Companies are embracing AI but they are encountering problems in deployment, with nearly three-quarters of data management decision-makers citing issues with model explainability, according to a survey from Capital One and Forrester.
The report said 73% of respondents faced difficulties in transparency, traceability and explainability of data flows when they try to deploy and scale machine learning to more use cases.
“Businesses see massive potential in applying machine learning but encounter headwinds in their data,” said Dave Kang, senior vice president and head of Data Insights at Capital One. “This can hinder businesses from seeing actionable insights, and perversely shy away from adopting and operationalizing ML solutions in the first place.”
Another key obstacle was breaking down data silos, with 57% believing internal silos between data scientists and practitioners inhibit ML deployments. Also, 38% said that data silos across the organization and external data partners pose a challenge to ML maturity.
Diverse, messy data sets (36%), difficulty translating academic models into deployable products (39%) and reducing AI risk (38%) were pain points for data decision-makers as well.
“To overcome challenges, organizations must focus on the business outcomes of ML and build partnerships with proven leaders in their ML journey,” according to the report. Moreover, “without better explainability and transparency” top executives and directors “have trouble seeing business benefits after adopting AI/ML solutions."
“If there’s no clear connection to ROI, executive buy-in decreases, which reinforces data silos, creates struggles in driving actionable insights and inhibits operationalization,” the report warned.
Capital One and Forrester surveyed 150 North American data decision-makers to determine their ML goals and challenges.
Partnerships To Drive ML Maturity
According to the survey, 67% said they intend to leverage partnerships to fill ML staff gaps. Around 37% said they’re currently partnered with a third party for ML model development and plan to grow that collaboration.