ElevenLabs Agents Alternative UK: Defaults to Change

If your team is evaluating ElevenAgents, there is a fair chance it did not start out shopping for a call platform. It may have started with ElevenLabs voices: a narration project, a product demo, a prototype agent in a browser tab that sounded better than anything the team had heard before. Then someone asked whether the same agent could answer the main phone line, and the question changed from "how good is the voice" to "what does this system do with a UK caller's data, and who picks up when it cannot help".

This guide is for the Head of Customer Operations who now owns that second question. It does not rank voices or argue about text-to-speech quality; the TTS voice selection guide owns that. It reads the ElevenLabs documentation as it stood on 31 October 2026 and asks one thing: what does ElevenAgents do by default, and what must a UK production phone line change before it carries real customers? Four decision layers matter most: how calls reach the agent, where data is stored, how long it is kept, and how a human takes over. Six documented defaults sit across those four layers. Pricing figures are ceded to the best AI voice agent 2026 comparison, transfer law to the international transfers guide, and the suite-versus-call-platform frame to the Parloa alternative guide.

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. Or see DATS, our five-stage method for placing AI inside enterprise operations.

What is ElevenAgents, and why are UK teams comparing it with call platforms?

ElevenAgents is the agents platform from ElevenLabs, the company best known for synthetic voices. Its documentation describes four core components working together: speech recognition, a language model you choose or bring yourself, low-latency text-to-speech, and a proprietary turn-taking model. UK teams compare it with call platforms because a strong voice is only one layer of a phone service, and the other layers carry the operational and data risk.

The ElevenAgents overview positions it as a full omnichannel agents platform, distinct from two lighter products the same page describes: Realtime models, a standalone speech-to-speech API, and Speech Engine, a speech orchestration layer for teams that already run a text-based chat agent and want to add voice. That distinction matters before any comparison starts. A team that only needs better speech on top of an existing bot is comparing Speech Engine, not ElevenAgents, and a voice-first call platform is the wrong benchmark for that decision.

For the team that wants a complete agent on the phone, the overview lists the parts you would expect: visual workflows, a knowledge base with retrieval, tool calling, testing, experiments on live traffic, analytics, and a command-line tool for managing agents as code. ElevenLabs also states a library of 10,000+ voices. On languages, the same overview page gives two different figures: its configuration table says 31 languages for voice and language settings, while its architecture section says 70+ languages for text-to-speech. Neither number should go into a business case until the vendor confirms which applies to your agent's configuration, and the language guide adds that a new agent defaults to English on the Flash v2 model and switches to the v2.5 Multilingual model when you add languages.

None of this is a weakness. It is a reminder that the product a team fell in love with in the browser and the product that answers a regulated UK phone line share a voice, not a configuration. The rest of this guide is about the configuration. If you are still deciding between building on components, orchestrating them or buying a platform, the build, orchestrate or buy guide sets out that earlier decision.

What does ElevenAgents do by default that a UK phone line must change?

ElevenAgents ships with defaults that suit fast building better than a UK production phone line. Its documentation states that conversation data is retained for two years by default, that audio recording is on by default, that customer data is stored in the United States as standard, and that Twilio numbers route through the US region unless configured otherwise. Each default is reasonable for a prototype and each needs a deliberate UK decision before launch.

The table below sets each default against the decision it forces. Every default is taken from ElevenLabs' own documentation, retrieved on 31 October 2026, and every one is configurable. The point is not that ElevenAgents gets these wrong; the point is that nobody should discover them after go-live, which is why a pre-launch review belongs in any rollout plan, whichever platform or delivery partner such as DATS you use.

The rows split into two groups. Retention, recording and storage are data decisions, owned by whoever signs off your data protection impact assessment, the step the UK and EU voice AI compliance guide walks through. Routing and handover are telephony decisions, owned by whoever runs the contact centre platform. Where these are different people, the agent configuration sits between them, and a buyer who sends the vendor questionnaire to one team and not the other risks signing off half of the table.

The same split applies to any platform, including ours. The enterprise AI voice agents guide walks through the full production checklist; this post stays on the six rows above because they are where ElevenAgents' documented behaviour is most specific.

How does an ElevenAgents agent reach a UK phone number?

ElevenAgents reaches the phone network through several routes, and the route you choose shapes everything after it. The documentation lists a native Twilio integration, generic SIP trunking, named carriers such as Telnyx, Vonage, Plivo, Bandwidth, Sinch and Vobiz, and contact centre connectors for Genesys, Five9 and Amazon Connect, plus Microsoft Teams. The route decides which handover types are available and, on Twilio, which routing region applies, so pick it first.

