How did it do that? Over half of IT leaders not confident explaining AI model decisions
A new report from Anaconda shows 51.4 percent of IT leaders say they’re not very confident in explaining AI model decisions to regulators, executives, or customers.
As a result they’re over-promising and under-delivering, and this trouble communicating creates unrealistic expectations for customers and stakeholders. 26 percent of respondents identify difficulty demonstrating ROI as a top concern. What’s more just 22 percent say they would describe their organization’s AI deployment as ‘strategic’.
The report, based on a survey of over 200 data scientists, engineers, researchers, and consultants working with AI and machine learning models, finds the most common ways people showcase ROI from AI initiatives are productivity improvements (58 percent) and cost savings (47 percent).
However, AI isn’t a technology that can immediately deliver these results. In fact, over 57 percent report that it takes over a month to move from development to production of AI projects. 75 percent of respondents say employees and customers don’t begin using AI tools until one to six months after they’re deployed. Plus over 30 percent say trust is either still developing or is non-existent after three months or longer.
Although 57 percent report that it takes over a month to move from development to production, organizations leveraging open source AI tools are gaining competitive advantages in innovation speed. 92 percent of respondents use open-source AI tools, with 76 percent saying there's more priority on open source this year than last. Organizations are realizing they need access to the best tools available globally, not just what's in their vendor's catalog.
“There’s a line between collecting metrics and measuring ROI with those metrics. Understand where AI makes you better. You should always be measuring and looking at the outcomes, just perhaps not evaluating AI initiatives strictly on what metrics move. Now is the time to hone these tools,” says Steve Croce, field CTO at Anaconda.
You can get the 8th Annual Anaconda State of Data Science and AI Report from the company’s site.