AI pilots keep crashing, mostly because firms skip the prep, survey finds
(2025/11/13)
- Reference: 1763051807
- News link: https://www.theregister.co.uk/2025/11/13/survey_ai_projects_failure_no_prepare/
- Source link:
It is the best of AI times; it is the worst of AI times, depending on whom you ask. Nearly a third of firms are seeing almost total failure of their AI proof-of-concept (PoC) projects, while 46 percent are successfully moving more than 10 percent of theirs into operational use.
These figures are from Omdia's [1]2025 AI Market Maturity Survey , which claims that its research reveals a diverse and nuanced picture of the progress of AI adoption in corporations, with some managing to get projects into production while others are still struggling to get theirs off the ground.
That said, the largest single category is the 31 percent of enterprises that experienced a success rate lower than five percent. Only nine percent reported that more than half of their PoCs were accepted into production.
[2]
According to Omdia, the chief reason that PoC projects fail is not because of flaws in the particular AI technology being tested, but because both customers and vendors fail to appreciate the complexity involved in an AI deployment.
[3]
[4]
It says that firms need to grasp that a successful implementation starts with identifying a business challenge or an opportunity that AI could, in theory, address, and that much of the critical effort comes before the PoC even kicks off and will continue after its completion.
A recent [5]report from Cisco backs this up, finding that many companies simply don't know what they are doing with AI, and only 32 percent bothered to identify which human tasks they wanted AI agents to supplement or take over.
[6]
US Environmental Protection Agency CIO Carter Farmer also claimed that organizations are hastily rushing to deploy AI projects [7]without defining a clear use case , leaving them to wonder why it doesn't work.
[8]Superintelligence probably not happening, but AI will still reshape society, expert panel says
[9]Russia's first autonomous humanoid robot staggers and falls on debut
[10]Microsoft is building datacenter superclusters that span continents
[11]OpenAI GPT-5.1 adds more personalities, loses inhibitions
Having access to greater resources also helps. Omdia found that many firms in its survey that reported revenue below $100 million are running fewer than five PoCs, while at the opposite end of the scale, just four percent of enterprises are managing over 100 at the same time, and these were the largest companies in the survey.
In other words, the survey shows that AI tools are neither a total disaster nor the greatest thing since sliced bread, with a growing number of companies now managing to implement them in production workloads.
Reading between the lines, it does appear that actually getting AI to function properly can be a painstaking and complex task for many organizations, and so questions remain over whether the gains are actually worth all that investment.
Earlier this year, research from IT firm Lenovo found that [12]business leaders remain unconvinced that it is worth the expense. Most AI use cases have met business expectations, it said, but proving the return on these investments remains challenging and is one of the biggest barriers to adoption.
[13]
For its part, Omdia says that 30 percent of respondents in its survey reported that AI deployments aimed at increasing productivity had exceeded expectations, while 49 percent claimed they are meeting expectations. ®
Get our [14]Tech Resources
[1] https://omdia.tech.informa.com/blogs/2025/nov/ai-pocs-to-production-a-balanced-perspective
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.theregister.com/2025/10/15/cisco_ai_readiness_index/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2025/07/09/ai_projects_need_planning/
[8] https://www.theregister.com/2025/11/11/ai_experts_forecast/
[9] https://www.theregister.com/2025/11/13/aidol_russia_robot_fail/
[10] https://www.theregister.com/2025/11/13/microsoft_fairwater_dataceter_superclusters/
[11] https://www.theregister.com/2025/11/13/openai_gpt51_adds_more_personalities/
[12] https://www.theregister.com/2025/02/06/lenovo_ai_report/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[14] https://whitepapers.theregister.com/
These figures are from Omdia's [1]2025 AI Market Maturity Survey , which claims that its research reveals a diverse and nuanced picture of the progress of AI adoption in corporations, with some managing to get projects into production while others are still struggling to get theirs off the ground.
That said, the largest single category is the 31 percent of enterprises that experienced a success rate lower than five percent. Only nine percent reported that more than half of their PoCs were accepted into production.
[2]
According to Omdia, the chief reason that PoC projects fail is not because of flaws in the particular AI technology being tested, but because both customers and vendors fail to appreciate the complexity involved in an AI deployment.
[3]
[4]
It says that firms need to grasp that a successful implementation starts with identifying a business challenge or an opportunity that AI could, in theory, address, and that much of the critical effort comes before the PoC even kicks off and will continue after its completion.
A recent [5]report from Cisco backs this up, finding that many companies simply don't know what they are doing with AI, and only 32 percent bothered to identify which human tasks they wanted AI agents to supplement or take over.
[6]
US Environmental Protection Agency CIO Carter Farmer also claimed that organizations are hastily rushing to deploy AI projects [7]without defining a clear use case , leaving them to wonder why it doesn't work.
[8]Superintelligence probably not happening, but AI will still reshape society, expert panel says
[9]Russia's first autonomous humanoid robot staggers and falls on debut
[10]Microsoft is building datacenter superclusters that span continents
[11]OpenAI GPT-5.1 adds more personalities, loses inhibitions
Having access to greater resources also helps. Omdia found that many firms in its survey that reported revenue below $100 million are running fewer than five PoCs, while at the opposite end of the scale, just four percent of enterprises are managing over 100 at the same time, and these were the largest companies in the survey.
In other words, the survey shows that AI tools are neither a total disaster nor the greatest thing since sliced bread, with a growing number of companies now managing to implement them in production workloads.
Reading between the lines, it does appear that actually getting AI to function properly can be a painstaking and complex task for many organizations, and so questions remain over whether the gains are actually worth all that investment.
Earlier this year, research from IT firm Lenovo found that [12]business leaders remain unconvinced that it is worth the expense. Most AI use cases have met business expectations, it said, but proving the return on these investments remains challenging and is one of the biggest barriers to adoption.
[13]
For its part, Omdia says that 30 percent of respondents in its survey reported that AI deployments aimed at increasing productivity had exceeded expectations, while 49 percent claimed they are meeting expectations. ®
Get our [14]Tech Resources
[1] https://omdia.tech.informa.com/blogs/2025/nov/ai-pocs-to-production-a-balanced-perspective
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.theregister.com/2025/10/15/cisco_ai_readiness_index/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2025/07/09/ai_projects_need_planning/
[8] https://www.theregister.com/2025/11/11/ai_experts_forecast/
[9] https://www.theregister.com/2025/11/13/aidol_russia_robot_fail/
[10] https://www.theregister.com/2025/11/13/microsoft_fairwater_dataceter_superclusters/
[11] https://www.theregister.com/2025/11/13/openai_gpt51_adds_more_personalities/
[12] https://www.theregister.com/2025/02/06/lenovo_ai_report/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aRYOpl3L8mit-q54wJjf7AAAARc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[14] https://whitepapers.theregister.com/