What this page covers
- Do we have to rebuild our phone system first?
- How calls reach the agent from your current phone system
- When a SIP trunk or BYOC connection makes sense
- What happens when a caller needs a person
- What changes in the phone tree, and what stays
- How a phone tree replacement works in steps
- Which phone systems Flexbone works with
- How a Flexbone deployment works, step by step
You do not need to rebuild your phone system to run an AI voice agent. The agent is one more destination your phone system can already route to. That destination is an outside number, a SIP endpoint, or the target of a menu option or queue rule. A practice starts by sending one slice of calls to it. That slice is often after-hours calls, overflow from the front-desk queue, or a dedicated line for refills or Spanish-speaking patients. The carrier, the numbers, the desk phones and the rest of the phone tree stay as they are.
Do We Have to Rebuild Our Phone System First?
No. One critical-access hospital told us that other vendors required a telecom rebuild before an AI agent took a single call. Its own phone upgrade had been deferred by a year. Flexbone starts from the other end. The first agent runs on the phone system you have today, through routing settings your administrator already uses. A rebuild is a separate decision, made later, and only if the current system blocks something the practice wants.
A pilot needs three things from the phone side. The agent needs a phone number, and one routing change sends the chosen calls to it. The third is a list of the numbers or queues the agent transfers callers to. Reversing the routing change takes minutes, because the rule points back at its old destination.
Get an outside read on your phone system integrations workflow
In 30 minutes we map your current volume, the payers and systems involved, where staff time goes, and the highest-ROI calls and follow-ups Flexbone can take off your team first, scoped to the work you actually run.
How Do Calls Reach the Agent From Our Current Phone System?
The agent is one more destination your phone system can route to. A deployment uses one of three routes, and a pilot usually starts with the first.
| Route | How It Works | Where It Fits |
|---|---|---|
| Call forwarding | A forwarding rule sends chosen calls to the agent's number over the public phone network. The rule can apply after hours, on no answer, or to one dialed number. | Pilots, after-hours coverage, and a dedicated refill or Spanish line |
| Menu or queue rule | One IVR menu option, or one queue setting such as a maximum wait, points at the agent. Other options and queues keep their destinations. | Handing over one call type, or front-desk overflow |
| SIP trunk or BYOC | Your phone system or carrier delivers calls to the agent over SIP, with the caller ID and dialed number attached. | Higher volume, or an agent that sits inside the phone system as an endpoint |
Forwarding keeps the numbers with your carrier, so nothing is ported. A practice can also give the agent its own dialed number (DID), such as a refill or Spanish line. The main number stays as it is.
How the call arrives is separate from what the agent does once it answers. It books and reschedules visits, takes refill requests to the clinical staff queue, and answers coverage questions after an eligibility check. Where the EHR connection is set up, results are written to eClinicalWorks, athenahealth or Tebra through a vendor API or a browser automation.
When Does a SIP Trunk or BYOC Connection Make Sense?
A SIP trunk connects your phone system or session border controller to the agent platform directly, instead of through a forward across the public network. It carries the audio along with call data such as the caller ID and the dialed number. It also handles higher volumes. Bring your own carrier (BYOC) is the variant where your current carrier delivers the calls over SIP, so the numbers stay on your carrier account.
Both take network and firewall work on your side, so they usually follow a forwarding pilot rather than start one. They fit a contact center with steady high volume, or a team that wants the agent to appear as an extension or queue member. They also suit a phone administrator who prefers SIP routes to external forwards. On hosted VoIP and on-premises PBXs alike, the routing primitives are the ones your administrator already manages.
What Happens When a Caller Needs a Person?
The agent transfers the call to your staff with the context attached. Before it connects the caller, it tells the staff member who is calling and what has already been checked. The full transcript is saved with the call. Clinical questions, billing disputes and anything outside the rules built from your SOPs go to a person this way.
There are two ways to move the call. A PSTN transfer dials the direct number or queue number you choose and connects the caller to it. That works with any phone system that accepts outside calls. A SIP REFER transfer applies when the call arrived over SIP. The agent asks your phone system to move the call to an extension or queue, and the agent leaves the call path. Where the phone system can show a screen pop, the summary can also appear on the staff member's screen. In both cases the caller does not start over.
What Changes in Our Phone Tree, and What Stays the Same?
One routing destination changes for each call type you hand to the agent: a forwarding target, a menu option or a queue rule. Everything else stays as it is. That includes the carrier and the numbers, desk phones and softphone apps, extensions, ring groups, queues and voicemail. It also includes the menu branches the agent does not take. Staff call recording settings do not change. The agent records and transcribes its own calls, with a recording disclosure your counsel approves.
The phone tree also shows the audit where calls go wrong. One practice described a nine-option menu that misrouted about 60 calls a day to its billing staff. In a case like that, the first change is to give the agent the menu branch callers pick by mistake. The Voice Room audit finds those patterns in your recordings before anything is built. Flexbone has pulled call recordings for audits from GoTo Connect, Talkdesk and VoIPX.
Have Flexbone run your phone system integrations workflow on real cases
A complimentary six-week pilot once the BAA is signed, with a weekly review of what the agents handled, which results were written back to your system, and the hours your team got back.
How Does a Phone Tree Replacement Work?
In steps, with the old tree kept as a fallback. The agent first takes one branch, such as scheduling or after hours, while the rest of the menu runs as before. Transcripts from those calls show where the agent resolves the request and where it hands off. As it proves out on live calls, more branches route to it and the menu gets shorter.
The end state can be inversion. The agent answers first, in natural language, and resolves or routes each call. The old tree sits behind it for the calls the agent sends on. Spanish-speaking callers get the same path in Spanish instead of a separate menu option. For a fuller comparison of the two models, see voice agents versus IVR.
Which Phone Systems Does Flexbone Work With?
Flexbone works with hosted VoIP services, on-premises PBXs and contact center platforms. Each of them can send a call to an outside number, and many can also route to a SIP endpoint. These guides cover the settings on specific platforms: RingCentral, GoTo Connect, Nextiva and Microsoft Teams Phone.
Practices also tell us they run Cisco, Zoom Phone, Dialpad, Vonage, Mitel, CallTower or Twilio. Contact centers run platforms such as Talkdesk, Five9, Genesys Cloud, NICE CXone, Amazon Connect and 8x8. The same three routes apply to these, with setting names that differ by platform. The general routing patterns are covered in how voice agents connect to phone systems.
How Does a Flexbone Deployment Work?
Every deployment starts from the work as your team does it today, not from a product demo. The steps below are the same whether the first agent answers phones, checks eligibility, or files incoming documents.
- Audit first, then automate. We listen to a sample of your calls and read the work queues, then agree with your team on which work to automate first. The audit is complimentary.
- Forward-deployed engineers map your SOPs. An engineer sits with the staff who do the work and turns your written and unwritten rules into agent logic.
- Agents act on your rules, and exceptions go to a person. Each step follows logic built from your SOPs. A case outside those rules, or one the agent is not confident about, goes to your team with the context attached.
- Every action leaves a record. A check in a web application saves a screenshot, a call saves a transcript, and results are written back where your team already looks.
- Start with one or two agents on the systems you already run. The agents work in your EHR, practice management system, phone system, and the web applications your staff use, as they are configured today, so nothing is migrated or rebuilt.
- The same team stays after go-live. We monitor results, tune the agents as payers and workflows change, and add agents as new work is found.
The engagement model is described in more detail in forward-deployed engineering for healthcare contact centers.