Industry insights
A first-person evaluation of seven Dialogflow alternatives for teams that need production AI voice agents with native CRM integration, multi-channel follow-up, and voice pipeline management — ranked by migration fit.
Last updated
I evaluated seven platforms that teams most commonly consider when migrating away from Google Dialogflow for AI voiceAI voiceAn artificially generated, natural-sounding voice produced by a TTS model. Thoughtly supports a library of AI voices and brand-specific cloning. agent use cases. Dialogflow launched as a capable NLU framework, but as conversational AI has shifted toward autonomous, multi-channel voice agents that qualify leads, book appointments, and write back to CRMCRMThe system of record for leads, contacts, deals, and activity. Thoughtly reads from and writes to your CRM continuously. records in real time, many teams have outgrown what Dialogflow was designed to do.
This is not a customer-support chatbot comparison. The platforms below were chosen specifically for teams that need production voice agents handling live phone calls — qualifying interest, executing follow-up sequences across SMS and email, and integrating directly into revenue workflows. If your Dialogflow implementation is mostly a text chatbot, some of these platforms will still apply, but the evaluation criteria are weighted toward voice-first use cases and revenue operations.
Dialogflow is still a viable NLU layer for simple chatbot use cases, but several recurring limitations push teams toward alternatives — especially when the use case involves live phone calls and revenue workflows.
Each platform was assessed across six criteria chosen specifically for teams replacing Dialogflow in voice-first, revenue-critical deployments. These are not abstract feature checklists — they reflect the operational gaps that cause Dialogflow migrations in the first place.
How easy is it to build, test, and iterate on conversation flows compared to Dialogflow CX? I looked for platforms where a RevOps or product manager can modify qualification logic, add new conversation branches, or update CRM mappings without writing fulfillment code. Dialogflow CX requires developer involvement for most non-trivial changes, so the bar here is whether the alternative gives business teams direct access to the conversation. Platforms with visual builders, drag-and-drop editors, or plain-language agent configuration scored highest.
Does the platform handle the full voice pipeline — telephony, speech-to-text, dialogue, text-to-speech, call routing, warm transfer — natively, or does it require stitching together multiple services like Dialogflow does? I evaluated latency benchmarks (sub-500ms is the floor for natural conversation), voice quality, interruption handling, and whether the platform can place and receive calls without third-party telephony middleware. Platforms that handle voice end-to-end without custom integration code scored highest.
How does the platform handle natural language understanding relative to Dialogflow? This matters because teams migrating off Dialogflow have already invested in intent taxonomies, entity definitions, and training data. I looked at whether the alternative uses modern LLM-based NLU (which reduces or eliminates manual intent engineering), supports hybrid intent+LLMLarge Language Model (LLM)A machine-learning model trained on massive text data, used as the reasoning engine that drives a voice agent's understanding and responses. approaches, or still requires the same manual intent/entity management that made Dialogflow cumbersome. Fewer manual intents with better accuracy is the goal.
Can the platform read from and write to CRM systems, scheduling tools, and workflowWorkflowAn automated, multi-step process — usually triggered by an event (form fill, new lead) and orchestrating one or more voice / SMS / email actions. automation platforms natively? Dialogflow relies on webhook fulfillment for every integration, which means maintaining custom Cloud Functions for Salesforce, HubSpot, Calendly, and every other system in the stack. I tested whether each alternative offers pre-built CRM connectors, native scheduling, and event-driven workflow triggers that eliminate the fulfillment-code bottleneck. Platforms with 20+ native integrations and two-way CRM sync scored highest.
How quickly can a team go from a working prototype to a production voice agentVoice agentAn autonomous, conversational interface that interacts with humans over the phone — answering, qualifying, and routing calls without human staffing. handling real calls? Dialogflow CX deployments often take 8-16 weeks because of the integration assembly required. I evaluated time-to-first-call, the testing/staging infrastructure, and whether the platform supports gradual rollout, A/B testing, and production monitoring without building custom tooling around it. Platforms where a team can ship a production agent in under a week scored highest.
Is the pricing model predictable as voice call volume grows? Dialogflow CX charges per session with separate audio and telephony costs that compound quickly. I compared per-minute pricing, per-session pricing, seat-based models, and enterprise custom quotes — looking specifically for whether a team can forecast monthly costs at 1,000 and 10,000 calls without needing a sales call to get the answer.
