Synthflow AI Review: Hands-On Testing and Insights

The future we dreamed about a decade ago is finally here, and tools like Synthflow AI are changing the game. This software promises “effortless, human-like AI phone calls,” and aims to transform how businesses manage customer interactions.

But is it truly as sophisticated as it sounds? Powered by ChatGPT, Synthflow AI boasts impressive strengths - but it also has its quirks. Setting up an AI agent feels straightforward, but getting it to work flawlessly can be a different story.

In this Syntflow AI review, I’ll take you through my hands-on experience. You’ll learn how I built my own AI agent, step by step, and discover the highs and lows of the process. By the end, you’ll have all the information you need to decide if Synthflow’s AI is the right fit for you.

Let’s dive in.

What Is Synthflow AI?

Synthflow AI is a platform that streamlines the creation and management of AI voice agents to handle live customer conversations. It offers friendly and efficient service anytime by making calls, understanding, and responding naturally like a human would.

Its no-code setup means anyone can create an agent, and its quick deployment makes integration with your existing systems simple.

Synthflow AI markets itself as your always-available, tireless, and highly capable team of virtual phone agents with a few key benefits:

• 24/7 Availability: Answer support calls 24/7 or reach out to leads from far-away time zones.
• Improved Customer Experience: With features like real-time booking and quick transfers to real humans, Synthflow AI resolves some of the annoying features people associate with AI chatbots.
• Boosted Sales Conversions: Sales opportunities are addressed immediately, which allows you to strike while the iron is hot.
• Cost Efficiency When Scaling: At $0.08-$0.13 per minute, Synthflow AI offers a cost-effective way to scale your calling capacity without breaking the bank.
• Easy No-Code Deployment: The platform's user-friendly interface allows you to deploy AI voice assistants without talking to your developers.
• Seamless Integration with Third-Party Tools: Synthflow AI integrates smoothly with CRM platforms and other tools. It can transfer important contact data and conversation transcripts across your marketing and sales stack so your whole team is on the same page.

Pssst! We also reviewed one of Synthflow's top competitors. Check out our Vapi AI Review and our comparison of Synthflow AI vs. Vapi

Getting Started With Synthflow AI

The core of Syntflow AI’s product is its agents, which are automated voice agents that can make or take calls.

Before you create your first assistant, I recommend setting up at least one phone number.

1. Setting Up a Phone Number

Phone numbers are essential for the Synthflow AI service. Without one, you can’t test, make, or receive calls. Also, keep in mind that each agent requires a unique phone number, so if you plan to build five agents, you need five phone numbers.

Currently, Synthflow AI supports numbers from the U.S., Canada, and Australia. Each phone number costs $1.50 per month in addition to your monthly subscription.

I found buying a phone number quick and straightforward. You simply select the number and pay for it:

Fortunately, you can connect local phone numbers from other countries through the Twilio integration. Just enter your Twilio Auth Token and SID, and the numbers will be populated in your Syntfhlow account, ready for use. Synthflow has also implemented support for other services like Telnyx, RingCentral, and even custom phones:

With your first virtual phone number purchased and set up, you're ready to create an agent.

2. Creating Your First Synthflow AI Agent

Regardless of which plan you're on, Synthflow AI allows you to create an unlimited number of agents. This feature is quite useful as you can create as many test agents as you need without worrying about plan limitations. However, you need to account for the costs of the additional phone numbers, as each agent requires a unique phone number.

You can build an agent from scratch, use a quick setup, or start with a predefined template.

As of this review, the platform provides only a limited selection of templates, but the list is growing. You can now access templates for:

• Recruiters
• Car dealers
• Mortgage consultants
• Recruitment
• Insurance
• Education
• Restaurants
• And more

You also get access to four types of inbound agents.

Starting with a pre-made template is a great way to understand how to structure the agent prompt (the instructions for your agent). Even if your business is in a different niche, I recommend you start with a pre-made assistant and adapt it to your industry.

