Organizations that are at the forefront of adopting generative artificial intelligence (GenAI) are running into a plethora of issues. A survey of 1,500 AI practitioners and decision-makers conducted by S&P Global Market Intelligence on behalf of WEKA finds that, on average, organizations that have funded artificial intelligence (AI) projects only have six AI projects deployed at scale, compared to 16 in limited deployment and 10 still in a pilot phase. In the average organization, 51 percent of AI projects are in production but lack large-scale delivery.
The single biggest AI challenge organizations are encountering is not, for example, finding and retaining AI skills. Instead, storage and data management (35 percent), followed by compute (26 percent), security (23 percent), and networking (15 percent) are the major inhibitors of adoption, the survey finds. Among organizations that identified data management as a challenge, the availability of quality data is the biggest impediment (34 percent).
Additionally, 80 percent of respondents anticipate a rise in the volume of data used to develop AI models over the next 12 months. About half (49 percent) are forecasting growth in data volumes of more than 25 percent.
Exploring the AI minefield
Data, of course, is not the only AI challenge organizations are encountering. Many are also finding it difficult to find the graphical processor units (GPUs) needed to train and deploy AI models (30 percent), with 40 percent now considering AI accelerators as an alternative option. Nearly half of respondents (46 percent) are relying on cloud service providers for model training, with just under a third (32 percent) relying more on specialist GPU cloud providers.
Demand for those resources continues to far outstrip supply. A full 88 percent of respondents are now actively investigating use cases for GenAI, the survey finds.
Nearly two-thirds (64 percent) of respondents said they are concerned about the impact of AI/machine learning (ML) projects on their energy use and carbon footprint, with a quarter (25 percent) being very concerned. A total of 42 percent said they have invested in energy-efficient IT systems to address the potential environmental impacts of their AI initiatives over the past 12 months. Of those, 56 percent believe this has had a “high” or “very high” impact. Making changes in data infrastructure vendors (59 percent) and AI project scope (57 percent) have had a “high” or “very high” impact.
The MSP’s role in scaling AI initiatives
The one apparent thing is most of the challenges organizations are encountering have little to do with data science. Instead, managing IT infrastructure as the amount of data that needs to be processed and analyzed continues to scale well beyond the skills and expertise of the average IT operations team is the biggest hurdle.
Not every organization, of course, is building GenAI applications but regardless of size almost all are at the very least researching at least one GenAI use case. Many of them, however, will in the absence of any help from an MSP be likely to conclude they don’t have the IT expertise required to succeed. It then falls to the MSP to make sure those organizations understand that with their added expertise it’s truly possible to achieve using AI.
Photo: Christina Morillo / Pexels