The SIP trunking documentation is the most useful page for a UK telecoms team, because it is specific. Your system must support at least one of the G.722 or G.711 (PCMU or PCMA) audio codecs, or resample audio itself. Signalling can run over TLS, and the RTP media stream can be encrypted; the page recommends TLS transport with required media encryption and describes UDP as experimental, to be replaced by TCP or TLS for production. For enterprise clients who allowlist by IP address, ElevenLabs offers static IP SIP servers drawn from a /24 block of 256 addresses spread across the US, EU, India and Singapore, and the whole block must be allowlisted.

The native Twilio route is simpler to stand up and unlocks features SIP does not, which the handover section below covers. It also carries the routing default that catches most UK teams. The Twilio regional routing page states that Twilio numbers route through the US region, us1, by default, "even if the phone number itself is from another country", and gives the example of an Italian number still routing through the US. If you then configure ElevenLabs for a different region and do not tell it which Twilio region the call is in, the page warns that operations on live calls, such as transfers, fail because the requests go to us1.

That is a Twilio behaviour, not an ElevenLabs flaw, and any team importing Twilio numbers into any platform should check it. For a UK deployment the practical step is the same everywhere: confirm the edge location on every imported number before the first live call, and record it in the change log.

For an estate already standardised on a contact centre suite, the CCaaS connectors are the more natural path. Genesys is reached through its Audio Connector, Five9 through AI Agent Connect, and Amazon Connect through its native third-party agent integration, according to the ElevenLabs documentation index. These connectors keep routing, queues and agent desktops in the suite you already run, and put ElevenAgents behind them as the conversational layer, a split the orchestration versus platform guide examines in more depth.

Where does ElevenAgents store UK call data, and what does EU residency change?

ElevenLabs states that customer data is hosted and stored in the United States as standard, with isolated environments available in the EU, India and Singapore as an Enterprise feature. Its data residency page does not list a UK environment. An isolated environment is a separate workspace with its own address, API endpoint and API key, so choosing one is a migration decision, not a toggle.

The ElevenLabs data residency page is careful to separate storage from processing, and UK buyers should be equally careful. The page says:

"While storage will take place in the selected location, processing may nevertheless occur outside of the selected location, including by ElevenLabs' international affiliates and subprocessors, for support purposes, and for content moderation purposes."

That sentence is the most important line in the residency documentation, and it is to ElevenLabs' credit that it is written down. It means an EU isolated environment answers the storage question and leaves the processing question open, which is why every vendor review, and the AI operating model that owns it, should ask for the two answers separately. The same page goes further for the EU: users "may restrict processing to the EU" by combining Zero Retention Mode with the API, with an exception for optional integrations such as a custom language model or post-call webhooks that need out-of-region processing. In other words, the strictest EU configuration depends on settings that also change how you get call records back, which the next section covers.

For a UK controller, none of this is disqualifying. US storage raises a restricted transfer question, set out in the international transfers guide linked above, and the sub-processor chain behind any vendor is covered in the sub-processor authorisation guide. What is disqualifying is assuming that "EU residency" means "nothing leaves Europe", or that a UK telephone number implies UK processing.

There is also a route that removes the question entirely. ElevenLabs' private deployment page says ElevenAgents can run inside your own cloud account on AWS or GCP, with Azure planned later in the second half of 2026, and that text, audio and transcripts never leave your network. The technical documentation sits behind an NDA, so a buyer cannot assess it from public material, but for a large regulated estate it is a genuine option and should be on the shortlist alongside any hosted alternative.

How long does ElevenAgents keep transcripts and recordings by default?

ElevenLabs' retention documentation says conversation data is kept for two years by default, and that transcripts and audio recordings can each be set to a number of days, kept indefinitely, or scheduled for deletion. Audio saving is on by default. Zero Retention Mode stops recordings and PII-bearing transcripts being stored at all, but call records then arrive only through post-call webhooks, and debugging becomes harder.

The retention page is clear about the mechanics. You can set any number of days, set the value to -1 for unlimited retention, or set it to 0 for scheduled deletion, and you choose whether a change applies to existing conversations or only new ones. The page also offers two pieces of guidance that a UK reader should weigh differently. Its GDPR guidance is to set retention periods that align with your processing purposes, which matches the ICO's storage limitation principle. Its HIPAA guidance is to retain records for a minimum of six years, which is US healthcare law and has no bearing on a UK contact centre. A team copying settings from a US template can import the wrong number.

