Is Dialogflow CX, Azure AI, or Bedrock Right for You?
By Christi AkinwumiSeptember 28, 2026In this article
A CX leader evaluating agent platforms in September 2026 runs into three finalists almost every time: Google’s Dialogflow CX, Microsoft’s Azure AI Foundry Agent Service, and Amazon’s Bedrock AgentCore. Vendor pages compare feature counts. None of them answer the question a CX team actually has to decide: where does the work of building and running this agent land on our team, and what happens the day it needs to hand a customer to a person.
ICX builds on all three platforms. This post is a decision table built from that build experience, not a feature review, plus the naming confusion a buyer needs to see through first.
Why does the naming keep changing?
Before comparing the platforms, it helps to know that two of the three names in this post’s title are already out of date, which is itself a data point about platform risk.
Google renamed Dialogflow CX to Conversational Agents and merged its console with the older Vertex AI Agent Builder console into a single interface. Existing agents migrated automatically, a change documented in Google’s own Dialogflow change history. Then, in April 2026, Vertex AI Agent Builder itself was renamed the Gemini Enterprise Agent Platform. A buyer searching “Dialogflow CX” today lands on pages that already use both newer names, and a job posting written a year ago may reference a product that no longer carries that name.
Amazon went further than a rename. The original Bedrock Agents feature, launched in November 2023, was rebranded Bedrock Agents Classic and stopped accepting new customers on July 30, 2026. Amazon Bedrock AgentCore is the replacement, and it is a different kind of product: a managed runtime, identity layer, and observability layer, rather than a single agent-building feature. AgentCore added a managed consent portal for OAuth flows in September 2026, removing infrastructure teams used to build by hand.
Microsoft’s Foundry Agent Service is the youngest of the three names, following the 2025 to 2026 rebrand of Azure AI Studio into Microsoft Foundry. Its Build 2026 edition added a stable orchestration framework and general availability for its VS Code toolkit, while features like Toolboxes and multi-type memory remain in public preview.
None of this means any platform is unstable. It means a comparison written a year ago is already partly wrong, and a team should confirm current names before a procurement conversation, not after.
The five-question platform framework
ICX uses five questions to compare agent platforms for CX work, because they are the decisions a team actually has to make, not the features a vendor chooses to highlight.
| Question | Dialogflow CX (Conversational Agents) | Azure AI Foundry Agent Service | Bedrock AgentCore |
|---|---|---|---|
| Where does the conversation logic live? | In a visual flow builder: pages, intents, and routes, with generative fallback available inside a flow | In code, through the Microsoft Agent Framework, called via a single Responses API | In code, through your own agent framework; AgentCore runs it, it does not define it |
| How is escalation to a human built? | Native transfer-to-human step inside the flow, built for contact center handoff | A custom tool call, typically wired to Teams, Dynamics, or Omnichannel | A custom tool call the team writes; no built-in contact center handoff |
| How are guardrails and evals attached? | Design-time flow constraints, plus Vertex safety filters on generative fallback | Native evaluation and tracing inside Foundry, plus Azure AI Content Safety | Attached via Bedrock Guardrails and AgentCore’s identity and observability components |
| What must the team already run? | A Google Cloud project, often paired with an existing contact center integration | Microsoft 365 and Teams investment, for the agent to publish anywhere useful | AWS infrastructure and a team comfortable building the harness themselves |
| How fast has the product renamed? | Twice in under two years: to Conversational Agents, then folded into the Gemini Enterprise Agent Platform | Once, from Azure AI Studio, with several features still in public preview | Once, with the original feature retired to “Classic” and closed to new customers |
Read the table by row, not by column. A platform that wins on one question can lose the deal on another, and the answer that matters most depends on what the team is short on: conversation design time, engineering time, or an existing platform investment.
What ICX has observed building on all three
The build experience differs less in what each platform can technically do and more in what it hands you by default versus what it expects you to assemble.
Dialogflow’s flow builder gets a working demo in front of a stakeholder fastest, but the guardrail is the flow design itself. A rule enforced by a flow node is not a rule you can point to and test independently; changing one page can quietly change what the agent is allowed to say two steps later. Teams that add generative fallback on top of an existing flow need to re-test every node the fallback can reach, not just the new one.
Building on Bedrock AgentCore means writing the escalation path yourself from the start, because the runtime has no opinion about contact centers. ICX treats that as a feature disguised as a gap: a team that writes its own handoff logic on day one tends to design a better one than a team that inherits a default and only touches it when something breaks. That matches ICX’s read on harness engineering, where the platforms with the least built-in structure often produce the most deliberate one, because nothing gets skipped by default.