For instance, if you’re selling software, you can take Jessica, below, who is made for real estate. With some changes to Jessica’s phone script and agenda, you can easily adapt it to your needs.

Inbound and Outbound Agents

Syntfhlow AI supports three main types of agents:

• Inbound Agents: These handle incoming calls from customers, such as support inquiries or general questions.
• Outbound Agents: These agents can initiate calls on your behalf, whether for cold outreach, follow-ups, or updates.
• Widgets: An embeddable chat widget for your website that enables real-time conversations and engagement directly with your visitors.

Now, let’s create a simple inbound assistant.

Configuring Your AI Agent

The initial set of options for your agent might look intimidating, but don't worry. Most of these features are easy to grasp.

Once you choose your agent's name and avatar, you can select the voice engine and AI model. Synthflow AI supports most of the latest AI models, such as GPT-5, GPT-4o, and GPT-4o-Mini. They also offer their own native model optimzed for conversations:

Synthflow AI doesn’t require you to set up the LLM model or pay extra for its subscription. The models are built into Synthflow and its pricing structure. How well the models work depends on how your prompt is written (more on that later).

You can further fine-tune the conversations by adding filter words, a custom vocabulary with words, and connecting your knowledge base and resources for the assistent to learn and use. You can also import PDF files, paste a website URL, or simply copy-paste your text into Synthflow’s interface.

Something that raises an alarm here is the lack of native integrations with help doc platforms like Zendesk and Helpscout, as well as the lack of an option to import a CSV database. I tried scanning a help docs homepage, but received a nasty error.

4. Giving Your Agent a Voice

Synthflow AI's out-of-the-box AI voices are powered by ElevenLabs, the leading speech synthesis software. Thanks to advanced deep learning technology, they offer natural-sounding speech and sound remarkably lifelike without the robotic tone often associated with AI. You can have a test conversation here.

I found the voice of my agents to sound almost perfect and, if not for the longer delays, almost indistinguishable from a real person.

That said, the longer pauses and the inability to interrupt the voice agent mid-sentence reveal that you're speaking to an AI bot rather than a real human. So, if flawless human experience is your top priority, Synthflow AI might not be the best fit for you at this time.

Synthflow AI supports English, Spanish, French, German, Portuguese, Italian, and Dutch, with additional multilingual support now including Hindi, Russian, and Japanese. Twenty-four other languages are in beta.

The platform allows you to select male and female voices, as well as a number of pre-set tones. You can also create a custom voice clone by uploading five audio samples. Since Synthflow AI uses ElevenLabs, you can upload one of the voice presets available on the ElevenLabs marketplace. For example, if you need an Australian accent, you can upload a preset with that specific voice.

I was impressed by the additional options available for fine-tuning the voice, though I found the default settings to work quite well overall. The only option I was tempted to experiment with was the latency of the responses to reduce the awkward pauses during conversation. However, it’s worth noting that faster response times can sometimes compromise the quality of the answers.

You can also configure:

• The pause before the agent speaks.
• How long the phone rings before the agent answers.
• The call duration limit (useful for when people forget to hang up the call, as you have a limited number of concurrent calls on your plan).

You can even enable your agent to use filler words to sound more natural.

5. Creating and Structuring Your Agent Prompt

Your agent’s prompt is the set of rules and instructions it will follow when talking to people. If you've used ChatGPT in the past, you'll feel at home. Even so, it’s best to start with one of the prompt templates as they are quite comprehensive. You can also use Synthflow’s “Prompt Co-Pilot” feature, a chatbot that hand-holds you through the process of building your prompt.

In your prompt, you want to mention things like:

• Agent tone and personality.
• The goal of the call and the agent.
• Conversational style.
• Company information.

Most importantly, you need to provide the most complete script instructions you can.

Below are the exact prompt structure and instructions I gave my sales assistant for a hypothetical project management cloud platform.