| Platform | Best for | Voice native? | Pricing model | Key limitation |
|---|---|---|---|---|
| Thoughtly | Revenue teams converting inbound leads across voice, SMS, and email | Yes — full pipeline | Per minute | Best when GTM workflow is clearly defined |
| Rasa | Engineering teams wanting full NLU control and self-hosting | Via integrations | Open source + enterprise tiers | Requires ML engineering to deploy and maintain |
| Amazon Lex | AWS-native teams with existing Connect/Lambda infrastructure | Via Amazon Connect | Per request + audio | Limited outside AWS; voice requires Connect setup |
| Voiceflow | Conversation designers building multi-channel bots with visual tools | Via integrations | Free tier + per-seat plans | Voice requires third-party telephony integration |
| Kore.ai | Enterprise IT deploying virtual assistants across business units | Yes — via SmartAssist | Enterprise custom | Complex setup; long implementation cycles |
| Cognigy | Multi-language contact centers with deep telephony needs | Yes — via Voice Gateway | Enterprise custom | Contact-center focused; less suited for lean RevOps teams |
| Microsoft Copilot Studio | Microsoft ecosystem teams running Dynamics 365 and Teams | Via Azure | Per-session + capacity packs | Tightly coupled to Microsoft stack; limited outside it |
Thoughtly is a voice AI platform built for autonomous inbound lead conversion. Where Dialogflow gives you an NLU engine and expects you to assemble the rest, Thoughtly delivers the complete pipeline: AI agents that answer calls, qualify leads, follow up by SMS and email, book meetings, and write outcomes to your CRM — all without fulfillment code. The platform supports 34 languages natively, delivers sub-350ms response latency, and ships with 24+ certified integrations including Salesforce, HubSpot, Pipedrive, Calendly, and Zapier.
For teams migrating from Dialogflow, the most relevant difference is operational ownership. Dialogflow requires developers to maintain intent graphs, fulfillment webhooks, telephony integration, and STTSpeech-to-Text (STT)The system that turns the caller's speech into text the agent can reason over./TTSText-to-Speech (TTS)The system that turns the agent's generated text into spoken audio — the voice the caller actually hears. configuration separately. Thoughtly collapses that stack into a single platform where RevOps and GTM managers can build, test, and deploy voice agents directly — including the conversation logic, the CRM write-back, and the multi-channel follow-up sequence.
Revenue operations and GTM teams at mid-market or enterprise companies in high-consideration industries — insurance, mortgage, real estate, automotive, education enrollment, elective healthcare, home services, and financial services — that have outgrown Dialogflow and need a production platform that handles the full lead conversion lifecycle without assembling infrastructure.
Per-minute pricing. Contact Thoughtly for a quote based on volume and channels.
Rasa is the most established open-source conversational AI framework and the most common destination for engineering teams leaving Dialogflow. The platform gives developers full control over the NLU pipeline, dialogue management, and deployment infrastructure — including self-hosting. Rasa extends LLMs with custom business logic to create reliable AI agents across millions of conversations, and customers include Autodesk, BNP Paribas, Swisscom, and Providence Health.
The key difference from Dialogflow is ownership. Where Dialogflow locks you into Google Cloud and a proprietary intent format, Rasa lets you train custom NLU models, run them on your own infrastructure, and retain full data sovereignty. The trade-off is that Rasa requires ML engineering skill to configure, train, evaluate, and maintain the models — it is a developer platform, not a turnkey business tool.
Engineering-led teams at mid-market or enterprise companies that need full control over the NLU stack, data sovereignty, and self-hosting — typically in regulated industries like healthcare and financial services where cloud-hosted third-party NLU raises compliance concerns.
Open-source (Rasa Open Source is free). Rasa Pro and Rasa Enterprise are priced per deployment — contact Rasa for a quote.
Amazon Lex is AWS's conversational AI service and the most direct cloud-provider alternative to Dialogflow. It uses the same deep learning technology that powers Alexa and integrates natively with Lambda for fulfillment, Amazon Connect for telephony, and the broader AWS ecosystem. If your team already operates on AWS and wants to avoid Google Cloud dependencies, Lex is the closest architectural equivalent.
The migration path from Dialogflow to Lex is conceptually familiar — you are still defining intents, slots, and fulfillment functions — but the AWS service mesh replaces Google Cloud Functions and Dialogflow-specific APIs. The voice channel runs through Amazon Connect, which provides a full contact center stack. However, Lex shares many of Dialogflow's structural limitations: manual intent engineering, separate fulfillment code, and no built-in CRM or follow-up execution.
Engineering and IT teams at companies with deep AWS investments — especially those already using Amazon Connect for telephony — that need a cloud-native conversational AI layer without adding a Google Cloud dependency.
Per-request: $0.004 per speech request and $0.00075 per text request (with a free tier for the first year). Amazon Connect telephony is billed separately per minute.
Voiceflow is a visual agent-building platform that has become one of the most popular Dialogflow migration targets for teams that want a drag-and-drop conversation designer without giving up programmatic control. The platform earned a 2026 Best Software Award from G2 for Agentic AI Products and is used by design, product, and engineering teams to build AI agents for customer support, internal helpdesk, and conversational commerce use cases.
For Dialogflow migrants, Voiceflow's biggest draw is the visual canvas: you can see and edit the entire conversation graph in a single workspace, connect to external APIs via built-in API steps, and prototype on the same platform where you deploy. The trade-off is that Voiceflow does not include native telephony — voice deployments require integrating a third-party telephony provider and STT/TTS services, which adds back some of the assembly complexity that drove the migration in the first place.
Product and design teams building conversational AI agents primarily for web chat, messaging, and customer support channels — especially teams that valued Dialogflow's intent framework but need a more visual, collaborative design environment. Less ideal for teams whose primary use case is live phone calls.
