Insighto.ai

What is Insighto.ai?

Insighto.ai is a code-free AI voice and chat agent platform designed for businesses and agencies. Teams can use existing customer data, FAQs, and knowledge bases to create 24/7 phone assistants, website bots, and customer service agents for messaging channels.

These agents can answer questions, filter leads, make appointments, conduct outbound calls, send messages, run various tools, and transfer calls to human agents when necessary. The platform also offers custom domain names, sub-accounts, and the ability to charge clients separately, making it suitable for service providers to use.

A one-sentence summary

Insighto.ai brings together knowledge base Q&A, voice calls, chat components, multi-channel messaging, business workflows, and agent whitelisting under a single AI agent platform.

Main functions

Code-free AI assistant

Users can create assistants in the backend, write instructions, select models, and connect one or more knowledge bases. After testing is completed, the assistant can be linked to phone apps, web pages, or messaging components.

Phone AI agent

The phone assistant supports natural voice conversations, handling of incoming calls, outbound calling tasks, detection of voicemails, and recording of calls. It can determine, based on instructions, when to use a call transfer tool in order to connect the customer with a human agent.

Knowledge base chatbot

Chat assistants can answer questions based on websites, PDFs, Word documents, help centers, and text snippets, and they can also display citation markers. Multiple knowledge bases can be linked to the same assistant at the same time.

Multi-channel messaging

The platform supports various communication channels such as website chats, phone calls, text messages, emails, and WhatsApp, and can be expanded through additional components and integrations. WhatsApp makes use of the Meta Cloud API directly, rather than the WhatsApp interface provided by Twilio.

Business tools and workflows

Assistants can utilize SMS, calendars, APIs, Webhooks, and other tools to transform conversations into appointments, notifications, data synchronization, or business actions. Steps that fail within a workflow are not automatically retried; it is necessary to implement idempotency and compensation mechanisms manually during the design process.

Manual transfer

Phone conversations can be transferred to a human operator at any point, and chats can also be handled by live agents. When upgrading, it is necessary to provide a summary of the conversation, the user’s details, and the actions that have been taken, so as to avoid the customer having to repeat information.

Voice libraries and voice cloning

The voice library includes Azure Speech, Cartesia, ElevenLabs, PlayHT, as well as custom cloned voices; it is possible to filter and listen to these voices by language, provider, and gender. Custom cloning requires access to the appropriate subscription plan and high-quality, authorized audio files.

Session and usage analysis

The management backend stores chat records, phone call recordings, transcriptions, information on tool usage, and billing details, which facilitates the assessment of quality and costs. The team should establish access permissions, retention periods, and rules regarding notification for recordings.

Which types of assistants are supported?

Knowledge base and response control

The knowledge base is the foundation for ensuring the accuracy of responses; the information contained within it must be accurate, accessible, clearly labeled as to its version, and continuously updated. Enabling citations allows visitors to view the corporate materials related to a particular response, but it does not guarantee that the model fully understands the content.

Voice processing chain

Traditional phone assistants combine speech recognition, language models, and speech synthesis separately, with the platform handling the scheduling of calls. The real-time OpenAI assistant integrates listening, thinking, and speaking into a single sequential process, resulting in lower latency; however, its cost is higher and the available voice options are limited.

Model selection

The cost of models is calculated on a multiplicative basis; it is generally advisable to start with the model that offers the lowest cost while still being capable of completing the task. More advanced models should only be used when actual testing shows that the basic model is not sufficient.

In the current model dropdown list, the main options are OpenAI, Azure OpenAI, and Groq; the documentation specifies that there is no Claude option available for now. The available models may change, so it is necessary to check them in the account before deployment.

Steps to create a chat agent

1. Register an account and define the objectives for customer service, sales, scheduling, or internal Q&A.
2. Create a knowledge base by importing websites, documents, FAQs, and the data that can be used.
3. Create a new Chat assistant and define the roles, tasks, response boundaries, and conditions for transferring to a human agent.
4. Select a model and compare the quality and cost of basic and advanced models in relation to the actual problem at hand.
5. Connect to forms, SMS, calendars, CRM, Webhooks, or custom tools.
6. Configure the colors, brand, welcome message, quotes, and privacy notices for the chat component.
7. Tested error data, unauthorized requests, prompt injection, identity confusion, and tool failures.
8. Publish to a website or messaging channel, and continuously monitor session details and cost breakdowns.