Azure’s Foundry Agent Service folds cleanly into Teams and Microsoft 365 once an agent is built, which matters for internal service desks more than customer-facing support. ICX treats its newer pieces, the expanded memory types and the Toolboxes feature, as preview-stage rather than production dependencies until Microsoft moves them to general availability. A platform in public preview can still power a live agent; it should not be the only path to an escalation decision.
None of these observations replace a pilot. They are the questions ICX asks before recommending one, drawn from work across all three rather than a single vendor’s documentation.
Which platform fits which team?
Start from what the team already runs, not which platform looks newest in a demo.
A team already on Google Cloud, especially one with an existing contact center integration, gets the most value from Conversational Agents, because the native handoff and flow builder do work the other two make the team build. A team already inside Microsoft 365 and Teams gets the most value from Foundry Agent Service, because the distribution is already solved. A team on AWS building custom agent logic, or one that wants full control over the harness rather than inheriting one, gets the most value from AgentCore.
In every case, the platform is not the hard part. ICX’s measurement framework for agentic AI and the production gap in agent pilots both point to the same conclusion: platforms rarely fail a pilot. Conversation design, escalation rules, and evaluation do. A platform choice made without those in place will surface the same problems regardless of which of the three a team picks.
Buyers researching this decision are also asking AI search engines who to trust on it. ICX’s own AI Citation Index found that comparison and best-practices pages earn far more of these citations than short opinion posts, which is part of why this post is a table and a framework rather than a verdict.
Where to start
Pick the platform that matches what the team already runs, then spend the saved time on the conversation design and escalation rules that determine whether customers trust the result, the same discipline behind ICX’s guide to enterprise chatbot design. If a pilot is stalling on the platform question rather than the design question, that is usually a sign the wrong question is being asked first.
If your team wants a second opinion on which of these three fits your build, see ICX’s services or get in touch. ICX would love to hear what you’re weighing.
ICX is building something new for readers who want these comparisons as they update: a newsletter is coming soon, so bookmark the blog and check back.
How ICX verified this post
Every product name, rename date, and feature status in this post was checked against vendor documentation on September 27, 2026, linked inline. The Dialogflow rename and console consolidation are documented in Google Cloud’s own change history. The Bedrock Agents Classic retirement date and the AgentCore identity feature are documented in AWS’s own developer guide and blog. The Foundry Agent Service feature list reflects Microsoft’s Build 2026 announcement; several named features remain in public preview as stated.
ICX drafts with AI assistance and reviews every post before it goes live. Read the full policy on the AI disclaimer page.
Sources
Frequently asked questions
What's the actual difference between Dialogflow CX, Azure AI Foundry, and Bedrock AgentCore?
Dialogflow CX gives you a visual flow builder where conversation logic lives in pages and intents. Azure AI Foundry Agent Service and Bedrock AgentCore are both code-first: you write the orchestration in a framework, and the platform provides the managed runtime, memory, and identity layer around it. Dialogflow is the only one of the three built primarily for conversation design rather than general agent infrastructure.
Does Dialogflow CX still exist, or was it renamed?
The product still exists but the name is being phased out. Google renamed Dialogflow CX to Conversational Agents and merged its console with the old Vertex AI Agent Builder console. Vertex AI Agent Builder itself was renamed the Gemini Enterprise Agent Platform in April 2026. Existing agents migrated automatically, but a team searching for 'Dialogflow CX' today will land on documentation that already uses the newer names.
What happened to Amazon Bedrock Agents?
The original Bedrock Agents feature, launched in November 2023, was rebranded Bedrock Agents Classic and stopped accepting new customers on July 30, 2026. Amazon Bedrock AgentCore is now the primary managed runtime for building and running agents on AWS. Teams starting new work should build on AgentCore, not the Classic feature.
Which platform has built-in escalation to a human?
Dialogflow CX is the only one with a native transfer-to-human step inside the flow builder itself, because it was designed as a contact center product first. Azure AI Foundry Agent Service and Bedrock AgentCore both treat escalation as something the team builds as a tool call or a custom integration, since neither comes as a turnkey contact center flow.
Which platform is easiest for a non-technical team to start with?
Dialogflow CX, because the flow builder does not require writing orchestration code. Azure AI Foundry Agent Service and Bedrock AgentCore both assume a developer is assembling the agent, even though each has lowered the amount of custom code needed over 2026.
How should a CX team actually choose between these three?
Start with what the team already runs, not which platform looks newest. A team on Google Cloud with an existing contact center integration gets more value from Conversational Agents. A team already inside Microsoft 365 and Teams gets more value from Foundry Agent Service. A team on AWS building custom agent logic gets more value from AgentCore. Conversation design and escalation rules matter more than the platform choice in every case.
Get an email when we publish a new post.