TL;DR
- Retell AI is the strongest starting point for inbound qualification that needs current CRM context and a human transfer path. Connected integration tools and custom functions can act during the call.
- Synthflow fits an operations-led enterprise deployment with visual flows, booking and conditional transfers. Current new-customer pricing starts at $30,000 annually; old low-cost self-service figures are not the purchase basis here.
- Bland fits structured pathways and enterprise warm handoff. Its low-cost Start and Build rates should not be presented as including the Enterprise warm-transfer feature.
- ElevenAgents fits a team that wants a voice platform with minute allowances, webhook tools and telephony transfers. Pro is $99/month for 1,238 minutes; model and telephony costs are additional.
- Twilio ConversationRelay fits engineers who own the qualification application and carrier workflow. Its $0.07/minute relay rate sits alongside voice and model costs, plus the work of building and maintaining the application.
What the agent must decide
An inbound caller is not automatically a new sales lead. They may be an existing customer, a buyer already speaking to an account executive, someone asking about a job, or a supplier. The first useful decision is where the call belongs. Qualification comes after that.
A practical agent should identify the purpose, ask enough to route, look up relevant CRM context, and either transfer or agree a next step. A call transcript is evidence of the conversation; it is not a substitute for the assigned owner, qualification result and follow-up.
Two integration patterns matter. Live tools can read a record or write a result during the conversation. Post-call sync records the finished call for reporting or follow-up. Both are useful. If an open opportunity should route to its owner immediately, a transcript sent after the call cannot make that decision in time.
This distinction does not make every inbound agent a complex CRM project. A small reception workflow can collect a callback request without live deal lookup. The five options below cover different levels of ownership, from visual enterprise deployment to an application your engineers build.
Shortlist comparison
US-dollar public bases checked on 30 September 2026. These are buying units and qualifying scopes, rather than an assumption that every displayed minute price contains the same services.
| Product | Live qualification and CRM pattern | Human handoff | Current price basis | Best fit |
|---|---|---|---|---|
| Retell AI | Connected provider tools or custom HTTP functions during calls | Cold, warm and agentic warm modes | Published $0.07–$0.31/minute range; displayed $0.11 configuration is an example | CRM-aware platform build without owning the entire voice stack |
| Synthflow | Visual flow, custom live HTTP actions and extracted variables | Conditional/dynamic phone or SIP destinations | Enterprise from $30,000/year; configured usage terms | Operations-led enterprise implementation |
| Bland | Pathway branches and live webhook/tool calls | Warm transfer requires Enterprise | Start $0.14/minute; Build $299/month plus $0.12/minute; Enterprise quoted | Structured enterprise qualification and briefing a human |
| ElevenAgents | Authenticated webhook tools and agent/workflow configuration | Phone/SIP transfers with transport-specific behaviour | Pro $99/1,238 minutes; additional $0.08/minute, model/telephony extra | Voice platform with a team-owned CRM endpoint |
| Twilio ConversationRelay | Your WebSocket application owns logic, CRM calls and state | Application ends relay session and routes using TwiML | $0.07/relay minute plus voice, number and model costs | Engineering team already operating Twilio |
The five choices
1. Retell AI: best default for CRM-aware qualification
Retell's current connected integration tools can act against supported providers including HubSpot and Salesforce. They can run during a conversation, use provider authentication managed by Retell, and retain returned values such as a contact ID for later actions. A custom function calls an endpoint you host for an unsupported system or your own business logic. Integration tools, custom functions.
That is a useful architecture for checking whether a caller is already attached to a deal, preserving the owner and writing a structured qualification result. A CRM logo alone does not define that result: the buyer still chooses fields, permissions and the conditions for creating a new record.
Transfer tools include cold, warm and agentic warm modes. A warm handoff is relevant when the receiving rep should hear why the caller qualifies before joining. The fallback matters just as much: if the destination is unavailable, collect a clear callback commitment rather than describing the transfer as completed. Retell transfer modes.
Its public range is $0.07–$0.31/minute. The displayed $0.11 example combines platform, selected model and voice with zero telephony in the calculator; it is not a universal all-in carrier rate. Retell pricing.
Choose Retell for a team that wants a hosted voice platform while retaining control of CRM logic. It is less suitable for a buyer expecting a finished qualification process without anyone owning integration and routing. Its connected tools reduce plumbing; they do not decide your sales policy.
2. Synthflow: best for a visual enterprise workflow
Synthflow's actions can call external HTTP systems during a live call. Extracted data can flow through a booking action, transfer whisper and backend request. Transfers support phone and SIP destinations, including dynamically resolved targets. This suits a visual flow that separates existing customers, new demand and specialist requests. Actions, call transfers.
For a qualification buyer, the advantage is operational ownership: an explicit sequence can ask the needed questions, run a lookup and select the destination. CRM reads and writes still depend on the configured action and its endpoint. A visual node is not proof that the correct record was changed.
Current new-customer pricing starts at $30,000 annually, a $2,500 monthly allocation before interpreting the configured agreement's allowances. The public price is a supported purchase floor, not enough evidence to invent an all-in per-conversation rate. Synthflow pricing.