The audio saving page says recordings are enabled by default and that disabling them still leaves transcripts viewable. For many UK workloads, such as appointment changes or delivery queries, a transcript answers every review question and the audio adds risk without adding evidence. For others, such as complaints handling or regulated sales, the recording is the evidence. That is a per-agent decision, and ElevenAgents lets you make it per agent.

The Zero Retention Mode page describes the strictest setting. With it on, no call recordings are stored and no transcripts or call metadata containing personal data are logged or stored after the call. To retrieve anything about those calls, you must use post-call webhooks. The page warns that Zero Retention Mode "may impact ElevenLabs' ability to debug call-related issues", and notes that under it the only Realtime model available is Eleven Realtime v1 Max, in the US and EU only. The practical consequence is that the strictest privacy setting moves the system of record onto your side: your webhook receiver, your storage, your retention schedule. That is often the right design, and it is also a build task someone has to own. The voice AI data retention guide covers how to set the schedule itself.

ElevenLabs also documents conversation history redaction, which removes detected sensitive entities before storage. The page says it is available to select enterprise clients only and that its detection rate is not 100%, so treat it as harm reduction for stored logs, not as a substitute for a retention decision.

How do human handovers work on each ElevenAgents telephony route?

ElevenAgents supports three ways to pass a call to a human: a conference transfer, which is the default; a blind transfer, which keeps the caller ID; and a SIP REFER transfer. Which ones you can use depends on how the number reached the platform. The documentation says warm transfer messages to the receiving human only work when the number is imported through the native Twilio integration.

The transfer to number documentation sets out the trade-offs plainly. A conference transfer calls the destination, adds them to a conference with the caller, then removes the agent. A blind transfer sends the call straight to the destination and preserves the original caller ID, but it is only available on numbers imported through the native Twilio integration and is configured through the JSON editor. A SIP REFER transfer works with phone numbers and SIP URIs, needs your SIP trunk to allow REFER, and does not support warm transfer messages. It is only available when SIP is used during the conversation, and when the destination is a SIP URI it can carry User-to-User Information to a receiving platform such as Genesys Cloud or Talkdesk, which is how context travels when the receiving system can read it.

For a UK contact centre this produces a real design choice. If the team wants the agent to brief the human before the caller is connected, the native Twilio route is the one that supports it, and the Twilio regional routing default from the earlier section then applies. If the estate runs on SIP into an existing contact centre platform, the context has to travel as SIP headers or UUI and the receiving desktop has to display it. Neither is wrong, but a buyer who assumed "warm transfer" was a single feature will find it is a property of the route.

The handover is also where a voice agent earns or loses trust with callers. The warm transfer and context handoff guide covers what the human should receive and how to measure it, and applies whichever platform you choose.

The same diagnostic logic underpins our AI operating model design, which decides who owns each of these settings once the agent is live.

When is ElevenAgents the right choice for a UK team?

ElevenAgents is a strong choice when voice quality and breadth are the deciding factors, when the team has engineers who want agents managed as code with testing and experiments, when an existing Genesys, Five9 or Amazon Connect estate should stay in charge of routing, or when a private deployment inside your own AWS or GCP account is required. In those cases a UK team should configure it carefully rather than replace it.

Four scenarios favour it clearly. First, brand-led voice: a library of 10,000+ voices, by ElevenLabs' own account, gives a marketing or brand team a wide palette to choose from, and the persona and brand voice guide explains when that room matters. Second, engineering-led teams: branches, merge proposals, simulated conversations, automated tests and live experiments are all documented, which suits a team that already ships software with review and rollback. Third, suite-first estates: the CCaaS connectors let ElevenAgents sit behind the routing you already trust. Fourth, sovereignty-first estates: a private deployment answers residency in a way no hosted service can.

ElevenLabs also states on its data residency page that it maintains SOC 2 certification and offers Business Associate Agreements for HIPAA-eligible services, and it runs a public trust center. Those are claims a procurement team can test against documents, and they count in its favour.

The scenarios that favour a different platform are narrower but common in UK operations. A team without engineers to own webhooks, routing regions and transfer types may prefer a platform where more of those decisions arrive as default settings. A team running outbound campaigns needs calling windows, opt-out handling and screening to be part of the product, not a build. The Vapi alternative guide covers the outbound duties that stay with the deploying business whichever platform you pick.

How does Dilr Voice differ from ElevenAgents for UK phone work?

Dilr Voice is a multi-agent voice AI platform built for phone calls rather than for voices. It chains specialised agents, such as a greeter, qualifier and customer care agent, into a single call, ships per-country compliance rules by default, and runs telephony on Twilio or Vobiz. It does not show a SOC 2 or other certification on its live pages, which ElevenLabs does state for itself.

