Voice AI18 min read

Vonage AI Studio Alternative UK: Who Owns the Build

Dilr Voice is an enterprise voice AI platform from DILR.AI that answers inbound calls and runs outbound campaigns, for UK teams weighing Vonage AI Studio. This guide tests AI Studio on its fixed engine choice, who owns the build, call handover, default retention windows and outbound limits, and when staying on Vonage is the right call.

On this page 12 sections
Vonage AI Studio Alternative UK: Who Owns the Build DILR VOICE Vonage AI Studio Alternative UK: Who Owns the Build 01 Pick the engine 02 Name the owner 03 Test the handover 04 Fix retention dilr.ai/blog

Vonage has built a virtual agent studio on top of its communications APIs. Its UK AI Studio page offers two ways in: start with a fully managed deployment by Vonage's AI experts, or build your own agent in a low-code, no-code flow builder. For a UK contact-centre or digital lead whose numbers, messaging and perhaps contact centre already run on Vonage, that looks like the shortest path to an AI agent on the phone.

The shortest path is not always the cheapest one to change later. Vonage's own documentation sets out an engine choice that is fixed when an agent is created, three different routes to ownership of the build, and default retention windows for transcripts and recordings. Those decisions shape how easily a phone agent can be changed once it is live, and this guide is built on them rather than on a feature tour. It is written for the Head of Customer Operations who has to decide whether the phone line goes to AI Studio, to a phone-first voice platform, or stays with people.

The scope is deliberately narrow. The general question of whether to build, orchestrate or buy voice AI belongs to our build, orchestrate or buy guide, and choosing a carrier belongs to our telephony provider selection guide. Designing an intent taxonomy is covered in our natural language understanding guide. No Vonage or DILR prices appear here. What this guide holds is the ownership test: who builds the agent, on which engine, and what happens to the calls and the data once it is live.

This guide is shipped by the team behind Dilr Voice, a multi-agent voice AI platform that chains specialised agents into a single phone call and warm-transfers to a human with full context. Or see DATS, our five-stage AI consulting system, if the harder question is which calls should be automated at all.

What is Vonage AI Studio, and how do UK teams buy it?

Vonage AI Studio is a builder for virtual agents on voice, WhatsApp, SMS and HTTP channels, sitting on Vonage's communications APIs. Vonage says the agents can be bought standalone or bundled with other Vonage products such as Vonage Contact Center. A team can build its own agent in a low-code flow builder, or have Vonage's AI team design, develop and maintain it.

The bundle line comes from the AI Studio FAQs, and the managed route from the UK product page, which describes a fully managed option in which the Vonage AI team will design, develop and maintain the solution, plus an AI Studio Services offer of conversation designers and AI services professionals. The bundle is a purchasing choice. The build routes are an operating-model choice: building it yourself, a fully managed deployment and AI Studio Services each put the knowledge of how your agent works in a different place, and that knowledge is what you need when call volumes or policies change.

On the phone side, the voice getting-started guide says a voice agent is created by choosing the Telephony option, and that only an agent with a phone number can be published. The numbers come from a Vonage API account. Agents can be inbound, handling questions such as opening hours, or outbound, sending notifications and updates. The same guide lists a range of Amazon Polly voices plus premium neural voices, with Microsoft Azure and Almagu voices for Hebrew.

For answers drawn from your own content, Vonage offers Knowledge AI. Its Knowledge AI guide describes a retrieval-augmented generation system that combines semantic search with Google Gemini models, fed by uploaded files, single URLs or public cloud documents. The guide asks builders to convert tables into flat sentences and replace images with descriptive text, and says table content and image-embedded text are not indexed. That is good advice for any phone agent, because a caller cannot see a table either. Our enterprise AI voice agents guide sets out what a production rollout needs whichever vendor you choose.

Which AI engine should an AI Studio voice agent start on?

Vonage AI Studio offers three engines for understanding callers: Traditional NLU, Hybrid NLU and Agentic NLU. The choice matters more than it looks, because Vonage's documentation says the engine is fixed when an agent is created. A team that starts on one engine and later wants another rebuilds or duplicates the agent, and the Agentic engine cannot be reached that way at all.

The plainest statement is in Vonage's NLU engine guide, which warns: "Once a Virtual Agent is created, the AI Engine cannot be changed later." Between Traditional and Hybrid, the documented workaround is to duplicate or import the existing agent and pick the other engine during duplication, then fix any custom entities that need the new Description field. That is a migration project with its own testing, not a setting.

