Five areas to focus on to ensure AI success

Artificial intelligence has rapidly become another ‘rush ahead’ innovation, according to Forrester analyst Craig Le Clair. That is, organizations want to start AI projects before they have grasped the key issues involved with success.

To boost the likelihood of a rewarding project, Le Clair says chief information officers, chief technology officers and chief data officers should focus on five key areas: business, learning, technology, project and organization. AI efforts should be clearly defined, limited, and have understood goals, data inputs and source knowledge, he says.

AI risk.jpg
Green LED lights and rows of fibre optic cables are seen feeding into a computer server inside a comms room at an office in London, U.K., on Tuesday, Dec. 23, 2014. Vodafone Group Plc will ask telecommunications regulator Ofcom to guarantee that U.K. wireless carriers, which rely on BT's fiber network to transmit voice and data traffic across the country, are treated fairly when BT sets prices and connects their broadcasting towers. Photographer: Simon Dawson/Bloomberg

“There is no single road map for implementing AI,” Le Clair stresses. “Training the machine, integrating data, building intelligent agents, and setting design and deployment milestones will vary based on your application area, the scale and scope of your use case, and the AI building blocks you use.”

Despite that, Le Clair says there are 17 risks across these five areas that are common to all AI projects.

“They capture issues that range from ethics and learning bias to preparing the post-AI infrastructure. CIOs must pay attention to these areas to avoid getting burned,” Le Clair explains.

Le Clair breaks down the 17 top risks to an artificial project in his new report “Tackle Five Key Areas of Risk For Successful AI Projects.” They are:

“Business risks are, unsurprisingly, of great concern”:

  • Your AI initiative provide a poor customer experience
  • The application doesn’t actually engage intelligently with the real world
  • AI efforts fail to meet overall business goals
  • AI introduces new compliance, legal and ethical challenges

“AI learning requirements introduce critical risk”:

  • The recurring learning process doesn’t work well
  • You introduce training bias
  • Milestones for the learning process are unclear

“Careful build-versus-buy decisions reduce technology risk”:

  • You make the wrong decision to build or buy
  • You overreach, pushing for AI when a simpler technology will do
  • Bots expose your firm to security problems
  • You don’t fully understand your vendor’s capabilities

“Project risks center on poor estimates, overoptimistic skills assessments, and data traps”:

  • You poorly assess the implementation time and budget
  • The AI knowledge base stumbles over data problems
  • Project skills fall short

“Organizational readiness offers strong protection against AI’s pitfalls”:

  • Your foundational infrastructure is inadequate
  • You may too many decentralized and uncoordinated projects
  • Poorly defined technology roles keep the CIO and her team out of the loop

As Le Clair points out, “AI’s possibilities are much clearer than the specific steps and issues that CIOs must tackle, yet those who take a risk on AI projects will be rewarded, even though many initiatives will fail.”

“For the 17 specific risk areas, determine whether you are ready to deal with them to avoid being burned,” Le Clair says. “Select AI projects with focused goals and limited data sets. Remember: The best way to minimize the risks from your AI initiatives is to pick projects with clearly defined, limited, and understood goals, data inputs and source knowledge.”

For reprint and licensing requests for this article, click here.
Artificial intelligence Data management Data and information management
MORE FROM DIGITAL INSURANCE