Introduction

This is where you explain who the agent is and provide its main objectives. Here’s what I wrote:

Tone and Behavior

This is where you instruct your agent on how to act. I provided the following:

Product Information

This is where you share how your product works. Here’s what I said about my hypothetical company:

Target Audience

This is the audience your bot will be speaking to. Here’s the detail I gave:

Value Proposition

Share the unique selling points (USPs) of your product or service. Here’s what I said (most companies will have more details to share):

Objection Handling Instructions

Write down the most common objections you usually hear on your phone calls. This helps train your AI agent on these specifically.

Objection: “I’m already using a project management tool.”

Objection: “I’m new to project management software.”

Objection: “I’d prefer a self-hosted solution.”

Script Instructions

This is a step-by-step script on how a typical call should go. The best part about this is that you can copy and paste your existing sales script, and your agent will follow it to the letter.

1. Initial Contact

2. Introduction and Initial Query

3. Highlighting Benefits

4. Organizing a Detailed Discussion

5. Confirming Phone Number

6. Closing and Follow-Up

One drawback I noticed with the prompt builder is that it was quite laggy and occasionally caused my browser to crash.

6. Testing Your AI Agent

Before you deploy your new agent, you'll want to test it. I found testing incredibly easy. You simply enter your phone number and click “Call me.”

The agent will give you a ring, and you can talk to them freely and test different questions and answers. Alternatively, you can call the number yourself and start a conversation.

On the free trial, you get 15 minutes of free testing.

Once the call is over, you get a transcript, a recording, a summary of actions performed, and an analysis of the call.

All calls can be later viewed in an organized table:

To test your agent through chat, click the “Chat” tab on the right sidebar and enter a prompt. The main benefit of chat testing is that it’s completely free but it’ll still show you how your agent responds in certain situations.

Synthflow AI uses voice APIs to convert these prompt responses into spoken dialogue. In other words, what you see in your chat prompts is exactly what people calling your bot will hear on the phone.

Since test calls also cost credits (between $0.08 and $0.13 per call minute, depending on your plan), it's recommended that you conduct your more extensive agent tests through text chat instead.

Here are some things you should consider testing. I’ll also show you how my AI agent, Jordan, handled my inquiries.

Testing Your AI Agent’s Introduction

This is where you test how your agent starts a call.

The agent’s introduction should differ depending on whether you have an outbound or inbound agent. Since Jordan is an outbound agent, I expected her to take the lead and start the conversation. However, that was not the case.

In the chat test, the agent was waiting for me to start the conversation:

Attempting a phone call test didn’t make any difference.

I tried giving her more explicit instructions like: “Make sure you start the conversation first; do not wait for the person to talk.” It made no difference. No matter how I fine-tuned the prompt, Jordan still acted like an inbound agent and didn’t initiate the call proactively.

Changing the LLM model also didn’t improve this.

I’m unsure if this was a bug in the testing process or a mistake in my settings, but the issue left me questioning the reliability of outbound agents.

It’s worth noting that when evaluating these prompts, we're essentially testing ChatGPT (or any other LLM being used). As a result, any issues with answer quality are more likely tied to the LLM model itself rather than to Synthflow AI.

Testing Your AI Agent with Off-Topic Conversations

The next big thing you want to test is how the agent handles off-topic conversations.

My first tests here failed as well:

I quickly found that you must be more explicit in your instructions to keep your agent in line. I added the following rule to the prompt:

This approach worked better, but I was disappointed that I had to be so precise with my prompts. Even then, some off-topic questions managed to slip through.

I expected Synthflow AI to include built-in rules for this within the pre-made templates. I know it’s possible because some of the walkthrough videos showed a bot handling off-topic questions much better than mine.

Testing Your Call Script

Last but not least, you want to test your on-topic script. My AI agent, Jordan, handled this great without issues and seamlessly arranged a follow-up call to book a demo with me.

It’s still important to remember that this is a ChatGPT-based bot, and avoiding unwanted minor deviations or hallucinations from the script is impossible. For instance, I was expecting Jordan to mention our case study with IBM when talking about cost savings, but it was omitted. Because these bots aren’t totally reliable, it’s best to leave the handling of sensitive data, such as credit card details, to a human or secure interface.

