APPLICATIONS
Cost pressures are compounded by the high failure rate of AI projects. Research from RAND Corporation shows more than 80 % of AI projects fail to deliver their intended business value, but Markus asserts the technology itself is rarely the main culprit:“ Reasons for failing can be customers not having well-specified outcomes they want to achieve; organisational challenges because people don’ t come together; or challenges with underlying data and access to data.
“ Companies get carried away with what AI might be able to do reliably, but that’ s what happens in hype cycles.”
Networks built for humans, not agents Beyond organisational and cost pressures, Markus identifies a more physical problem: much existing network infrastructure was never designed for the way AI agents behave.
Unlike human users, AI agents can generate dozens of simultaneous queries, connecting to sensors, data sources and other AI services all at once.
aimagazine. com 101