Each engine carries its own documented trade-offs:

EngineHow Vonage describes itLanguagesMoving an existing agent onto it
Traditional NLUIntents trained on example user expressionsShorter list, several in betaDuplicate or import from Hybrid
Hybrid NLURule-based logic plus Google Gemini models; labelled public beta and described as production-readyLonger list, including English (United Kingdom)Duplicate or import from Traditional
Agentic NLUGoal-driven agents with a perceive, plan and act loop; a gated featureVoice agents and English onlyNot possible from either other engine

Source: Vonage AI Studio documentation, NLU engine, Agentic NLU and Languages Available pages, read on 20 November 2026.

Two Hybrid figures are Vonage's own. The guide claims up to 20% better intent detection, based on Vonage's internal testing, and tells builders to expect about 100ms extra response time, which it calls imperceptible to callers. Neither has been tested independently here, so treat both as a starting hypothesis to check on your own calls.

The Agentic NLU guide is candid about its limits. It is available for voice agents and English only. Its responses are generated by language models and cannot be defined manually, and its retry and reconfirmation behaviour is predefined. Agents built on it cannot be converted to the other engines, agents on the other engines cannot be imported into it, and it is not supported by Vonage's Virtual Assistant Historical Analytics, so session reporting needs one of the other two engines. A team choosing an engine should therefore decide first what it must report on and in which languages, and only then which engine sounds most capable.

Who owns the build: your team, Vonage or a voice platform?

Vonage AI Studio supports three ownership models. Your own team can build and maintain flows in the studio. Vonage's AI team can run a fully managed deployment. Or Vonage's services professionals can work alongside your team. A separate phone-first voice platform is a fourth option. The right one depends on who will change the agent after launch, not who builds the first version.

A phone agent keeps changing after go-live. Opening hours move, a product is withdrawn, a regulator updates a script, and the escalation rules need tightening after the first complaint. In a self-built studio agent, those changes sit with whoever on your team knows the flows, intents and Knowledge AI sources. In a fully managed deployment they sit with Vonage, and the speed of every change depends on the terms of that service. Neither is wrong, but the business case should say which one you are buying, and the contract should say how quickly a change request is turned round.

Four questions settle ownership before a pilot:

  • Who will edit flows, intents and Knowledge AI sources after launch, and how many people on your side understand them?
  • If Vonage runs a fully managed deployment, what is the written turnaround for a content change, and for an urgent change such as a withdrawn product or a safety script?
  • Which engine will the agent use, and who decides if it ever needs to move to another, given that a move means duplicating or rebuilding?
  • If you leave Vonage, what can you take with you? The documented duplicate and import routes work inside AI Studio, so ask whether flows, intents and Knowledge AI sources export in a form another platform can use.

The generic economics of building against buying are covered in our in-house versus vendor operating model guide. The same diagnostic logic underpins our AI operating model consulting, which decides who owns a phone agent's content and changes once it is live.

The ownership test before building on AI Studio
01Pick the engineReporting and language…02Name the ownerWho changes the agent03Test the handoverWhat reaches the human04Fix retentionHow long, and where
Settle each step in writing, in order, before a pilot; each answer names an owner the steering group can hold to it.

What happens when an AI Studio agent hands a call to a person?

In Vonage AI Studio, a voice agent hands a call to a person through a Call Routing node, which can send the call to a saved contact, a phone number or a number collected during the call. The node can pass the caller's keypad entries to the person who answers, but Vonage says that option is disabled when the destination is a SIP endpoint.

The detail is in Vonage's Call Routing guide. It describes a Failed outcome for calls that do not connect, and notes that even an answering machine picking up counts as connected. A Record Call option lets you listen to the routed conversation in the call reports, with the transcript saved in a parameter. In the documentation index, voice agents use this Call Routing node, while the WhatsApp, SMS and HTTP channels each list a separate Live Agent Routing node.

The practical test is where the context lands. If your agent routes calls to a SIP endpoint in your contact centre, the keypad option is disabled, and Vonage's SIP integration page shows how calls reach a virtual agent over SIP as well. If the person who answers sees only a ringing line, and the transcript sits in AI Studio's reports, the caller repeats themselves. Before a pilot, place a test call that fails on purpose and check what the human sees on their own desktop: the caller's name, the reason for the call, and what the agent already tried. Our warm transfer guide sets out what a good context handoff should carry, whichever platform makes it.

