As early as January 2025, companies approached artificial intelligence (AI) like a shopping spree. They bought copilots, niche tools that were vertical or horizontal, bots for single channels, summarizers, insight engines, knowledge assistants, and a growing list of point solutions that promised quick wins with customers and employees.
The test and learn mindset was exactly what software buyers were looking for.
Fast forward to Q1 2026, and the market is still moving quickly, internal capabilities are still forming, buying looks easier than building, and the operating model hasn’t changed to fit these new capabilities. Yes, AI tools are still trying to fit into old models.
I think that phase is ending by late 2026, and we’re already starting to see the change happening.
When CX Network asked its community how they acquired new AI capabilities in 2025, 74 percent said they worked with third-party vendors. Now in 2026, that figure has fallen to 50 percent.
The change doesn’t mean outside vendors are disappearing. On the contrary, it means the market is changing shape. Companies are becoming more selective, buying fewer standalone tools, and are expecting more from the platforms they already use. And in this time, the flashiest demo no longer is winning the pitch.
Finally, companies are getting more confident building on their own too, limiting the need for some software, and I have personally seen this happen with engineering teams building survey tools and LLM-based text analytics capabilities.
There is also a “grey middle” emerging where many companies aren't "building" from scratch (e.g., writing code), but they aren't just "buying" (SaaS) either. Instead, they’re taking open-weights models and refining them on private data; this is more of a “compose” strategy.
Three shifts are driving this change
The first is that AI is being absorbed into the software stack. Gartner has predicted that by the end of 2026, 40 percent of enterprise applications will include task-specific AI agents, up from less than five percent in 2025. These numbers are a clear indication of how companies are changing their buying behavior. So instead of adding yet another vendor to solve a narrow problem, companies will increasingly ask whether CRM, service, workflow, analytics, or collaboration platforms can handle the job natively.
For CX teams, that means AI will show up less as a bolted-on solution and more as a built-in capability inside the systems they already use every day.
The second shift I’m seeing is that model choice is widening while model costs are falling. Stanford’s 2025 AI Index showed that smaller models got much better, and the cost of using capable models dropped sharply. At the same time, enterprise use of open source AI has become mainstream. When you add in recent McKinsey insights that found more than half of organizations using AI are already using open source technologies somewhere in the stack, often alongside proprietary tools, vendor power shifts to the buyer.
When model performance becomes more interchangeable, and when open models become credible options, vendors can no longer rely on model access alone as their moat much like they did in the early days of ChatGPT. Vendors now need to compete and differentiate on workflow, integration, governance, and measurable business value.
The third shift is that interoperability is becoming a serious business issue, and is no longer something just your CTO cares about.
As AI moves from chat interfaces into orchestrated actions, tools need to work across systems, data sources, and models. That is one reason open standards are gaining traction. In late 2025, OpenAI, Anthropic, and others helped establish the Agentic AI Foundation under the Linux Foundation to support open, interoperable infrastructure for agentic AI. By early 2026, the Foundation had expanded quickly, with dozens of new member companies joining to reduce fragmentation and advance shared protocols.
The message from the market is pretty clear: that even the largest players understand that customers do not want to get trapped in closed ecosystems for every AI decision. And in a market as dynamic as this one, picking the right pony (read: unicorn) means the difference between winning and losing in business.
When you put these three signals together, I think the direction of travel becomes much clearer. And that is that the AI vendor market is moving from a chaotic land grab to a layered market.
At the top will be large platforms that embed AI into everyday work. In the middle will be orchestration, governance, and infrastructure players that help companies connect models, systems, and controls.
Finally, around the edges will still be specialists, but they will need to prove that they do something meaningfully better than a platform feature that is already included in the enterprise contract.
What it all means for a CXO about to go to RFP or buy a new solution
First, stop showing up as the buyer of shiny tools and start showing up as the executive who understands where AI can improve real customer outcomes. This sounds really obvious, but it is where many CX leaders get pushed out of the room. If the conversation is only about model choice, architecture, or cost per token, the CIO, CTO, CDO, and procurement team will dominate it. If the conversation is about resolution rates, trust, speed to answer, agent effectiveness, complaint reduction, retention, and effort, the CXO is a critical player in the room.
Second, the CXO needs a stronger vendor evaluation model. In the next phase of the market, the smartest question will not be, “Which AI vendor should we buy?” It will be, “Where should we buy, where should we extend what we already have, and where should we keep optionality?” That requires a different skill set. CX leaders will need to assess vendors on features yes, but also on data access, integration requirements and lift, governance needs, portability, pricing and total cost of ownership, and how quickly the tool can be integrated into daily work by users.
Third, the CXO’s new love language has to be about risk. The governance for general-purpose AI under the EU AI Act are already in force, and the broader regulation becomes fully applicable on August 2, 2026, with some exceptions. Even for leaders outside Europe, it’s clear that buying AI solutions is as much about innovation as it is a governance decision. For CXOs, that means customer advocacy now includes pushing for explainability, transparency, human oversight, and sensible escalation paths when automation fails.
Now for the practical implications of all this
The future CXO will need to be part operator, part translator, and part market decoder. Operator, because AI only matters when it improves the day-to-day service and support deliver as an example. Translator, because the CXO has to connect technical choices to customer and business outcomes. Market decoder, because the vendor landscape will keep shifting and the winners will not simply be the loudest or the largest. CXOs need to become futurists.
OpenAI’s 2025 enterprise report found that enterprise use is not just increasing, it is deepening into structured workflows and more complex tasks. While this research can be seen as self-serving, it goes without saying that this is also the opportunity for CX leaders. So forget about experimenting with another chatbot, think about redesigning how work gets done across service, operations, and customer-facing teams. Then you can consider which AI solutions amplify your people in the new operating model.
AI solutions are only successful when they do three things effectively: deliver a financial return, increase operational delivery/velocity, and amplify your people’s ability to execute.
The AI vendor market is not becoming simpler. But it is becoming easier to read. If you can define where AI belongs, where it does not, and how to keep the customer represented when the buying decisions get technical, your voice in the organization only becomes more critical.