The good news is, Synthflow recenlty added a large-scale automated feature called Test Center. I'll explore it in-depth later in this review, but it should help speed up and improve agent testing.

7. Setting Up AI Agent Actions

Apart from actual calls, your agents can perform a number of handy additional actions:

• Send an SMS after the call to share follow-up information. Unfortunately, text messages don't currently support dynamic variables, so you're limited to a fixed text for all contacts.
• Book a call using Cal.com and GoHighLevel integrations.
• Transfer a call to a real person when the person requests it, or based on predefined characteristics.
• Extract call information to be displayed in the call summary. This is especially useful for outbound calls and lead qualification, where you need to skip through a large number of calls.
• Custom actions to connect with API and other advanced options.

I like the granularity of the action lists and the ability to see in which agents a specific action is used.

You can also now select an action from a list of templates. This makes setting up some of the more standard actions even easier.

8. Deploying Your Agent

It’s time to get your agent live! Synthflow AI offers a few options:

For inbound agents, the easiest way to deploy is to share your phone number with your customers and leads. When they call the number, the AI agent will answer.

For outbound agents, you can import a CSV file with numbers and select a specific outreach cadence and time zone for calling. This feature is called “Batch Campaigns.”

If you’d like to automate additional actions across other tools, GoHighLevel and Zapier offer more advanced deployment integrations. Synthflow also supports webhooks and Rest API.

So far, we’ve covered how to create, configure, and deploy your first agent in Synthflow AI.

In the next section, we’ll explore some of Synthflow AI’s standout features.

Key Synthflow AI Features

Book Meetings in Real Time

Syntflow AI’s agents can book meetings in your calendar during phone calls so that you capture every opportunity. Currently, the platform supports integrations with HubSpot, Cal.com, ClickFunnels, Make, Zapier, and GoHighLevel. But chances are great that by the time you are reading this, there are more integrations available.

In my conversations with Jordan, the agent was able to schedule a product demo call directly on the call instead of asking the person to leave a phone number and have a salesperson follow up to arrange a call. This saves one additional step in the sales process.

Transfer to Real Humans

Synthflow AI’s agent can forward a call to one or more phone numbers whenever necessary. You simply specify the trigger words, for instance, “I want to talk to the manager.” In my opinion, this is a must-have feature as some people simply don’t like to talk to robots. You can also ask the agent to share more information about the incoming transfer so the person handling the call can be more prepared.

When testing my agents, the transfer feature worked smoothly, easily identifying my inquiries for a transfer, regardless of how I worded them: “Transfer me to a human,” “I want to talk to the manager,” etc.

Enrich Leads With Information Extraction

One of the post-call actions available to agents is the ability to extract key information from the call. This can be used for lead qualification, customer research, or to gather more information on support issues.

You get the option to collect information in three different formats: Yes/no answers, single-choice answers, and open-question answers:

The answer to the question must be mapped to a unique identifier (think of it as a contact field). For instance, if you ask about the first name of the person in an open question, that identifier is “user_name.”

The platform also thoughtfully shows you what other agents, if any, use that identifier:

This feature is critical for outbound calls and salespeople. It allows them to prospect cold leads and collect valuable information that can be later used in follow-up calls. Information such as company size, pain points, and objections can be extracted from the call and recorded in a CRM like HubSpot. This allows salespeople to prepare for calls much faster than by sifting through dozens of call transcripts.

Advanced Custom Actions

With Synthflow AI, you can design custom post-call actions to fit your specific requirements, business processes, and workflows.

These actions can range from collecting precise data and integrating smoothly with your preferred tools to triggering advanced workflows. Keep in mind this is an advanced feature that will require your developer’s input.

Below is an example of a custom action that gets the product details from Hubspot, querying the database using the product name.

Assistant Teams

With Assistant Teams, you can create a whole fleet of AI agents that work together.