Free tier available. Pro and Team plans are per-seat. Enterprise pricing is custom.
Kore.ai is an enterprise virtual assistant platform that competes directly with Dialogflow CX at the large-organization tier. The platform offers pre-built industry templates for banking, healthcare, retail, and insurance, along with a visual dialog builder, advanced NLU with hybrid intent+LLM support, and a full contact center automation module called SmartAssist. Kore.ai has been recognized as a Leader in the Gartner Magic Quadrant for Enterprise Conversational AI Platforms.
The migration appeal for Dialogflow teams is the pre-built vertical templates: rather than rebuilding intents from scratch, teams can start with industry-specific conversation frameworks and customize from there. SmartAssist adds native voice capabilities including call routing, IVR replacement, and agent desktop — features that Dialogflow requires you to build with separate Google CCAI and telephony integrations.
Enterprise IT and digital transformation teams deploying virtual assistants across multiple business units or customer service channels, particularly in regulated industries where pre-built compliance templates and enterprise security controls are non-negotiable.
Enterprise custom pricing. Contact Kore.ai for a quote. No public self-serve tier.
Cognigy is a German-headquartered enterprise conversational AI platform built for contact center automation at global scale. The platform supports 100+ languages natively, offers a visual flow editor, and includes Cognigy Voice Gateway for direct telephony integration without third-party middleware. Cognigy has strong traction in European enterprises and global contact center operations where multi-language support is not optional — it is the primary requirement.
For Dialogflow teams that built multi-language bots using Dialogflow's language variants, Cognigy offers a significantly more mature multi-language architecture with per-language NLU models, locale-aware conversation management, and region-specific telephony routing. The voice gateway handles SIP trunking, PSTN connectivity, and call transfer natively — replacing the Google CCAI + third-party telephony stack that Dialogflow voice deployments typically require.
Global enterprises with multi-language contact center operations — particularly European companies or organizations with contact centers serving 10+ languages — that need native voice with telephony integration and strong data residency controls.
Enterprise custom pricing. Contact Cognigy for a quote. Typical deployments include professional services.
Microsoft Copilot Studio (formerly Power Virtual Agents) is Microsoft's low-code conversational AI platform built into the Power Platform ecosystem. For teams running Dynamics 365, Teams, and Azure, Copilot Studio offers the tightest integration path: agents can read CRM records from Dynamics, trigger Power Automate workflows, and deploy directly into Teams channels without third-party connectors.
The Dialogflow migration case for Copilot Studio is strongest when the organization is already a Microsoft shop. Rather than maintaining Google Cloud + Dialogflow alongside a Microsoft-primary stack, teams consolidate on a single ecosystem. The platform now includes generative AI capabilities, knowledge base grounding, and plugin extensibility — bringing it closer to modern agent architectures than the original Power Virtual Agents.
IT and business operations teams at Microsoft-first organizations running Dynamics 365, Teams, and Azure — especially those that want to consolidate conversational AI into their existing Power Platform governance and licensing without adding a Google Cloud dependency.
Per-session pricing starting at $200/month for 25,000 sessions (Microsoft 365 Copilot Studio plan). Additional capacity packs available. Enterprise pricing varies by licensing agreement.
The right alternative depends on what made Dialogflow fail for your team in the first place. Here is the decision map:
Dialogflow ES (the original Essentials edition) has been in maintenance mode since Google launched Dialogflow CX. Google continues to invest in Dialogflow CX and its newer Vertex AI Agent Builder, but the direction is clear: Google is shifting conversational AI development toward Vertex AI and away from the standalone Dialogflow interface. Teams starting new projects should evaluate whether the long-term Google Cloud roadmap aligns with their needs.
It depends on the target platform. Some platforms like Rasa and Kore.ai offer intent import tools or migration guides for Dialogflow intent/entity definitions. However, most teams find that migrating to a modern LLM-based platform means rethinking their intent architecture entirely — which is often the point. Platforms like Thoughtly eliminate manual intent engineering altogether by using LLM-based conversation understanding.
Amazon Lex has the lowest entry cost with per-request pricing and a free tier for the first year. Voiceflow offers a free tier for prototyping. However, adding voice telephony to either platform requires additional services and costs. For teams that need production voice agents with minimal assembly, Thoughtly's per-minute pricing is often more predictable than assembling separate services with per-request, per-minute, and per-session charges.
In most cases, yes — but that is usually an advantage rather than a cost. Dialogflow's intent/entity/flow model is a different paradigm from modern LLM-based agents. Teams that try to port Dialogflow flows 1:1 into a new platform typically end up with the same brittleness they were trying to escape. Start with the use case and the desired outcomes, then design the new agent from those — not from the old intent graph.
Picking a platform that solves only one of Dialogflow's limitations. For example, migrating to another flow-based NLU engine (like Amazon Lex) fixes the cloud lock-in problem but preserves the manual intent engineering burden. Migrating to a visual builder (like Voiceflow) improves the design experience but may not solve the voice pipeline assembly problem. The safest migrations happen when teams clearly identify which Dialogflow limitations are causing the most operational pain and choose an alternative that addresses all of them — not just the most visible one.
