Glut of AI agent tools faces paradox of choice, skills gap

The AI agent buyer's dilemma

AI agents, which have become dominant in the tech industry only in the last year, remain a high priority for most IT organizations long-term. However, according to a Gartner Tech FutureSight report this month, the market is already overloaded. 

"The current supply of agentic AI models, platforms and products far exceeds demand. …The tech industry’s growth forecasts for next-generation agentic AI systems aren’t sustainable," the report reads. "While investment in agentic AI continues to grow at an exponential rate, the technology faces significant headwinds in the short term."

Among these challenges is a disconnect between AI agent providers' pricing, which Gartner described as "opaque, unpredictable and insufficient to cover provider costs," and what customers want: "pricing that is simple, predictable and tied to business value."

The report didn't specify what the ratio of providers to consumers is, but this gap between pricing, revenue and cost is uniquely problematic for AI agent tools, according to Will Sommer, an analyst at Gartner who was among the authors of the report.

"The critical metric here isn’t the number of consumers relative to the number of providers. It’s the revenue each provider can generate from consumers relative to their costs," Sommer wrote in an email to Informa TechTarget. "Unlike the early days of cloud computing, for example, where excess data center capacity could be sold by the current class of hyperscalers to consumers at relatively affordable prices, agentic AI systems require significant investments in new IP underpinned by historically large investments in infrastructure and compute."

The rapid evolution of the market cited by Adobe's Jacobsen is a problem for AI agent tools providers as well, Sommers wrote.

"Gartner interviews with startup founders and CEOs indicate that the window for tech-innovation-based competitive advantage is shrinking rapidly to a matter of months," he wrote in the email. "Monetizing new agentic AI features will be very challenging as each innovation rapidly becomes table stakes."

Still, pressure to find a competitive edge using AI agent tools means enterprises can't simply sit on the sidelines amid this uncertainty, the Gartner report cautions.

"Avoid pulling back on agentic AI investments as markets calibrate to each generation of AI technology," the report states. "Companies that deprioritize product development will fall behind their peers."

IT leaders' early bets set the stage for AI agents

For Adobe's Jacobsen, finding value amid the noise around AI has meant focusing on overarching patterns and principles, rather than specific tools.

"Most immediately, what we're looking at is how we wrap MCP [Model Context Protocol] around all of our services so that people who don't know the code or the technical aspects of our various technologies could speak to these services using natural language," he said during the HashiConf keynote presentation. "How do we standardize on the usage patterns, instead of the actual tooling?"

For another IT decision maker, preparing for AI agents sometimes involves prioritizing features that help modify existing workflows for the transition, rather than which vendor adds AI features to a product first.

Hireology, a hiring software maker in Chicago, has already integrated some generative AI (GenAI) features in its customer-facing apps to generate job descriptions and message responses and has begun building its own AI agents using Microsoft Semantic Kernel, now Microsoft Agent Framework.

 As AI agents pop up in IT automation tools, Hireology sits at the nexus of several vendors adding them to their workflows, including Microsoft's GitHub Copilot, Anthropic's Claude Code, LaunchDarkly's upcoming Vega agent, unveiled Oct. 8, and IT automation tools from Application Performance Management vendor Datadog. It's also considering AI workflow automation from specialist vendor N8N.

The company's vice president of engineering, Scott Gainous, hasn't finalized most of the decisions between those tools, except for one -- a move from Datadog to LaunchDarkly for observability, based on LaunchDarkly's acquisition of Highlight.io earlier this year. LaunchDarkly officially rolled out its first integration of IP from Highlight this month, including integration with LaunchDarkly's Guarded Releases.

Before the acquisition, Highlight, founded in 2020, didn't have as high a profile in observability as Datadog , founded in 2010.  LaunchDarkly's Vega lags Datadog's Bits AI Dev Agent, already in public preview. Datadog also acquired its own feature flag IP with Eppo this year. But stronger human relationships with LaunchDarkly and the integration of Highlight with Guarded Releases made it Gainous's choice for troubleshooting AI-generated code releases.

How can you help me release faster? How can you help me recover faster? I'm not really interested in buying AI add-ons -- to me, that feels like a fad that will go away.

"What I'm really looking at is, does the [product], regardless of whether it uses AI or not, help me solve the problems I need it to?" Gainous said in an interview with Informa TechTarget. "How can you help me release faster? How can you help me recover faster? I'm not really interested in buying AI add-ons -- to me, that feels like a fad that will go away."

One agent-driven tool, LaunchDarkly's AI Configs, has also lent finesse to developing GenAI apps for Hireology.