What should a UK buyer confirm about retention, region and processors?

Vonage's AI Studio documentation gives two answers on retention. Its FAQ says Insights API data such as transcripts is wiped after 90 days and recordings after 30 days, while its Reports page says all Insights API data, recordings included, is wiped after 30 days, with an increase to 90 days planned. A UK buyer should get the current windows in writing.

Both pages, the FAQ and the Reports guide, were read on 20 November 2026, and the FAQ points teams that need records for longer to the Insights API, to store the data in their own application. Short windows cut both ways. Short default windows suit the storage limitation principle: the ICO's storage limitation guidance says you must not keep personal data for longer than you need it, and that you need a policy setting standard retention periods wherever possible. The ICO notes that this guidance is under review following the Data (Use and Access) Act 2025. But a firm that must keep call records for complaints, disputes or its own regulatory reasons may need them for longer than 30 days, and then the copy in your own application becomes the record. Whoever builds that export owns its security and its own retention schedule. Our voice AI data retention guide covers how to set those periods.

Region is the second question. Vonage's outbound call guide lists API endpoints for agents deployed in an EU region and in a US region. Ask which region your agents will run in, and whether that covers recordings, transcripts and Knowledge AI sources as well as the agent itself.

Processors are the third. Different pages name different components: the UK product page refers to proprietary Vonage speech recognition and language understanding engines, the NLU guide says Hybrid NLU uses Google Gemini models, Knowledge AI uses Gemini too, and the voice guide lists Amazon Polly voices. None of that is a problem in itself, but your data protection officer will want each one mapped. Ask four questions before a pilot:

  • Which sub-processors handle speech recognition, language understanding, answer generation and speech synthesis on our calls, and in which locations?
  • Does the engine we choose change that list, for example Hybrid or Agentic against Traditional?
  • Which retention windows apply today, given that the FAQ and the Reports guide disagree, and can they be changed for our account?
  • Which region will our agents, recordings and Knowledge AI sources sit in?

The UK GDPR questions behind those four are set out in our UK and EU voice AI compliance guide. They apply to every vendor on a shortlist, Dilr Voice included.

How do outbound calls work in AI Studio?

In Vonage AI Studio, an outbound voice campaign is built through an API rather than a campaign screen. A team creates an agent with an Outbound Event, publishes it with a phone number, and calls Vonage's make-call API for each number. Vonage limits outbound calls to one per second by default and says it can raise that to three or five per second on request.

The outbound call guide adds that calls over the approved limit fail with a 429 error, and that a status URL can receive a notice when the recipient does not pick up. The guide also offers an optional setting, labelled as in beta testing, that hangs up when an answering machine picks up. Each request carries a destination number and an agent ID, so the guide describes outbound one call at a time.

What that guide does not describe is the campaign around the calls: the contact list, the permitted calling hours, retries and pauses, and screening against the Telephone Preference Service. With an API-first design, those sit with your team or with whoever builds the integration, so ask Vonage what it provides and plan for the rest.

That screening matters in the UK. The ICO's telephone marketing guidance says that, in general, you must not make marketing calls to any number listed on the TPS or Corporate TPS unless that person has specifically consented, and that automated marketing calls fall under the stricter rules in regulation 19 of PECR. The ICO notes that this guidance is also under review following the Data (Use and Access) Act 2025. Whichever platform places the call, the duty binds the organisation making it, not the software vendor. Our UK GDPR and PECR outbound calling guide sets out the full picture.

When is AI Studio the right choice?

Vonage AI Studio is the right choice when your numbers, messaging and contact centre already run on Vonage, when the first use cases are frequently asked questions and simple self-service, and when you want Vonage to build and maintain the agent. In that case one supplier for telephony, channels and the agent is a real advantage over adding a separate platform.

It is also a sensible choice for a team that wants the same studio to run WhatsApp and SMS agents beside the phone agent, since AI Studio builds those agents in the same studio. And a team that has no conversation designers of its own, and no plan to hire them, gets a managed route from a supplier it already holds a contract with. Our view is that such a team should give AI Studio a fair pilot on one line before looking further.

