decisions and any issues that arise from using the AI solution? For example, it is important to understand the limitations of the solution (and the underlying training data) and to have ‘human intervention’ (also known as the human in the loop) when required.
4. How can the solution be delivered?
There are two elements to consider. First, it is important to structure the design and delivery roadmap to include ‘Discover’ stage and to include ‘quick wins’ early on to build momentum. Second, the delivery team needs to be assembled. There may already be a strong internal AI team. However, scarcity of talent and implementation timelines may require supplementing the team through an external delivery partner or providing focused training and coaching to develop the required specialist skills.
5. What is the business case?
The final stage is to develop a business case the clearly highlights the following:
- Value to the customers
- Value to the internal business
- Implementation plan including timelines and delivery resources
- The investment required and the ROI
- Key risks and how these can be mitigated
What are the additional insights?
Considering all of the above, there are three underlying themes that surface:
- AI is not always the solution — The data supply chain may not support an AI-based solution or it may be that a non-AI based analytical solution will provide what is required, at a better value for money
- (Some) Knowledge of AI is required from the very beginning — This could be in the form of ‘technology interpreters’ — those that have the dual understanding of business problems and AI (as explained in a previous post)
- Leadership buy-in is a requirement for success — as with other change initiatives, senior executives must define the business objectives and must commit the time required to provide ongoing support and guidance to the project
Feel free to drop me a line at firstname.lastname@example.org — always happy to chat.
ConscientAI works with clients across different geographies and industries. We have a pragmatic way of working that helps our clients identify key problems and understand the potential benefits before investing in the right solutions.
Title image photo by rawpixel.com from Pexels