Synthflow bills connected usage by its documented rules and stops AI billing on successful transfer. Use the contracted scope for platform minutes and other services rather than adding a guessed low per-minute plan to this enterprise floor. Usage basis.
Choose Synthflow when deployment support and visual operations justify an enterprise commitment. It is a poor economic fit for a small occasional inbound queue bought solely to save receptionist minutes. A larger business can justify it through coverage, workflow ownership and implementation value, but those benefits need to exist in the business case.
3. Bland: best for structured pathways and enterprise warm transfer
Bland pathways can branch and run webhook nodes during a live call. A lookup can return an account owner or eligibility result, and a later action can write the agreed outcome. Custom API tools are another way to reach a backend; post-call webhooks are a separate reporting mechanism. Live webhook nodes, custom API tools.
Bland's warm-transfer design calls the human, briefs them, then merges the conversation. Its documentation explicitly requires Enterprise. That is the relevant qualifying purchase when the recommendation depends on briefing and introducing the receiving rep. Warm transfer.
The current commercial page lists Start at $0.14/minute without a platform fee, Build at $299/month plus $0.12/minute, and Enterprise by agreement. Start is a lower-cost platform entry, not the advertised warm-handoff package. Bland pricing.
Choose Bland Enterprise for a business wanting explicit pathways and contextual human handoff, with its configured commercial scope. It is less suitable when the buyer only wants the cheapest displayed minute rate and assumes every enterprise node or transfer option is included. This recommendation makes no blanket certification claim; qualification and handoff are the reasons it belongs on the shortlist.
4. ElevenAgents: best for a voice platform with flexible tool integration
ElevenAgents' webhook tools can query external data or perform authenticated actions during a conversation. Tool configuration controls the request, parameters and authentication. That provides a practical way to call a CRM lookup or qualification endpoint while keeping business rules in your own service. Webhook tools.
The transfer-to-number tool supports phone and SIP destinations, but behaviour depends on transport. A warm operator message is supported through the native Twilio integration; SIP REFER does not provide that warm message. A team whose handoff requires a spoken briefing should choose the corresponding telephony arrangement rather than treating every transfer as identical. Transfer to number.
Current Pro is $99/month with 1,238 call minutes and twenty concurrent calls. Additional minutes are $0.08; LLM and telephony costs sit on top. Scale is $299 for 3,738 minutes and thirty concurrent calls. Burst pricing is $0.16/minute, so bursts can change the bill even where total monthly minutes appear modest. ElevenAgents pricing.
Choose ElevenAgents when its voice and agent tooling fit your team and you have someone to own the CRM endpoint and transport. It is not a ready-made sales-qualification service, and the minute allowance should not be mistaken for a complete phone-and-model bundle.
5. Twilio ConversationRelay: best for engineers owning the application
ConversationRelay handles speech recognition, speech generation and the live session while your WebSocket application receives transcribed speech and supplies responses. Your application owns the model, qualification state and CRM calls. This is a building block for a service rather than a complete sales agent. ConversationRelay architecture.
For handoff, the application can end the relay session with context and use the Connect action callback to route the call to a human. Twilio documents this separation between the virtual-agent session and the human leg. Your code must also deliver the right CRM context to the receiving rep. Handoff pattern.
The relay is $0.07/minute, with voice billed separately. US local inbound voice is $0.0085/minute and a local number $1.15/month. The model and application infrastructure remain your costs. Do not add a separate generic TTS bill automatically: the relay already performs speech conversion; budget the actual selected provider configuration. Relay pricing, US voice rates.
Choose ConversationRelay when your engineers already operate Twilio or need full application control. It is a weak fit for a sales operations team without an engineering owner. Low infrastructure rates do not pay for an unowned routing application.
A qualification and handoff design
For a B2B sales line, a useful flow is:
- Identify the purpose and a usable person/company identifier. Phone number helps lookup, but should not be treated as certain identity when shared or unknown.
- If a current customer or open deal is found, preserve the existing owner and route accordingly.
- For new demand, ask about the problem, company fit and timing. Gather only the information needed to select a useful next step.
- If a rep is available and the caller wants to speak now, transfer with a short context package. Otherwise offer a real calendar booking or an owned callback request.
- Record the result, reason, owner, next action and call ID. Attach a summary without substituting it for those fields.
For example, an existing account asking about renewal should reach its owner. A new business with a relevant need can qualify for a discovery call. A candidate asking about a vacancy should reach the recruitment route rather than create a sales lead.
If a CRM lookup times out, the caller should still receive a useful response. Collect a callback request and mark the lookup unresolved internally; do not create a second record merely because the first search failed. If a transfer fails, return to the agreed fallback. If a booking tool does not confirm success, do not announce a confirmed appointment.
These are workflow decisions the buyer implements, not guarantees supplied by every platform. The smallest effective design has a clear owner and a few reliable actions, rather than asking eight questions before anybody helps the caller.
A thousand-call budget
Assume 1,000 answered inbound calls, each using three AI minutes: 3,000 AI minutes. Suppose 500 are new enquiries, 300 existing accounts and 200 support/other requests. Of the new enquiries, 400 qualify: 250 reach a rep live and 150 accept a booking or callback. These results are illustrative assumptions held constant across the comparison.
