Source: S&P Global Media Portal by S&P Global.
Investment momentum for AI in data science and analytics appears strong over the next 12 months: More than two-thirds of surveyed organizations (68%) expect spending to increase, while only 1% expect it to decline. Current adoption is characterized by an assistive approach using natural language instructions to conduct data science and analysis tasks within a vendor platform. However, investment in multi-step analytical workflows, in which agents plan, reason and execute autonomously, is a priority in the next year. Meanwhile, data privacy and security remain major organizational concerns.
A recent study conducted by 451 Research by S&P Global provides actionable data and insights into the key dynamics driving the data science and analytics markets for organizations, IT vendors, suppliers and investors.
The Take
The assistive nature of AI in data science and analytics reflects the first phase of adoption, which our survey data suggests is likely to continue for the next 12 months. However, during this period, investment in multi-step analytical workflows is a key priority. This suggests a significant increase in the sophistication and maturity of use cases, as well as the bona fide adoption of AI agents. Current “agentic” implementations are largely rebranded features in vendor platforms used with human agency.
We expect autonomy to land first in high-volume, repeatable, routine and low-risk workflows that have already been delivered by agentic features and are therefore proven and trusted, carrying minimal risk if they fail. One example is a data prep workflow involving autonomous data discovery, profiling and cleansing to remove duplicates and null values. Automation will also be suited to use cases where data inputs, objectives and constraints are machine-executable once clearly defined and grounded in the associated business logic, such as predictive analytics. However, the successful execution of multi-step autonomous workflows will also require further investment in AI infrastructure, including data privacy and security — concerns that will be magnified by agentic autonomy.
Summary of findings
AI adoption for assistive features is established and projected to grow in the next year. Almost two in five organizations (38%) use assistive data science agents, and more than a third (36%) have adopted assistive business intelligence/analytics agents. Moreover, growth remains healthy, with approximately one in five organizations (21%) piloting data science agents and a similar proportion (22%) in proof of concept with BI/analytics agents.
In the next year, an additional 21% of organizations plan to implement data science agents and 17% plan to implement BI/analytics agents.
Human oversight and control prevail in current use cases. Partial autonomy governed by pre-defined rules is the prevalent approach today, in use by more than half of organizations — 52% for data science and 55% for BI/analytics. About a quarter of use cases in both data science (27%) and BI/analytics (24%) involve minimal AI autonomy, with users making decisions based on AI-generated insights or recommendations.
Assistive agent adoption is highest for data-related tasks and coding. A third of organizations (33%) have implemented AI agents for data profiling or cleansing, and a third (33%) for writing code such as SQL queries and Python scripts, making these the top use cases today. This pattern reflects that adoption is strongest among data practitioners and expert users, such as data scientists and analysts, for assistive, labor-intensive activities.
Multi-step autonomous analytics workflows are a top investment priority in the next 12 months. Just over one-fifth of organizations (22%) intend to implement agents that plan, reason and execute autonomously in the next year, ranking it second among next-12-months investment priorities.
The top planned use case — documentation of models and workflows (24%) — also supports this shift in adoption, as documentation is critical to the context engineering required for autonomous agent execution.
Data privacy, security risks and high cost are major barriers to adoption. The top current or anticipated obstacle to AI agent adoption, by a wide margin, is concern about privacy and security risks, cited by almost half of organizations (45%). High implementation/operational cost is another major obstacle, cited by nearly a third of organizations (31%). Cost is also key in prioritizing use cases — it is cited by almost half of organizations (45%) as an important factor in determining new uses, the top overall response.
Most organizations assume responsibility for failures of autonomous AI agents, but formal accountability policies are not ubiquitous. Almost two-thirds (64%) of organizations hold an internal team member or end user responsible when an autonomous AI agent fails. Just over a third of organizations (36%) place responsibility on an internal technical team, while 29% assign accountability to an end user such as a data scientist, analyst or business stakeholder. However, around one in seven organizations (15%) have no accountability strategy in place.
Sustainability is not a dominant consideration. Fewer than one in five organizations (17%) emphasize the expected energy or carbon impact of AI use cases when deciding which projects to prioritize, and just 18% identify sustainability concerns as an adoption barrier.
Furthermore, nearly a quarter of organizations (23%) have no plans to implement tools to analyze environmental impact or energy use.
Are we the data? Data ownership and control dynamics in the GenAI age
Want insights on AI trends delivered to your inbox? Join the 451 Alliance.
The author used a proprietary S&P Global AI platform in the production of the report that this blog post was based on. It was subsequently peer-reviewed, fact-checked and edited before publication.
This content may be AI-assisted and is composed, reviewed, edited and approved by S&P Global in accordance with our Terms of Service.
