Your Digital Human Provider Shut Down. What to Evaluate Before Choosing the Next One.

In February 2026, one of the best-known legacy CGI digital human platforms entered receivership with over $19.6 million in debt. Approximately 1,500 enterprise customers were left actively searching for an alternative — many of them with active contracts, live integrations, and a digital human built around the face and voice of someone at their company.

If you’re in that situation, the problem isn’t simply “find a new provider.” It’s replacing an entire infrastructure, preserving the cloned identity of your digital human, and doing it before the impact reaches your end users.

What you actually lost when your provider shut down

The contract is the easiest thing to replace. Everything else is what costs you.

• The conversational infrastructure. A digital human in production has a configured knowledge base, a conversational model tuned to your specific use cases, and responses refined over months. There is no way to export that into a standard format another provider can directly import. It has to be rebuilt.
• Integrations with your systems. If your digital human accessed your HRIS, CRM, or LMS in real time, those integrations are configurations specific to the provider’s system. When the provider shuts down, those connections shut down with them.
• The cloned identity. The face and voice of the real person from your company who brings the digital human to life require a cloning process that starts from new material — images, video, audio recordings. If the previous provider did not deliver the source files to you, the process starts from zero. Cloning also requires renewed explicit consent with the new provider.

The 3 traps of rushed migration

Under time pressure, the most common evaluation mistakes are predictable:

Trap 1: Choosing by name recognition. In this market, a well-known name is no guarantee the product is what you need. There are high-visibility vendors that generate pre-recorded video — not real-time conversation. Before moving forward with any provider, confirm that voice synthesis and facial animation are generated at the moment of response, with latency under 2 seconds.

Trap 2: Not evaluating identity cloning as a standard capability. Many providers offer image and voice cloning as an “add-on project” — not as part of the standard implementation process. That means more cost, more time, and more quality risk. Ask in the first conversation: is identity cloning included in standard onboarding, or is it a separate project?

Trap 3: Not verifying integrations before committing. “We can integrate with any system” is an answer that tells you nothing. What you need is technical documentation of available connectors and, if possible, a live demo where the digital human accesses test data from your actual systems. An enterprise-ready provider can do that in the second conversation.

The criteria your new provider must meet

After a forced migration, evaluation criteria need to be more rigorous, not faster:

1. Real-time synthesis with verifiable latency. The enterprise standard is less than 2 seconds between the user’s question and the digital human’s response — voice and facial animation generated simultaneously. Request this in a demo with questions the provider has not prepared for.
2. Identity cloning as a standard capability. Not as an add-on project, not as a professional services engagement. The process of cloning appearance from photos or short video, and voice from audio recordings, must be part of the standard implementation workflow.
3. Deployment in days, not months. The new provider must be able to have a first digital human operational in under a week, with primary integrations configured in parallel.
4. Documented integrations with your systems. Not generic promises. Real technical documentation — native connectors with leading HRIS, LMS, and CRM platforms, plus access via Zapier, Make, or n8n for everything else.
5. Clear data ownership for cloned assets. Biometric data — the cloned image and voice of real people — is yours. The contract must state this explicitly. It must also include portability clauses and a prohibition on using conversation data to train the provider’s models without consent.

A note on evaluating financial stability

The platform in question had raised over $135 million in venture capital before entering receivership. Funding raised is not a proxy for sustainability. When evaluating a new provider, ask questions that go beyond the product demo: What is the company’s revenue model? Does it generate recurring revenue from live enterprise deployments, or is it primarily pre-sales and pilot-stage? Is the business built around a platform that can scale independently of professional services?

A provider that depends on large implementation fees per deployment is structurally fragile at scale. A provider with no-code deployment and self-service onboarding can grow revenue without proportionally growing headcount — and that structural difference matters when you’re choosing a platform you’ll depend on for years.

How UNITH approaches migration

UNITH is designed for exactly the requirements a migration demands: deployment in days through a no-code process that requires no IT team, visual and voice identity cloning as a standard capability within the implementation flow, and real-time synthesis with latency under 2 seconds.

Integrations with enterprise systems — HRIS, LMS, CRM, and over 7,000 apps via Zapier, Make, and n8n — are documented and configured without custom development. The knowledge base is built through a structured onboarding process with your content, not a blank-slate prompt engineering exercise.

On the data ownership question: every cloned asset belongs to the client. UNITH does not use conversation data to train models without explicit written consent. This is in the contract, not just in the marketing deck.

In deployments with identity-cloned digital humans, interaction completion rates are observed to be 40 to 50% higher compared to text chatbots in the same use cases. Hypercustomized identity is not an aesthetic detail — it is what determines whether users actually engage.

What a migration engagement looks like in practice

Week one typically covers knowledge base reconstruction and identity cloning kickoff. The cloning process for a digital human requires the person’s participation in a controlled recording session — roughly 60 to 90 minutes of video and audio capture. This is not a bottleneck unique to any provider; it is the nature of the technology.

Week two covers integration configuration. If your previous deployment connected to an HRIS or LMS, UNITH’s native connectors for Workday, SAP SuccessFactors, BambooHR, Moodle, and others mean the integration is reconfiguration, not a build. For non-native systems, Zapier and Make connectors cover the vast majority of enterprise tooling without custom code.

By end of week two, most migrations have a fully operational digital human with live integrations, ready for QA and phased rollout. The timeline compresses or extends primarily based on content readiness — how organized your knowledge base source material is before the engagement begins.

Frequently asked questions

How long does migration to a new digital human provider take?

With a provider that has no-code deployment and documented connectors, a first operational digital human can be ready in 3-5 days. Integrations with HRIS, CRM, or LMS systems add 1-2 weeks. Knowledge base reconstruction is the most time-consuming part if you don’t have content documented outside the previous system.

Can I recover the cloned identity — face and voice — of my previous digital human?

It depends on what the previous provider delivered to you. In practice, most cases require starting the cloning process from new material — images, short video, and audio recordings of the person. This process requires explicit consent from the person whose image and voice are being cloned, even if a prior clone exists in another system.

What happens with the integrations I had configured?

Integrations are specific to the previous platform and are not directly portable. However, if the new provider has native connectors with the same systems, reconfiguration is faster than the original implementation. Request connector documentation before committing.

Is it possible to have a digital human operational within days?

Yes, with a provider that has designed their implementation process to be no-code. The bottleneck is not the platform — it is content preparation (knowledge base) and availability of material for identity cloning. What adds weeks or months is dependency on an IT team or provider professional services for basic configuration.

Ready to see how migration works for your use case? We’ll walk you through the migration path and show you a live deployment built around your requirements — not a generic demo. Talk to our team.