Teams are a great way to combine agents with different functions. For instance, you might have a lead qualification agent who refers qualified leads to a real human and unqualified leads to a second agent who can share automated information about your product or service.

You can specify when to transfer the person to another agent and how long to wait before the transfer (to simulate a real transfer):

Workflows

Synthflow AI offers a visual workflow builder that allows you to automate a set of actions, similar to how you would use Zapier. It also supports a broad range of actions and integrations such as native agents, third-party agents, FTP file uploads, and webhook triggers.

What surprised me was the sheer number of third-party agent implementations: I counted more than 60, from popular ones like Anthropic’s Claude to more obscure ones like Comfy ICU.

For example, let’s say you run a marketing agency and have Facebook ads promoting your content marketing service. When a lead enters their contact information through a Facebook ad, the Synthflow AI agent can make an instant call to qualify them. If the lead is qualified, the agent can add them to HubSpot CRM and create a task to follow up with them in a day. If not, the lead can be added to your email marketing tool for further lead nurturing.

On the native side of Synthflow agents, the workflows support an AI image generator, an MCP responder, a utility agent (that can classify and extract data), and even an AI video generator. For instance, you can automatically generate a personalized video for each client after every agent call. Workflows even support subflows, so you can automate your fleet of agents in a more manageable, organized way.

Frankly, I was expecting a fairly limited workflow builder centered around Synthflow's agents. I was proved me wrong. In a nutshell, we have a mini Zapier that’s deeply integrated with Synthflow’s native agents' capabilities. Good job, Synthflow!

This feature is only available on Pro plans and above, starting at $450/month.

Analytics Dashboard

Synthflow's relatively new analytics dashboard is great, offering several helpful indicators at a glance. You can get insights into your AI voice agents' performance, with key metrics like:

• Call volumes
• Success rates
• Sentiment analysis
• Call durations
• Reasons why the call ended

You can filter data by assistant, agent, call type (inbound/outbound), and custom time ranges to track trends and compare performance.

Personally, I find the sentiment analysis and failed calls most insightful, as I can grasp how well the agents are received and how often they fail.

Test Center

Test cases are a new, much-needed feature that signals Synthflow's ambition to serve enterprise customers by making its agents more reliable and production-ready.

Simulations let you test your voice agents in a safe environment before real customer calls happen. Instead of discovering issues live or manually, you create test scenarios where a simulated customer (persona agent) interacts with your voice agent. Synthflow then records the conversation, generates transcripts, and evaluates whether your agent met predefined success criteria.

Test cases take simulations to the next level by letting you run multiple scenarios at scale, up to 50 test cases in a single batch. You can create custom scenarios or auto-generate 5-10 cases from your agent's configuration, set clear success criteria, and track performance over time.

I wanted to try it out, so I decided to go with an AI-generated test case. I can choose between accuracy for responses, CSAT score, script compliance, and more. I’d definitely try script compliance, as this was something I struggled with in my manual tests, but for this exercise, let’s go with an angry customer, because it sounds fun.

These are the personalized test cases generated for one of my outbound agents, and I love it. I think it’s spot on both in terms of client reactions and the success criteria I outlined.

Now, with the test cases generated, I can run them and test my agent in a simulated environment:

Simulations take 3-10 minutes ... in theory. I did two test runs, and both got stuck at “Call in progress.” Maybe it was because I didn’t have enough credits, but I didn’t receive any error messages. Unfortunately, I couldn’t test the simulations in practice, which was disappointing!

Logs

The logs panel is another handy new utility feature. It is a list of all your agents’ calls, with a detailed view for each one. You can check the call outcome, its transcript, and more advanced features such as call analysis and any automation actions executed during or after the call. Overall, I find this very helpful.

Whitelabel Solution for Agencies

If you’re looking to start an AI-powered phone agency, Synthflow AI provides a flexible and fully customizable white-label option with advanced features to fit your needs.

