LLMs & AI Agents · Project strategy
Why 40% of AI agent projects will be cancelled and how to avoid that outcome
Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027. The reasons are rising costs, unclear business value and inadequate risk controls. Our own projects point to six factors that get an agent past the pilot stage.
Julia Rose
Key takeaways
- Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, mainly because of cost, unclear value and weak risk controls.
- Many initiatives start under trend pressure, without a defined goal or realistic expectations.
- A tightly scoped first use case delivers visible results quickly and lowers the risk.
- Data quality, missing expertise and employee acceptance belong on the table before the project starts.
What Gartner predicts
In June 2025 Gartner published a forecast: more than 40% of agentic AI projects will be cancelled by the end of 2027. The analysis names rising costs, unclear business value and inadequate risk controls as the main reasons.
Source: Gartner press release, 25 June 2025
Agentic AI means AI agents that pursue goals on their own, make decisions and act over extended periods. According to Gartner, many of these projects are driven by hype. They lack a clear goal, a strategic framework and realistic expectations of what is technically and organisationally feasible.
What we see in projects
We help companies introduce AI agents, from setting the goal to running the system in production. Gartner’s assessment matches what we see in that work. Many organisations are eager to “do something with AI”. What exactly often stays open, because trend pressure, the wish to innovate or outside expectations started the initiative.
Before the start, three questions then remain unanswered:
- What exactly should improve?
- Which processes are suited to AI and AI agents at all?
- Where is the best place to start?
Across many projects, six factors have emerged that decide whether an AI agent delivers lasting value.
Six factors behind successful AI agent projects
1.Start with a business objective
Successful projects begin with a challenge that matters to the business. A tool or a particular technology is a weak starting point. AI agents pay off when they improve processes, speed up decisions or make better use of resources. Without a concrete objective, the technology quickly loses focus.
2.Be realistic about data
Every AI agent works with data, and not every dataset is usable. An early assessment shows whether the structure, quality and accessibility of the available data are good enough for the intended use case. Settle data protection and usage rights at the outset. That avoids delays once the project is underway.
3.Involve the right expertise early
Building and introducing an AI agent takes knowledge from several fields: process understanding, data strategy and technical delivery. Few companies have all of it in-house. In those cases we add our project experience to existing teams.
4.Start with a focused use case
Trying to build a comprehensive agent system from day one usually runs into complexity and inflated expectations. Many companies come to us expecting a single AI agent to solve many very different problems at once. That ambition to tackle everything together is what stalls projects.
A clearly scoped, manageable use case works better. It delivers visible results quickly, lowers the risk and shows where to go next. Examples are an agent that classifies incoming emails or a chatbot that handles complex requests.
A focused start does not limit what comes later. A successful first use case is a foundation to build on, step by step, until a robust system with real value takes shape.
In practice: RPP Group
The public affairs consultancy RPP Group works with ChatRPP, an internal system of five specialised AI agents. It is built for one clear area: the routine work around policy documents. ChatRPP summarises long documents, prepares data and writes first drafts in the consultancy’s tone. Time spent on that work dropped by 70%.
5.Involve users early
A system can run flawlessly and still fail if employees do not accept it. Bring in the people who will work with it early. Clear communication, transparent decisions and a real say in the design raise acceptance and build trust in the new technology.
6.Create orientation and build knowledge
The term AI is used loosely, and it is often unclear what sits behind it. In workshops and strategy sessions we work with companies to calibrate expectations, assess technical options realistically and prepare sound decisions. That produces projects that work technically and make economic sense.
Our AI workshops are one way in. If you already have a specific process in mind, the AI discovery workshop tests whether automating it is worth it.
Conclusion: AI agents need a clear direction
The technology is there and the potential is large. The reasons Gartner gives and our own project experience point the same way: initiatives fail because of missing clarity, unrealistic expectations and a lack of strategic focus. Without a realistic goal and clean execution, ideas stay stuck in the pilot or disappear altogether.
AI cannot solve problems nobody has understood and framed first. Anyone who starts with inflated expectations is likely to be disappointed. A clear goal, a solid data foundation and an experienced partner produce progress you can measure and build on.
Planning an AI agent?
In the process analysis we map your workflows, rank the possible use cases by return and propose an architecture. Fixed price from €3k.
Request a process analysisFrequently asked questions
Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027. The forecast dates from June 2025 and names rising costs, unclear business value and inadequate risk controls as the main reasons.
Agentic AI refers to AI agents that pursue goals on their own, make decisions and act over extended periods.
With a business objective and a tightly scoped use case, such as classifying incoming emails. Before the start, check the data, data protection and usage rights, and involve the future users.
Usually there is no clear goal, or the first agent is expected to solve too many different problems at once. On top of that come data that is not good enough for the use case and employees who do not accept the system.
Julia has been part of theBlue.ai since 2019 and has accompanied the development of AI applications in the enterprise environment since the company’s early days. In her role as Marketing Lead, she works closely with the engineering and consulting teams and makes complex technical topics understandable and accessible for decision-makers.
In her articles, she writes about practical experience from enterprise AI projects, as well as the challenges and opportunities of using AI in companies.