The organizations getting the most out of the CTI-to-SCV migration are the ones who understood, before they started, that they were making an architecture decision, not a technology refresh.
There is a common way enterprise organizations approach the Open CTI retirement announcement: find the migration path that most closely replicates existing telephony behavior inside Service Cloud Voice, execute the migration before February 2028, and return to normal operations. It is a reasonable response to a deadline. It is also a significant missed opportunity.
The organizations I have seen extract genuine competitive advantage from this transition share a different framing: they used the forced migration as the moment to answer a question they had been deferring: where does telephony actually belong in our Salesforce omnichannel architecture, and what does getting that right unlock?
That is a different question than “how do we keep the phone ringing inside Salesforce.” The answer to the first question makes the answer to the second much more valuable.
The Connector Era Versus the Platform Era
Open CTI was an integration bridge. A well-engineered one, for its time, but it is old and that time has passed. It allowed CCaaS vendors to embed their softphone UIs inside the Salesforce console without requiring desktop software installation, and it gave Salesforce access to call events; ring, answer, hang up, transfer, so that automation could trigger on those events.
What it could not do was make voice data first-class CRM data. Call recordings, transcripts, sentiment signals, silence analysis, these lived in the CCaaS platform. Getting them into Salesforce required heavy work in extraction, transformation, and synchronization. The latency was tolerable. The incompleteness was the problem.
Service Cloud Voice changes this at the data layer, not just the UI layer. Under SCV, each voice interaction generates a Voice Call object natively in Salesforce. Real-time transcription populates against the CRM record as the call happens. Post-call summaries write directly to the case. Sentiment data, silence analysis, and acoustic indicators are available as CRM attributes, not CCaaS exports.
The implication for AI is the one that matters most at enterprise scale. Agentforce operates on Salesforce data. An AI model that surfaces next-best-action guidance, drafts case summaries, or routes intelligently based on customer history can only do those things well if the voice interaction is fully represented in the data model it operates on. Under Open CTI, voice data was a second-class citizen, present, but only partially. Under SCV, it is first-class data from the moment of ingestion.
This is why Salesforce is not merely retiring an API. They are retiring the architectural assumption that voice is an external capability connected to the CRM. The new assumption is that voice is part of the CRM.
What SCV Actually Delivers and What It Does Not
Honest assessment of Service Cloud Voice requires distinguishing between what is available today and what is on the roadmap, and between what works at enterprise scale and what works in a demo environment.
What is production-ready and substantive:
Real-time transcription is the most immediate operational benefit our clients see value in. When agents stop taking manual notes mid-call, two things happen: conversation quality improves because the agent is fully present, and documentation completeness improves because the system captures everything. The productivity math on after-call work is not speculative, organizations deploying AI-generated call summaries consistently report ACW reductions of 50% or more. In a high-volume contact center, that recovered time is material headcount capacity without adding a single hire.
The unified data model produces better reporting and better AI performance over time. Organizations that have migrated and run SCV for twelve months or more report that the quality of their Agentforce recommendations improve because the training data is more complete. Voice interactions that previously existed as CCaaS exports; summarized, delayed, partially structured, are now full-fidelity records.
Einstein Next Best Action, when the voice interaction is visible to it in real time, surfaces guidance at the moment of need rather than after the call. The agent who handles a billing escalation and immediately sees a relevant retention offer is faster and more effective. The research on this is consistent: unified agent desktop deployments deliver 20-30% reductions in average handle time and 10-15% improvements in customer satisfaction. Given most organizations measure call volume and costs closely, ROI calculations can be done in a straightforward manner.
Where honest qualification is warranted:
Bring Your Own Telephony (BYOT) implementations are not equal. The depth of Service Cloud Voice integration varies significantly by telephony provider, and some providers like Genesys and Amazon have invested far more in their SCV connector than others. Before committing to a BYOT path with your current CCaaS vendor, require a technical demonstration against your actual use cases, particularly omnichannel routing, post-call data capture, and Agentforce compatibility. “SCV-certified” on the AppExchange listing does not tell you how complete the integration is. Additionally, keep in mind as you consider a migration that there will be a temporary dip in productivity as team members learn to use SCV. Overall efficiency improves over time, but it takes time for employees to become as reflexive on a new interface as they are on an existing one.
Also keep in mind that Agentforce Contact Center is genuinely different from Service Cloud Voice and that distinction matters for your migration planning. SCV is still the model where an external CCaaS platform carries the call and Salesforce receives the data. Agentforce Contact Center, launched in March 2026, runs the entire telephony stack inside Salesforce on Hyperforce. For qualifying use cases, it eliminates the external CCaaS platform entirely. As of May 2026 it is not yet enterprise-grade for complex contact center operations, but the trajectory is clear. Your SCV migration architecture should be designed to allow optionality on Agentforce Contact Center without requiring a re-architecture when you are ready to evaluate it seriously.
The Ownership Question Nobody Is Asking Their Vendors
Here is where the architectural decision becomes strategic rather than technical.
