Use the 'MVP approach' to increase success of digital commerce AI projects: Gartner
- 23 October, 2018 06:00
Organisations looking to implement AI in digital commerce need to start simple
Gartner predicts that by 2020, artificial intelligence (AI) will be used by at least 60 per cent of digital commerce organisations and that 30 per cent of digital commerce revenue growth will be attributable to AI technologies.
“Digital commerce is fertile ground for AI technologies, thanks to an abundance of multidimensional data in both customer-facing and back-office operations,” says Sandy Shen, research director at Gartner.
In a recent Gartner survey, about 70 per cent of digital commerce organisations report their AI projects are very or extremely successful.
Gartner says the survey covered 307 digital commerce organisations - including 30 in New Zealand and Australia - that are currently using or piloting AI to understand the adoption, value, success and challenges of AI in digital commerce.
Three-quarters of the respondents said they are seeing double-digit improvements in the outcomes they measure. The most common metrics used to measure the business impact of AI are customer satisfaction, revenue and cost reduction.
The survey found a wide range of applications for AI in digital commerce. The top three uses are customer segmentation, product categorisation and fraud detection.
Despite early success, digital commerce organisations face significant challenges implementing AI, reports Gartner.
Foremost of these are the lack of quality training data (29 percent) and in-house skills (27 percent) in deploying AI in digital commerce.
Thus, Gartner advises: “Assess whether there is sufficient AI talent in-house to develop and maintain a high performance solution yourself. If not, go with a commercial solution of proven performance.”
Gartner notes on average, 43 per cent of respondents chose to custom-build the solutions developed in-house or by a service provider. In comparison, 63 percent of the more successful organisations are leveraging a commercial AI solution.
“Solutions of proven performance can give you higher assurance as those have been tested in multiple deployments, and there is a dedicated team maintaining and improving the model,” says Shen, who is co-author, together with Mike Lowndes, of the Gartner research on How to Increase Chance of Success for Digital Commerce AI Projects.
Aim for under 12 months for a single AI project - from planning, development and integration to complete launch
“Organisations looking to implement AI in digital commerce need to start simple,” says Shen.
“Many have high expectations for AI and set multiple business objectives for a single project, making it too complex to deliver high performance. Many also run AI projects for more than 12 months, meaning they are unable to quickly apply lessons learned from one project to another.”
She advises organisations to aim for under 12 months for a single AI project - from planning, development and integration to complete launch.
If the project has to go beyond 12 months, divide the project into multiple phases, and aim for under 12 months for the first phase. This can be done by running multiple projects in parallel that won’t increase risk of failure, she states.
Organisations should also ensure there is enough funding to ensure they get high-performance solutions. “Budget more for talent acquisition, data management and processing, as well as integration with existing infrastructure and business processes,” says Shen.
Taking on the minimum viable product approach increases the likelihood of success in AI projects, reports Chen.
She says organisations can:
· Set no more than two objectives at once. Revenue increase, customer experience and cost reduction are common objectives — but trying to hit all birds with one stone is a formula for failure. Limit to one or two objectives and fully deploy the solution to drive home those objectives, she states.
· Set the right expectation of what AI can do. Use the MVP approach to break down complex business problems and develop targeted solutions to drive home business objectives. Use AI to optimise existing technologies and processes rather than to try to develop breakthrough solutions, she advises.
· Limit the number of innovations per project. Organisations often view AI as a high profile investment. They thus try to do new things in multiple places, such as channels, applications, data management, algorithms and architecture.
But doing so many new things at the same time with one project significantly increases risks. Limit to one or two new ways of doing things at a time, and do those fully, she says.
For example, an organisation can develop a data cleansing and tagging system to speed up the data cleaning process, but not pair this with a new approach to search.
At the same time, Gartner recommends application leaders to work on multiple focused projects and quickly apply lessons learned from these initiatives. This will improve their chances of success, says Shen.