90% of AI Agent Vendors are Scams
A week ago Gartner published a report that concludes with that more than 90% of AI Agent vendors are scams, and that software companies are doing "agent washing" to rebrand their existing legacy technology as AI agents. Their motives obviously is to leverage the hype surrounding AI agents in 2025.
The wording in the Gartner report is as follows; "Gartner estimates only about 130 of the thousands of agentic AI vendors are real", which implies that less than 10% are actually capable of delivering what they're promising. Doing simple math on that fact results in the following fact; "More than 90% of AI agent vendors are basically scams!"
What is an AI agent
In order to understand the above numbers, we need to first define an AI agent. This allows us to create a minimum requirement your technology must be able to implement in order to be considered an "AI agent". Below is the shortest possible definition that exists.
An AI agent is an LLM capable of executing tasks
With the MCP protocol implementing true AI agentic behaviour should actually be fairly easy. Still, 90% or more are cheating and claiming they're capable of delivering AI agents without having any substance to their claims. Hence, in order to determine whether your vendor is a scam or not, you can ask them two simple questions.
- Are you using an LLM?
- Can you make it execute tasks?
Unless they're answering yes to both of the above questions, walk away! We've seen "AI agents" without AI. We've seen PowerPoint presentations sold as "AI agents". And we've seen billion dollar companies lying to their customers and share holders, claiming they've got technology they simply don't have.
AINIRO AI Agents
At AINIRO we've had AI agents since April of 2024. The way our technology works is by allowing users to create and declare what we refer to as "AI functions". An AI function is basically just an old fashion "function", that's exposed to the LLM, in such a way that it allows the LLM to parametrise your function and execute it. Such functions can be for instance;
- Send an email
- Returns last year's total revenue
- Schedule a meeting the 1st of August
- Create a new customer in HubSpot
- Etc ...
The above are just examples of course, and what functions you need your AI agent to be able to execute depends upon your use case - But the point is "the LLM can do something".
Such AI functions are immensely useful since by combining such functions together, you can initiate complex workflows, that integrates with your existing systems, and does tasks on your behalf. Sometimes, these are tasks that would require hours of manual work, which you can now simply ask your LLM to perform on your behalf, offloading your workload, and making you more productive.
No-Code AI Agents
At AINIRO there's no "agent washing". We've had the above capabilities for more than a year, and in fact, we had it long before the MCP protocol emerged too. We were the first company on earth to deliver AI Agents in fact! And, not only do we have real AI agentic capabilities, but we've got technology that allows our users to use no-code and low-code constructs to create such AI agents. You can watch me create a minimalistic AI agent in the following video ...
Not only does the above illustrate how we've got real AI agent tech, but it also illustrates how we can use AI to implement AI agents. You can see a screenshot of an AI agent "function" in the screenshot below.
The above AI functions returns aggregate values from a database table, allowing you to ask questions such as.
- What is the smallest asset ever created?
- What is the largest asset we created in January?
- What is the sum of assets in our database?
- What's the average asset in the month of February?
The above are examples of what you'd previously need a dashboard with KPI capabilities to deliver. By simply dropping the entire idea of a "GUI", and instead relying upon natural language as the user interface, we can deliver AI agents with thousands of such "AI functions", replacing arguably entire suites of systems with a simple natural language interface.
Why vendors are "agent washing"
There is literally nothing through human history that has gotten the attention that LLMs and AI agents have been getting the last couple of years. In addition, a real AI agent is extremely powerful, and its promise is to significantly automate your existing job. This implies there's money to be made by being able to deliver AI agents.
At the same time, AI agents are extremely disruptive. Companies such as SalesForce and HubSpot have generated billions of dollars on their existing software, which is based upon a graphical user interface. This becomes a problem for existing "legacy vendors", since you can just create an AI agent accessing their APIs, and/or database directly - Resulting in that AI agents are slowly eating these existing company's lunch.
This becomes an existential threat for those already having legacy systems that are now being threatened by AI agents. In addition, it creates opportunities for those wanting to start their own companies and deliver AI agents to others. Basically ...
It's an AI Agent gold rush
Why we care?
At AINIRO we can deliver amazing AI-based solutions, including highly complex AI agents, capable of integrating with your existing systems, automating important aspects of your business workflows. We've seen how this saves our clients tens of thousands of dollars, every single month. In addition we've got technology that allows us to build such complex functionality without even having to code.
However, more than 90% of buyers of such AI agents are basically being ripped off today, because they talked to "the wrong guy", believing in the promise that "their guy" would deliver an AI agent to them - Not realising that "their guy" couldn't, not even in theory, deliver what was promised, because they simply don't have the required technology to deliver what they promised.
This makes it difficult for us to be noticed, because everybody claims they can deliver AI agents, while only a few are actually capable of delivering.
The fix
The fix is to verify that your existing vendor actually has AI agentic capabilities. The simplest way to do that, is to show them the above YouTube video by yours truly, and ask them the following questions; "Can your platform do this?" If the answer is no, just cut your losses, walk away, and find somebody else that can actually deliver what they're promising.
And if you're one of these vendors, incapable of figuring out how to make the LLM "do something", relax! You don't need to reinvent the wheel, we've already licensed our entire technology stack as open source, and you can find a link below.
Wrapping Up
The AI market is hot like the sun, and software companies are either in a complete state of panic, or in a complete state of euphoria because of the changes that are literally happening in front of our eyes today. However, according to Gartner, more than 90% of software companies promising you AI agents are literally incapable of delivering AI agents, because they don't have the technology. And they've been messing with it for almost 3 years now, so there are few reasons to believe that will ever change anytime soon.
Hence, as you're shopping around for AI agents, please take some time to evaluate the technology of your vendor, because not everything that shines is gold. If you want help from us to assess your current platform, or to implement an AI agent for you, you can contact us below.