Budget 250 live handoffs at four human minutes plus fifty reviewed calls at two minutes: 1,100 minutes, or 18.33 hours. At $40/hour, that is $733.33 of human capacity. It remains even when the agent answers every inbound call.
| Platform basis at 3,000 AI minutes | Supported or assumed component calculation | Component subtotal before operating labour |
|---|---|---|
| Retell displayed configuration | 3,000 × $0.11 | $330, before configured carrier/number costs |
| ElevenAgents Pro | $99 + (3,000 − 1,238) × $0.08 | $239.96, before model and telephony |
| ConversationRelay, US local inbound | 3,000 × ($0.07 + $0.0085) + $1.15 | $236.65, before model and application hosting |
| Synthflow | $30,000 annual floor ÷ 12 | $2,500 allocation; agreement determines usage scope |
| Bland Enterprise | Configured platform, minutes and transfer purchase | Quoted E; no cheap warm-transfer total invented |
For an operating example, allocate $26.65 to carrier/number costs for Retell and ElevenAgents, using the same illustrative US local rate basis as the Twilio row. This is a budget assumption, not a claim those providers include that carrier package. Give ElevenAgents and ConversationRelay a further $45 model budget at an assumed $0.015/minute. Retell's displayed $0.11 example already contains its chosen model; adding that model budget again would double count it.
Now assume hosted-platform integration maintenance takes two hours/month at $75/hour, or $150. The engineer-owned ConversationRelay application takes eight hours, or $600. These are workload assumptions, not vendor service fees.
- Retell: $330 + $26.65 + $733.33 + $150 = $1,239.98, or $3.10 per assumed qualified lead.
- ElevenAgents: $239.96 + $45 + $26.65 + $733.33 + $150 = $1,194.94, or $2.99 per qualified lead.
- ConversationRelay: $236.65 + $45 + $733.33 + $600 = $1,614.98, or $4.04 per qualified lead.
- Synthflow floor plus the same human and maintenance allocations: $2,500 + $733.33 + $150 = $3,383.33, or $8.46 per qualified lead, before interpreting its agreed allowances and other charges.
- Bland Enterprise: E + $883.33 for the same assumed human/maintenance work, plus anything outside the configured quote.
The recommendation is not a forty-five-dollar platform race. If your Twilio application already exists and only needs two monthly maintenance hours, its example falls by $450 to $1,164.98. Existing engineering changes the practical winner more than a small difference in hosted rates.
Likewise, if only 250 new enquiries qualify, Retell's same example is $4.96 per qualified lead, not $3.10. Better routing and sensible qualification affect economics more than increasing conversation length. One-time implementation, extra transfer legs, peak/burst capacity, optional services, taxes and some hosting costs remain outside these examples; their basis should be added to the selected design without inventing identical totals across providers.
What makes the recommendation change
Existing CRM state favours a platform whose live tool pattern your team can operate. Retell's connected tools reduce integration work for supported providers; other platforms can call your backend. A correct lookup is more useful than a long generic qualification script.
Peak demand can change capacity and price. Three thousand monthly minutes does not describe how many callers arrive simultaneously after an event. ElevenAgents publishes concurrency and burst terms; the other selected purchase must carry the expected peak. Route overflow to an owned callback queue rather than letting a warm caller disappear.
Handoff requirements can change the qualifying plan. Bland's Enterprise warm transfer and ElevenAgents' transport-specific operator message are concrete examples. A direct transfer without context is a different experience from briefing the rep first.
Internal ownership changes the economics. A team with engineers and existing telephony can make ConversationRelay economical. A team paying for implementation and operational support may prefer Synthflow despite its larger commitment.
Which platform to buy
Start with Retell AI for a CRM-aware inbound build where the team wants platform tooling and can own the routing rules. Choose ElevenAgents when its voice/workflow tools and supported telephony handoff fit the deployment better.
Choose Synthflow for enterprise visual operations and implementation value. Choose Bland Enterprise for structured pathways and the documented warm-briefing workflow. Choose ConversationRelay when your engineers own the application and can keep it reliable.
Use the signal-based outbound playbook to connect qualification to a useful sales next step. The buying outcome is a caller reaching the right owner with enough context to continue, rather than an agent completing the longest possible conversation.
FAQ
Does every caller need sales qualification?
No. Existing customers, current deals and non-sales enquiries should route appropriately. Qualify new demand after establishing the purpose.
Can these agents read the CRM during a call?
Yes, through the documented tool and application patterns described here. The exact operation still depends on the configured provider, endpoint, permissions and record mapping.
Is Synthflow still a low-cost self-service minute purchase?
Its current new-customer commercial page starts at $30,000 annually. This guide uses that basis rather than old self-service pricing.
Is Bland warm transfer included on Start?
No. The current warm-transfer documentation requires Enterprise.
Which is cheapest overall?
That depends heavily on model/telephony scope and existing integration work. The worked example shows how an engineer-owned application can cost more to operate despite a low relay rate, and how an existing application changes that result.