Steps to create a phone agent

1. Clarify the processes for incoming calls, outbound calls, appointments, reminders, or lead filtering.
2. Create a Phone or Realtime assistant and select the working language.
3. Configure speech recognition, language models, speech synthesis, and voice characters.
4. Set the welcome message, interruptions, silence period, voice mail, and termination conditions.
5. Connects phone numbers, calendars, CRM systems, text messages, and call forwarding tools.
6. Tests accent, background noise, numbers, names, interruptions, and line latency.
7. Explain AI, recording, and data processing to the caller, and comply with local consent rules.
8. Launch in small batches, and monitor calls connected, transfers, reservations, errors, as well as the cost per minute.

Supported business channels

Voice and language

Insighto’s annotation feature supports over 50 languages; ordinary phone assistants can be configured to use multiple working languages, with automatic detection of these languages. The actual quality depends on the speech recognition provider, the language models, the voice databases, the communication lines, and local accents.

Requirements for voice cloning

For custom cloning, ElevenLabs is used; the audio required has a duration of at least 30 seconds, a sample rate of 16kHz, and must be monophonic with no background music. Companies must obtain explicit consent from the speaker, and impersonation, fraud, or unauthorized replication of voices are prohibited.

Business tools and integrations

• Calendar appointments and the GHL Calendar tool;
• SMS notifications and customer follow-up;
• Phone transfer and human takeover;
• Webhooks and custom API actions;
• WhatsApp Business, web page, and message components;
• GHL contacts and business automation;
• Forms, intents, and collaborative agents;
• Outbound campaigns, call records, and session analysis.

Workflow reliability

Failed workflow steps are not automatically retried, nor are Webhooks sent again if they fail. The receiving system should queue requests quickly; business actions require the use of idempotent keys, status checks, and manual compensation mechanisms.

Which users is it suitable for?

• Local businesses that require round-the-clock telephone support, appointment scheduling, and lead screening;
• Small and medium-sized companies that wish to offer both telephone and website-based AI customer service;
• Marketing teams that use GHL to manage clients and calendars;
• Multinational companies that need multilingual call and messaging services;
• Digital marketing agencies that build and manage AI agents on behalf of clients;
• Service providers who wish to use white-label domain names, sub-accounts, and refill services;
• Development teams that need APIs, Webhooks, BYOK, and custom tools.

In what situations is it not very suitable?

• Individual users who require permanent free access with no limits on the number of calls or messages sent;
• Organizations that require all components to be open source and deployed in a completely offline environment;
• Critical financial processes that require native Webhook signing and automatic retry.
• Organizations with high security requirements that do not currently offer two-factor authentication cannot be accepted;
• Outbound calling activities that lack legitimate consent for recording and permission for telemarketing;
• It is hoped that AI can handle independently scenarios involving medical, legal, financial matters, or urgent decisions.

Product advantages

• Telephones, web chats, and messaging platforms can share the same AI agents and knowledge base;
• Code-free configuration reduces the barriers to getting started for ordinary business teams;
• Models, speech recognition, and speech synthesis can be combined based on quality and cost;
• Multiple languages, voice mail, call forwarding, and recording are suitable for telephone reception;
• BYOK allows companies to keep certain models and voice billing details in their own accounts;
• The package plus wallet option facilitates continued service once the limit is exceeded;
• The agent plans to offer white-labeling, sub-accounts, customer domain names, and rebilling;
• APIs, workflows, and tools enable connectivity to business systems.

Usage restrictions

• The current account does not have two-factor authentication, so the administrator should implement additional security measures;
• To delete an account, you need to contact support – it cannot be done directly through the backend.
• Webhooks lack HMAC signatures, and they do not attempt to retry in the event of a failure;
• Steps that fail in a workflow are not automatically retried, and compensation logic is required for critical actions.
• API keys are authenticated using query parameters, and it is necessary to prevent them from being written to logs or analysis systems;
• The interface for automatic wallet top-ups is not yet available; top-ups usually have to be done manually.
• The total cost of models, voice services, telecommunications, and messaging can be higher than the advertised basic price.
• Calls, text messages, and WhatsApp are subject to regional, carrier, and platform rules.

Price of individual packages

Insighto offers two billing options: monthly plans and a prepaid wallet, with the quota for these plans being reset at the end of each billing cycle. Once the limits on queries, voice usage time, interactions with robots, or storage space are exceeded, charges are deducted from the wallet based on the actual amount used. New accounts receive a $10 trial balance.

The package page and the user documentation present the details of the included quota in different ways; before making a purchase, it is necessary to check on the settlement page the number of seconds for voice usage, the number of text queries that can be performed, the amount of storage available, as well as the status of various robots and functions. Discounts applied to annual subscriptions from previous years cannot replace the current monthly pricing.

Agent and white-label prices

White-label solutions allow the use of one’s own applications, interfaces, and chat domain names. However, once the DNS settings are saved, they cannot be changed manually in the background; external assistance is required. Before final configuration, it is necessary to carefully verify the domain names, certificates, branding elements, as well as the plan for customer migration.