The picture changes in three situations. The first is when most calls need a warm transfer into human queues with context on the agent's screen, and your test call shows that context does not arrive. The second is when much of the work is outbound, such as reminders, renewals or collections, where list management, calling windows and do-not-call screening are the hard part. The third is when the engine trade-offs bite: you need reporting the Agentic engine does not support, or a language the Hybrid list does not cover. In any of those, a phone-first platform beside Vonage's messaging is a reasonable design, and our Intercom Fin Voice alternative guide runs the same handover test on a different vendor.

How does Dilr Voice differ from Vonage AI Studio?

Dilr Voice is a phone-first voice AI platform from DILR.AI rather than a studio built around a carrier's APIs. It chains specialised agents into a single call, such as a greeter, a qualifier, a knowledge agent and an action agent, each with its own prompt, model and tools. Its telephony runs on Twilio or Vobiz, not on Vonage, which is the first thing a Vonage customer should weigh.

That telephony point should be stated plainly for a team on Vonage. Dilr Voice does not run on Vonage numbers; its telephony runs on Twilio or Vobiz, and outbound campaigns require the customer's own Twilio or Vobiz account. A team that wants its phone agent on its existing Vonage numbers should confirm what moving to Twilio or Vobiz means for those numbers before a pilot.

On handover, Dilr Voice offers business-hours configuration, after-hours routing and warm transfer to a human with full context. On outbound, campaigns run from an uploaded contact list with scheduling windows, retry logic, auto-pause at the configured daily end time, and real-time outcome and sentiment analytics. Per-country compliance rules are shipped by default: recording consent, DNC registry checks, opt-out recognition and permitted calling hours, with full audit trails on every call. That covers much of the campaign layer the outbound section above describes; confirm for each country how the DNC registry checks map to your own screening duties, such as the TPS and CTPS in the UK.

On knowledge and after the call, Dilr Voice uses RAG knowledge bases built from uploaded documents, chunked automatically and searched during live calls, and post-call automation without code: CRM records, email, SMS and Slack notifications, and lead status updates, with HubSpot, Salesforce and Google Calendar named as live integrations. It also offers a platform API for agents, campaigns, leads, knowledge bases and call data. On hosting, it runs on Google Cloud Platform, encrypted at rest and in transit, with dedicated tenancy and regional data residency options for enterprise. It supports more than 30 languages, and its response time is under 500 milliseconds on DILR's own on-platform measurement.

Where Vonage AI Studio is stronger is the single supplier. Dilr Voice does not replace Vonage's numbers, messaging or contact centre, and a team that wants one supplier for all three gives that up by running its phone agent elsewhere. You can test the difference directly with a free Dilr Voice trial, which needs no credit card. For the delivery side, our AI execution office runs the pilot with your team when capacity is the constraint.

What is the best Vonage AI Studio alternative for UK teams in 2026?

The best Vonage AI Studio alternative for a UK team in 2026 depends on why AI Studio is not the answer. If the issue is outbound campaigns, warm handover or engine lock-in, a phone-first platform such as Dilr Voice is the natural comparison, tested on your own calls. If the issue is wanting a managed contact-centre AI suite, compare other suites instead.

Four criteria separate the options fairly, and AI Studio should be scored on them like everyone else:

  1. Change ownership. Who edits the agent after launch, how fast, and does any core choice, such as the engine, lock you in at creation?
  2. Handover into your real queues. Does an unresolved call reach the right human with the transcript and reason on that person's screen, over the telephony you actually use?
  3. Retention, region and processors. How long are recordings and transcripts kept by default, where are they processed, and which third parties touch them?
  4. Outbound readiness. If you need outbound calls, does the platform handle lists, calling windows, retries and do-not-call screening without custom build?

AI Studio wins outright on one thing a phone-first platform cannot match: a team whose numbers, messaging and contact centre are all on Vonage, and which wants Vonage to build and run the agent, should pilot AI Studio first. Other names on a UK shortlist may include PolyAI, Parloa, Vapi and Retell AI, each with its own balance of managed service and self-build, and each should be scored on the same four criteria. Our best AI voice agent 2026 guide compares the field on these criteria, and our Genesys Cloud AI agents guide covers the native-suite question for teams on a larger contact centre platform.

One wider number keeps the choice in proportion. McKinsey's State of AI 2025 found that about a third of organisations have AI in production. Production is still the exception rather than the rule. Our view is that ownership is the question to settle before comparing features. Our voice AI ROI framework shows how to value the calls an agent actually resolves.

Frequently asked questions

Does Vonage AI Studio support British English?