Both your CCaaS vendor and Salesforce now sell AI, analytics, workforce optimization, and automation capabilities. In the overlap between those capabilities, you are at risk of buying the same thing twice, implementing it in two places, and getting fragmented results from each – routing is a great example of this. Gartner research puts the SaaS overspend from duplicate capabilities at a minimum of 25% for organizations without centralized governance on this.
Before your first vendor conversation post-CTI-audit, your organization should have internal alignment on one question: what does each layer own?
- Obviously we are not unbiased, but here is the framework we have seen work for many clients. The CCaaS layer owns the infrastructure that earns the right to carry the call: carrier-grade SIP trunking, PSTN connectivity, uptime SLA enforcement, voice quality management, compliance for voice recordings, STIR/SHAKEN authentication. This is genuinely hard operational work that requires expertise, carrier relationships, and infrastructure depth that no CRM vendor should be expected to replicate. At a 1,000-agent contact center, a single hour of downtime costs can cost over $100 grand – this stuff matters.
- The CRM layer owns the intelligence that makes every interaction count: the customer record, case history, routing logic based on customer context, AI that resolves rather than deflects, the audit trail, the compliance record. In a fully connected ecosystem (note, Data Cloud and other consumption costs will apply), Agentforce can process a refund, update an entitlement, schedule a technician, but only because it has access to the data and business rules that authorize those actions. That data lives in Salesforce (often Data Cloud), not in the telephony platform.
The seam between these layers, omnichannel routing, is where Service Cloud Voice lives architecturally. The CCaaS platform handles transport and queue mechanics. The CRM provides the data-driven routing intelligence that determines which agent or AI agent receives the interaction based on customer context. This is the integration point that must be engineered well, and it is the primary design decision in any SCV implementation at enterprise scale.
Drawing this ownership line internally changes the nature of planning and architecture significantly. Vendors on both sides will propose capabilities that extend into the other layer. Having a clear internal answer on what you want to own in which system makes it far easier to evaluate those proposals on their merits rather than on vendor persuasiveness.
One important item to consider here is that you are undoubtedly going to be changing what data is stored in Salesforce. There are data compliance and privacy consdersations here. For example, if a call transcript includes someone validating their identity or discussing health care infomration, how much of that do you want to be stored in Salesforce, for how long and accessible by what humans and AI? As you begin to plan your migration, now is the time to have those discussions with your Legal and Compliance teams.
The Migration Sequencing That Actually Works
For large contact center environments, the migration that works is phased and sequenced around risk, not around speed.
Start with the audit, not the architecture. The CTI dependency audit described in the companion piece to this post is the mandatory first step. Architecture should not be designed before you know what you are migrating. Tightly coupled custom logic built on CTI events behaves differently in the Service Cloud Voice data model, and discovering that mid-migration is expensive.
Pilot before you scale. Select a single team or use case for the initial SCV deployment, ideally one with relatively simple call flows, all in a single language and with straightforward post-call requirements. Use the pilot to validate your SCV configuration against real call volumes, identify gaps between current behavior and SCV behavior, and build internal operational expertise before the scale deployment. Organizations that skip the pilot phase and go straight to full deployment consistently have higher costs and longer stabilization periods.
Treat AI enablement as a separate phase. The temptation is to activate real-time transcription, Next Best Action, and Agentforce capabilities simultaneously with the SCV go-live. Resist it. Get the telephony foundation stable first. Get that voice data populating correctly in Salesforce records, call routing working as designed, reporting validated – this stuff takes time. Then layer AI capabilities on a clean foundation. AI running on top of misconfigured telephony data produces worse outcomes than no AI at all.
The Broader Point
Every major platform transition generates a cohort of organizations that treat it as a compliance exercise and a cohort that treats it as a strategic inflection point. The CTI retirement is forcing a migration. The organizations that use that forced migration to make deliberate decisions about their contact center architecture, where voice data lives, how AI is governed, what each layer owns, will exit this transition in a structurally different position than those that simply kept the phone ringing. For those of you, like me, that have been around for a…..a bit, harken back to the Lightning migration days. Some organizations did a brute force lift and shift and didn’t gain any of the value from the improved user interface. More recently organizations did lift and shift migrations from Workflows to Flows, some were able to take advantage of the new power of flows, others did a brute force rip and replace and got no additional value.
The deadline is February 2028. The planning window is now. So which type of organization are you going to be – the one that benefits from the new power of SVC or the one that keeps doing work in 2027 the same way you did in 2015?
Ateko is the only partner in this conversation who sits on both sides of the ownership matrix. Our Salesforce practice has been implementing post-CTI architectures since SCV was released. Bell, our parent company, is a national carrier with decades of enterprise contact center infrastructure deployment, the pipes and the platform, all under one roof. If you would like a independent opinion before platform vendors start shaping your architecture, that’s a conversation worth having before the RFP goes out.
If your organization is rethinking where voice, AI, and customer data should actually live, now is the time to start that conversation before your migration path gets decided for you.