Some of the white-label features include:

• Select which configuration options to display to your clients. For instance, you can hide the “prompt” section for your clients altogether and build your own prompts.
• Custom pricing structure.
• Branded dashboard with logo, name, and custom domain.
• Assign specific agents, teams, workflows, phone numbers, and more to each sub-account.

Overall, I was highly impressed by the attention to detail Synthflow AI has invested in its white-label feature.

Synthflow AI Pricing

Synthflow AI offers flexible pricing that scales with the number of call minutes. Overages cost $0.12 or $0.13, depending on your plan. The per-minute price can go as low as $0.08, but that’s only on the Enterprise custom pricing plan.

In recent months, Synthflow has moved up a market and entirely removed the accessible Starter plan, which started at just $29/month. This could be an attempt to target a more profitable market segment after the company's Series A round in June 2025. Whatever the reason, it leaves solo builders like me out in the cold and makes Vapi's usage-based model at $0.05/min suddenly look much more appealing.

The Pro plan doesn’t come with sub-accounts, so Growth is essentially the cheapest plan serving businesses that need to manage multiple accounts and clients.

There's also a $500/month add-on called "boosted queueing for calls", though both Synthflow's support documentation and marketing site are completely silent on what this feature actually includes. After reaching out to support and waiting two days, this is what I got from them: “The boosted call queueing feature enables higher call throughput and is recommended for users managing higher call volumes.” Not super helpful.

Final Verdict

The biggest challenge I faced with the Synthflow AI was getting the outbound agents to deliver a proper proactive introduction on the call and initiate the script without input from the receiver. The agent kept on being silent and acting like an inbound bot.

Another notable drawback was the effort required to create a robust prompt capable of handling off-topic questions. I wish the platform offered more comprehensive prompt templates with built-in rules for such scenarios. That said, if you're skilled at crafting prompts, you'll appreciate Synthflow AI's flexibility for designing your assistants.

The shift in their market positioning—starting at $450/month—feels like a move away from solo builders like me. Maybe it makes business sense for them, but it's well beyond what I can justify spending on a voice agent tool.

Since my audience communicates via email or chat and doesn’t rely on phone calls, Synthflow AI isn’t a fit for my needs. Also, I still prefer genuine human-to-human communication and would avoid AI voice assistants unless the scalability needs or budget constraints require it.

That said, if you’re in a more traditional industry - like car rentals, real estate, mortgages, or any field that handles a high volume of calls - Synthflow AI could be a valuable asset to your business.

If you're ready to embrace AI-driven communication, you should definitely give Synthflow AI a go. It's easy to deploy and offers the customization features needed to build a capable AI voice agent that can improve your processes and help you convert and retain more customers on autopilot.

Synthflow AI Alternatives

Vapi AI

Unlike the other tools on the list, Vapi AI is a solution geared toward developers (although it also provides a no-code dashboard). It's a voice AI API that allows you to build, test, and deploy voice assistants, as well as other text-to-voice solutions. Pricing starts at $0.05 per minute, which is significantly cheaper than the price of Synthflow AI. However, you must also pay additional costs for transcriptions, voice generation, LLM models (ChatGPT, for instance), and making the calls. This makes price calculations more complex but also flexible. That said, the customization options you have with Vapi AI allow you to bring its cost lower than Synthflow’s and let you optimize latency for faster response times. You can check out our head-to-head comparison of Synthflow AI vs. Vapi to learn more about how they compare.

Califluent

The first noticeable difference between Synthflow AI and Califluent is that Califluent supports 30+ languages, including Japanese, Indian, and Russian. Another benefit of Califluent is that apart from SMS messages, you can also send emails during or after the call. The biggest drawback of Callfluent is that it doesn’t support a white-label solution, so if you're an agency looking to set up AI call services, you would be better off with Synthflow.

Bland AI

The main difference between Synthflow AI and Bland AI is that with Bland AI you cannot upload your own custom voices or ElevenLab presets. A benefit of Bland AI is that you can purchase a custom number of credits for your calls vs. the predefined number of minutes that come with Syntfhlow plans.