Vonage AI Studio lists English (United Kingdom) as fully supported on its Hybrid NLU engine, and says it reviews its language portfolio monthly. The Agentic engine is limited to English. Welsh appears only on the Traditional engine's list, marked as beta testing, and is absent from the Hybrid list, which matters for UK public services with Welsh-speaking callers.

The detail is on Vonage's Languages Available page, read on 20 November 2026. Language support for understanding is separate from the voice the caller hears, which in AI Studio comes from Amazon Polly or premium neural voices. Test both on real UK callers, including regional accents and noisy mobile lines, before deciding, and do the same for every platform on the shortlist, Dilr Voice included.

Where to go next

Vonage AI Studio is a credible choice for a team already on Vonage, and its documentation states its limits plainly enough to test. The decision turns on ownership: which engine the agent is built on, who changes it after launch, where the calls go when it cannot finish them, and how long the evidence is kept. Answer those four in writing, on one line, before anything is rolled out. If you would like a second view on the call mix itself, our approach explains how we sequence that work, and the AI placement diagnostic ranks which calls are worth automating first.

Want to see this in production? Read the best AI voice agent 2026 guide, browse more voice AI guides for UK teams, see our DATS methodology, or read about our approach to placing AI inside enterprise systems.

Talk to the operators

Decide who owns the agent first.

30-min scoping call · No deck · Confidential. We will map your call mix with you and tell you plainly whether AI Studio on your Vonage numbers already covers it.

Written by the Dilr.ai engineering team, practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed.

  • vonage ai studio alternative uk
  • vonage ai studio alternative
  • vonage virtual agent alternative uk
  • vonage ai studio vs dilr voice
  • voice ai reddit
  • best ai voice agent uk 2026
  • dilr voice

What does this article answer?

The questions this article answers, in short

  • Vonage AI Studio is a builder for virtual agents on voice, WhatsApp, SMS and HTTP channels, sitting on Vonage's communications APIs. Vonage says the agents can be bought standalone or bundled with other Vonage products such as Vonage Contact Center. A team can build its own agent in a low-code flow builder, or have Vonage's AI team design, develop and maintain it.
  • Vonage AI Studio offers three engines for understanding callers: Traditional NLU, Hybrid NLU and Agentic NLU. The choice matters more than it looks, because Vonage's documentation says the engine is fixed when an agent is created. A team that starts on one engine and later wants another rebuilds or duplicates the agent, and the Agentic engine cannot be reached that way at all.
  • Vonage AI Studio supports three ownership models. Your own team can build and maintain flows in the studio. Vonage's AI team can run a fully managed deployment. Or Vonage's services professionals can work alongside your team. A separate phone-first voice platform is a fourth option. The right one depends on who will change the agent after launch, not who builds the first version.
  • In Vonage AI Studio, a voice agent hands a call to a person through a Call Routing node, which can send the call to a saved contact, a phone number or a number collected during the call. The node can pass the caller's keypad entries to the person who answers, but Vonage says that option is disabled when the destination is a SIP endpoint.
  • Vonage's AI Studio documentation gives two answers on retention. Its FAQ says Insights API data such as transcripts is wiped after 90 days and recordings after 30 days, while its Reports page says all Insights API data, recordings included, is wiped after 30 days, with an increase to 90 days planned. A UK buyer should get the current windows in writing.
  • In Vonage AI Studio, an outbound voice campaign is built through an API rather than a campaign screen. A team creates an agent with an Outbound Event, publishes it with a phone number, and calls Vonage's make-call API for each number. Vonage limits outbound calls to one per second by default and says it can raise that to three or five per second on request.
  • Vonage AI Studio is the right choice when your numbers, messaging and contact centre already run on Vonage, when the first use cases are frequently asked questions and simple self-service, and when you want Vonage to build and maintain the agent. In that case one supplier for telephony, channels and the agent is a real advantage over adding a separate platform.
  • Dilr Voice is a phone-first voice AI platform from DILR.AI rather than a studio built around a carrier's APIs. It chains specialised agents into a single call, such as a greeter, a qualifier, a knowledge agent and an action agent, each with its own prompt, model and tools. Its telephony runs on Twilio or Vobiz, not on Vonage, which is the first thing a Vonage customer should weigh.
Written by Dilr.ai EngineeringEngineering team

Who can run AI voice agents like these for me?

Dilr Voice runs AI voice agents for inbound and outbound calls

Multi-agent handoff, RAG knowledge bases and per-country call compliance in one platform, priced per minute.