Tovie

Tovie is an enterprise AI agent platform from Tovie AI Limited, a London company whose site lists an office on City Road. It is a low-code builder for voice and chat agents, and the home page leads with voice agents that clear call queues for councils, financial services firms and contact centers. The company positions it for IT leaders at large and mid-sized companies, not for individual makers.

What you build and how

The home page lists three kinds of agent work: outbound campaigns and reminders, inbound triage, and collections. Outbound and inbound calls are carried "from hello to resolved" by an AI call bot, and the collections work covers arrears calls that agree a payment plan on the call and update the CRM on their own. The company names its collections product Recobra. The platform also covers chat, since its own description is "Chat, voice and actions on one stack we build and run ourselves", and the company calls it "Our own platform, not a wrapper".

The builder is where the product sits in this directory. The home page says "Design the call as a flow, not a script", and describes routing between specialized agents, calling your tools and APIs, and shipping straight to production. In the company's own comparison post, the interface type is a low-code visual builder, and functions can be written in JavaScript, Python or a proprietary DSL for deterministic scenarios. The post says a built-in agent template catalog exists, and it says a built-in AI assistant for workflow creation is still in development. A reader should take those as the vendor's description, since the post is a comparison against Dify and n8n written by Tovie.

Extensibility is described in the same post: MCP (Model Context Protocol) with external servers and self-hosted MCP deployment, custom functions, and a REST API. For lifecycle control, the post says agent versions can be published and rolled back to a previous version. The post also says Tovie combines chat, voice and action automation and integrates with various channels and applications, and its call to action mentions flexible ASR and TTS (speech recognition and speech synthesis) integrations and connection to existing systems.

Deployment is a choice. The home page says "Cloud or fully on-premise, with your models or ours - and a human on the line whenever a call needs one". The comparison post lists the license model as proprietary, with a vendor cloud, a turnkey on-premises deployment inside the client's own infrastructure, and a SaaS cloud called Tovie AI Cloud. There is no open-source edition listed.

Visibility is part of the offer. Under "See every conversation", the company says you get topics, automation rate, sentiment and cost, measured on real traffic from the calls your agents take, and it shows a live analytics view on platform.tovie.ai. Every call ends with the outcome written back, so your team sees each conversation. The call flow on the home page shows the agent taking the action on the call, handing to a person when someone asks, and writing the result back.

The pilot process is spelled out as four steps. You pick the queue that hurts most, such as arrears, a waiting list or renewals. You connect your telephony and CRM, either in the cloud or on-premise in your estate. The agents take the calls and you watch every conversation live. Then you count the results on your queue and decide. The company adds that if you outgrow the product, the same team builds custom on the same platform with no migration and no new vendor.

The menu lists Agent Platform and Data Agent as products, and there are industry pages for financial services and insurance, councils and public services, contact centers and outbound, and collections and recoveries. The Agent Platform is the builder this entry describes.

What the pages state

The pages show no prices. The pilot is quoted: the home page says the company will show what a pilot would cost for a queue of your size, and the main calls to action are Book a demo and Get a personalized demo.

Customer results are the main evidence, and they are the company's own. For Richmond and Wandsworth Councils, which use voice agents in adult social care for an occupational therapy waiting list, the page cites 74% of calls gathering what was needed, and says the service is live at two councils. For BNP Paribas Cardif, which uses voice agents in an insurer's contact center, the page cites 66% of interactions automated and $9m saved over two years in production. The collections section cites 56% of debts recovered after a single call. A third story covers Kodland, an edtech company, with a contact center bot. A testimonial from a director at the Royal Institute of Navigation praises how easily the implementation integrated with its front-end and back-end systems.

On compliance, the home page says the platform is "The first conversational AI platform validated on IBM Cloud for Financial Services", and promises enterprise-grade compliance at mid-market speed and cost. The footer lists a GDPR page, but the pages do not name SOC 2 or HIPAA. The comparison post describes support as an enterprise SLA with L1 and L2 levels, a dedicated team and implementation support, and the target audience as CTOs, IT directors and chief AI officers at large and mid-sized companies. It also says the platform suits large UK and EU companies with strict infrastructure and data requirements.

Who it suits and who it does not

Tovie suits a public body, insurer, lender or contact center that has one painful call queue and wants a vendor to carry it from pilot to production, with the option of keeping everything in its own infrastructure. A technical team that wants to write functions in JavaScript or Python and connect MCP servers will have room to do so.

It does not suit a solo founder or small business that wants to sign up, pick a template and publish a website chatbot the same day. There is no self-serve plan or price list on the pages, and the sales path starts with a demo. Anyone who needs an open-source stack should look elsewhere, since Tovie describes its own model as proprietary. The strongest evidence is voice-heavy and concentrated in the UK and Europe, so a team that mostly wants a text chatbot should ask for chat references specifically before committing.