The differences that matter for the six defaults above are these. On telephony, Dilr Voice runs on Twilio or Vobiz, and outbound campaigns use the customer's own Twilio or Vobiz account, so the number configuration check above belongs in any Dilr Voice rollout too. ElevenAgents documents more telephony routes and contact centre connectors than our live pages name, and a team that needs Genesys or Five9 in front should weigh that honestly. On compliance behaviour, Dilr Voice ships recording consent, DNC registry checks, opt-out recognition, permitted calling hours and full audit trails on every call as per-country defaults.

On handover, Dilr Voice offers warm transfer to a human with full context from the inbound front desk, alongside business-hours configuration and after-hours routing. On hosting, it runs on Google Cloud Platform, encrypted at rest and in transit, with dedicated tenancy and regional data residency options for enterprise; ask us which regions apply to your contract rather than assuming any location. On structure, each agent in a Dilr Voice call has its own prompt, model and tools, and passes context to the next agent on handover, which suits calls that move from qualification to an action such as a booking. Responses come back in under 500 milliseconds on DILR's own on-platform measurement, and the platform supports 30+ languages.

We will not claim a better voice library; ElevenLabs is the voice specialist. The fair comparison is narrower: if your problem is a UK phone workload where compliance rules and handovers matter more than voice range, compare the defaults, not the demos. Dilr Voice offers a free trial with no credit card, and the Dilr Voice product page sets out the call flows. Where a team wants an independent view of where AI belongs in its call operation before choosing any platform, our AI placement diagnostic exists for that.

What is the best ElevenLabs agents alternative for UK teams in 2026?

The best ElevenLabs agents alternative for a UK team in 2026 depends on which default it most needs to change. A team that wants per-country compliance rules and warm handovers as defaults should look hard at a call-first platform. A team that needs Genesys or Five9 in front, a private cloud deployment, or the widest voice library may be better served by configuring ElevenAgents carefully than by replacing it.

Four criteria decide it, and each maps to a section above:

1. Telephony route. If your numbers must stay on an existing contact centre suite, ElevenAgents' CCaaS connectors are a genuine advantage. If you run on Twilio or Vobiz directly, a call-first platform built on those carriers, such as Dilr Voice, is the closer fit.
2. Data location. If US storage with standard transfer safeguards is acceptable, the hosted service works. If you need storage outside the US, an ElevenLabs isolated environment or private deployment is required, and you should get the processing position in writing for any vendor.
3. Record ownership. If you want the vendor to hold transcripts and recordings for review, set retention deliberately. If you want nothing stored vendor-side, plan and staff the webhook receiver that Zero Retention Mode implies.
4. Handover design. If the human must be briefed before connection, confirm the route supports it. On ElevenAgents that means the native Twilio integration, and the warm transfer guide lists what the brief should contain.

Other names belong on a UK shortlist for different reasons. PolyAI suits teams that want a managed enterprise build, covered in the PolyAI alternative guide. Retell AI and Vapi suit developer teams that want an orchestration layer, compared in the Retell alternative guide. For a side-by-side view with pricing shape, the best AI voice agent 2026 guide linked in the introduction holds the full comparison.

Where ElevenAgents clearly wins, say so in your evaluation and keep it. Where it does not, remember that much of the hard work, the call design, the knowledge base and the handover rules, belongs to you and can move with you.

Does ElevenLabs keep UK call data in the UK?

ElevenLabs' data residency documentation lists the United States as the standard storage location and offers isolated environments in the EU, India and Singapore as an Enterprise feature. It does not list a UK environment. The page also states that processing may still occur outside the selected location. A UK team needing UK-only handling should ask about private deployment in its own cloud account.

The ElevenLabs page linked above is the place to confirm the current list, because isolated environments are added over time. Ask any vendor, including us through the Dilr.ai contact page, for the storage location and the processing locations as two separate written answers.

Does Zero Retention Mode mean ElevenLabs stores nothing about a call?

ElevenLabs' Zero Retention Mode documentation says that, with it on, no call recordings are stored and no transcripts or call metadata containing personal data are logged or stored after the call. Information about those calls is available only through post-call webhooks. ElevenLabs warns that it may limit its ability to debug call issues, so your own logging must fill that gap.

That makes Zero Retention Mode a design choice about where your system of record lives, not only a privacy switch. Read the observability and tracing guide before you enable it, so the evidence you lose vendor-side is captured on your side.

Want to see this in production? Try Dilr Voice live, compare platforms in the best AI voice agent guide, see our DATS methodology, or read about our approach to placing AI inside enterprise systems. More comparisons live in the voice AI library.

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.