"One of the things that I've run into with AI implementation is managing model targeting and prompts," Gainous said. "In your code, you might have text files, or it's embedded in the code, what the prompt should be, along with the data -- what this allows me to do is, just like a feature flag, be able to [swap] out the model we're using and what the prompt is … and it automatically retargets the model and the prompt."

Reckoning with an AI agent learning curve

In addition to pricing, another disconnect between sellers and buyers of AI agents is an emerging skills gap. Vendors often pitch AI agents to make up for a lack of qualified human workers, but AI tools require users to learn how to use them effectively, leading to further friction in the adoption process.

"While many companies are experimenting with agentic AI (and some have even implemented it successfully), most companies lack the talent and data to effectively implement AI solutions and deliver improved business outcomes," according to the Gartner report. "Gartner predicts that by 2028, over 40% of agentic AI projects will be canceled due to escalating costs, unclear business value or inadequate risk controls."

Respondents to a survey conducted between March and April 2025 by Enterprise Strategy Group, now part of Omdia, also raised concerns about AI agent skills gaps. Of 350 respondents, 33% said they have skill gaps in human-AI collaboration -- in other words, designing systems that allow humans and AI agents to work together effectively. Another top choice, for 32% of respondents, was that their organizations lacked AI and machine learning expertise.

These gaps create opportunities for vendors to step in, according to the Omdia report.

"Most organizations may find that hiring their way out of these skill gaps isn’t an easy near-term option, given the stringent competition for such talent at this time," the report stated. "Leaders should look to technologies for help, quickly ascertaining whether AI agent [tools] can help address those gaps with automated processes or low-code/no-code expertise."

But there are also growing communities that offer free and open source tools and training for human-AI collaboration and systems design workflows, such as the Breakthrough Method of Agile AI-Driven Development (BMAD). BMAD includes open source AI agent tools. There are also extensive training videos on YouTube by BMAD's creator, Brian Madison.Hireology's Gainous has looked to BMAD for guidance on using AI agents in software development.

"You can use [BMAD] with tools like Claude desktop or ChatGPT, or you can use it in IDEs [Integrated Development Environments] like Cursor and Claude Code," he said. "It's a good blueprint for this stage of AI engineering and development. It's providing me a lot of inspiration too, for building AI into our product."

Analysts predict market consolidation

Ultimately, industry analysts predict heavy attrition soon among vendors selling AI agent tools. According to the Gartner report, the likely winners will be those that are already large, powerful and have plenty of money to invest.

"My sense is that we will start to see attrition into the fourth quarter this year," said Chris Saunderson, an analyst at Gartner, in a separate interview with Informa TechTarget. Saunderson was not among the authors of the Gartner Tech FutureSight report on AI agents.

"On the startup side, it's fairly simple -- they're just going to run out of money," Saunderson said. "Established vendors are going to go through their product iteration cycle as well, where perhaps the first iteration of their agent is going to be replaced with something else, because they're learning at the same rate."

Brad Shimmin, an analyst at Futurum Group, said in an interview this month that big vendors that already have large install bases -- particularly those that already store large amounts of enterprise data -- will also have data gravity as an advantage as context engineering becomes increasingly important to using AI agents effectively.

"Everyone who is selling anything is building in generative AI. There isn't a separate market for generative AI where disruptors are entering the market with new products that are going to displace the existing vendors," Shimmin said. "That's really not what's going on here."

That's why a vendor such as IBM, not typically synonymous with IDEs and software developer tools, waded into the already-crowded agentic IDE market this month with Project Bob, according to Shimmin.

"One of the differentiators for that product is the way that IBM is managing the context engineering," he said. "So if a company has certain standards for documentation and code generation, Bob will build a corpus of knowledge around those standards… to make it specific to the company. That's the secret sauce that's going to make it successful."

Shimmin said other large vendors, such as Microsoft, Salesforce and AWS, will be able to capitalize on the same strengths.

In the meantime, while enterprises wait for the competitive picture to shake out, the ultimate vision of autonomous AI agents is quite far off, according to another recent Gartner report. Among 360 respondents to a survey the analyst firm conducted between May and June 2025, only 15% said they are currently considering, piloting or deploying fully autonomous AI agents.

"Although the long-term implementation of AI agents is less clear, most leaders do not expect them to replace applications or workers within the next two to four years," according to a Gartner statement published Sept. 30. "Only 12% strongly agreed AI agents would replace applications, and just 7% strongly agreed they would replace workers in that time frame."

Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.