Pay-as-you-go details

In principle, no refunds are provided for any subscriptions or add-ons; a downgrade takes effect during the next renewal. The wallet does not offer an automatic top-up feature for regular users, and insufficient balance may affect the continuity of services; companies should implement external monitoring systems.

How does BYOK work?

BYOK allows users to connect their own OpenAI, ElevenLabs, and Azure Speech keys; the costs associated with those models or services are charged directly to the user’s account. This feature is subject to the settings of the subscription plan, and telecommunications providers, certain voice service providers, and platform orchestration services may still charge additional fees.

Keys are stored in an encrypted form and are not included in the interface responses; however, no automatic verification is carried out when they are saved. Users should proactively conduct status checks, restrict the permissions of keys, rotate them regularly, and avoid assigning keys to sub-accounts.

API and Webhook

Insighto provides API keys, interface references, and Webhooks, which enable the management of assistants, sessions, voice functions, tools, as well as agent capabilities. The API key is displayed only once when it is created, and it follows a specific format with a fixed prefix.

Privacy and security

The platform handles account information, knowledge bases, visitor conversations, phone audio, transcriptions, tool outputs, and usage data. Customers own and control the data that is collected, and they are responsible for providing notifications, obtaining consent, ensuring lawful use of that data, and responding to requests from data subjects.

Communications regarding safety and compliance cannot replace a thorough procurement review. Before launching sensitive services, it is necessary to examine the existing trust credentials, penetration tests, sub-processors, data locations, incident response plans, and contractual responsibilities.

API, GitHub, and open-source status

Insighto has made its development documentation and APIs available, but no official core GitHub repository or complete platform code could be found. The projects with the same name that were discovered, as well as third-party connectors, cannot serve as a basis for considering this product as open source.

Go-live checklist

• Verify that the knowledge base is accurate, accessible, and free of any information that should not be made public;
• Testing all workflows for failures, timeouts, repetitions, and manual compensations;
• Obtain the necessary consent for phone recordings, text messages, WhatsApp, and outbound calls;
• Restrict the permissions of API keys, BYOK keys, and agent sub-accounts;
• Use an export IP allowlist and random paths to protect Webhooks;
• Set alerts for wallet balance, usage, concurrency, and carrier billing;
• Compare the quality of voice, latency, and actual cost across different models;
• Assign human agents to handle sensitive issues, complaints, refunds, and emergencies;
• Regularly conduct random checks on recordings, incorrect answers, transfers, and customer feedback.

Basic information

Frequently Asked Questions

Can Insighto.ai be tried out for free?

New accounts receive a $10 free credit to test voice and chat functions. To continue using these services, it is necessary to choose a subscription plan or top up the prepaid wallet.

Is the cost for basic voice services really only 0.06 dollars per minute?

$

Can I be connected to a live person?

Yes. The phone assistant can use transfer tools, and in chat or agent-based interactions it is possible to switch to a human agent; however, it is necessary to set the conditions, numbers, and context in advance.

Is voice cloning supported?

It is possible to create custom voices using ElevenLabs, but this requires the corresponding plan benefits as well as a sample of at least 30 seconds in length that meets the requirements. It is necessary to obtain the speaker’s permission and to comply with local laws.

Is WhatsApp supported?

Support is provided for direct use of the Meta Cloud API. When sending messages outside of the 24/7 customer service hours, approved templates must be used.

Is it suitable for reselling by agents?

There are Agency Starter and Agency Pro options, which offer white-label domain names, sub-accounts, concurrency limits, as well as features for managing campaigns and recurring billing. The number of customers and the amount of usage allowed are restricted by the plan chosen.

Are APIs provided?

APIs and Webhooks are provided, but there are specific restrictions on authentication, signing, and retry mechanisms. For critical processes, it is necessary to implement key protection, queue management, and compensation logic in accordance with the documentation.

Is Insighto.ai open source?

It is not open source. It is a commercial hosting platform; the availability of public development documentation does not mean that the core proxy operation system or the white-label control panel are open-source.

Summary

Insighto.ai is suitable for companies that wish to quickly set up AI agents for phone calls, web pages, and messaging; it also offers agents the options of white labeling, sub-accounts, and revenue sharing. Models, voices, workflows, GHL, and BYOK enhance the platform’s configurability.

Before use, the total cost should be calculated based on the actual model, audio, telecommunications, and messaging channels; special attention must be paid to current limitations such as two-factor authentication, Webhook signing, and retry mechanisms. Knowledge governance, consent for recording, key security, and manual upgrades determine the reliability of the system in production.