At a recent Comms vNext session, Momentum’s Mark Vale (Enterprise Architect, Microsoft Teams Phone Expert, Microsoft MVP) and Brandon Bradley (VP, Modern Work & UC Innovation) made the case for what should be obvious by now and somehow isn’t: voice deserves the same workflow treatment as every other business process. The future state of voice hinges on how AI enhances and replaces traditional workflows and call flows, and their talk, Why doesn’t your call flow? Bringing workflows into the modern call flow with Power Automate and voice AI, walked through the practical mechanics of getting there.
The full recap is below, broken down by section. The video has the live demos and audience Q&A.
Voice is the last workflow most companies haven’t automated
Quoting runs through CRM automation. Ticketing fires alerts and routes by SLA. Procurement, onboarding, and approvals all sit inside structured workflows. Then a call comes in, and someone manually opens three tabs to figure out who is on the line.
The premise is straightforward: treat the call flow like every other workflow in the business. Orchestrated, contextual, and connected to the systems that already hold the answer. Here’s what that looks like in practice.
Power Automate as the orchestration layer
Power Automate is the Microsoft automation platform that predates newer tools like N8N. It’s the plumbing between platforms, a low-code/no-code canvas that connects Microsoft services, third-party CRMs, AI services, and compliance platforms through certified connectors and custom HTTP endpoints.
The licensing math matters more than most people realize. The free tier handles a surprising amount. A single premium account at $15 per month supports thousands of flows and processes up to 40,000 requests per day across multiple SharePoint sites and CRMs. Most organizations encounter Power Automate through trivial demos like “move an email attachment to SharePoint” and miss the real value, which is orchestration at scale.
Every flow follows the same three building blocks:
- Trigger: the event that starts the flow (webhook, schedule, file upload, message received)
- Conditions: the logic branches (if/else, true/false)
- Actions: what happens next (post a message, send an SMS, write to a CRM, call an API)
That structure is what makes the call flow examples below possible.
Three workflows already running in production
Three real deployments illustrate the range.
A school district incident response system
A school district in Colorado sends drivers out at 2 to 3 a.m. during winter to monitor roads for closure decisions. The traditional process is phone tag across faculty, parents, and emergency staff.
The workflow replaces it. A single flow can be triggered from an Algo paging device, a Teams mobile app, an inbound phone call, a Teams message, or an SMS. The trigger fires via webhook, and the flow does the rest:
- Captures the current time to schedule the incident
- Branches on event severity (emergency versus alert)
- Posts an Adaptive Card to a Teams channel for faculty action
- Sends SMS messages to faculty, parents, and students through an SMS gateway
- Creates a Teams incident bridge with a dynamic title pulled from the Algo event
- Provides one-click actions on the Adaptive Card: join the bridge, dial emergency services, open the paging talk path
The entire response runs through a single Teams pane of glass instead of dozens of disconnected phone calls.
A best practice worth flagging from the session: initialize variables at the top of the flow before any logic runs. Update one string at the top and the whole flow updates. Hard-code data into individual actions and you’ll be editing 80 nodes to change one phone number.
Customer appointment reminders by SMS
A common request. The CRM doesn’t natively support SMS, or the team uses a separate SMS provider. Power Automate bridges them.
The flow runs on recurrence (every minute, hour, or batched daily), looks 24 hours ahead, queries Salesforce for appointments in the window, and sends a personalized SMS using the contact’s first name, appointment subject, and time. Set it and forget it.
AI-powered document processing
A file lands in SharePoint, Azure OpenAI analyzes it, a human in the loop validates the AI output, and the document is processed. Covered briefly in the session due to time, but the pattern (storage trigger plus AI analysis plus human approval) is one of the most common production Power Automate workflows in the wild.
Tabs for Calling, the agent never swivel-chairs again
Tabs for Calling shifts the focus to the call itself. It’s a Microsoft Teams capability that lifts an app directly into the call dialogue. Instead of a screen pop opening a browser, the CRM or ticketing interface lives inside the call. The agent never switches windows.
Two scenarios show what changes.
Unknown caller
An inbound call arrives. The dialogue opens. Salesforce is queried for the caller and returns no match. The agent captures the caller’s name, email, and company directly in the dialogue. On hang up, Power Automate holds the agent’s Teams status as “in a call” until wrap-up completes. When the agent submits the disposition, the flow sets them available again, with a configurable forced-available timer (default five minutes). Salesforce auto-creates the account, contact, call log, and disposition record in the background.
Known caller
The CRM is dipped instantly. The caller is identified, pending appointments and history surface in the dialogue, the agent sets a disposition (for example, “callback requested”), and Power Automate writes the result back to the CRM. A post-call survey dispatches automatically.
Agent desktop and reporting layer
The same Teams app functions as a lightweight agent desktop. Agents can manually set wrap-up time, with status expiration visible on their profile. The service account behind the flow operates with delegated permissions, so administrators keep central control.
Because the entire interaction runs through Power Automate, structured BI follows naturally:
- Track agent availability and wrap-up duration
- Identify outliers (agents consistently exceeding wrap-up time)
- Integrate Azure OpenAI to summarize and score agent performance
- Log everything into spreadsheets, databases, or dashboards
Voice AI and the future of call queues
Start with what customers actually want. They don’t want AI. They want the results of AI. The voice industry is on the back foot adopting voice AI because customers have been trained to hate it. Traditional IVR bots run on rails, and the customer’s job becomes navigating the bot as quickly as possible to reach a human.
Reimagining the inbound flow
Strip away the IVR gateways. Drop the caller into the queue immediately and offer an AI agent that helps while they wait. The reframe sounds like this: “Thank you for calling. You’re in the queue and will be heard by a human. While you wait, can my AI agent help with whatever you’re calling about?”
Nine times out of ten, callers say yes. The alternative is hold music. The AI agent queries knowledge base articles, attempts to resolve the issue, and often the caller never needs to formally enter the queue. The customers who do reach a human become higher priority because their issue genuinely requires it.
Voice agent demo
Here’s what a complete inbound resolution looks like:
- Caller ID is matched silently against the ITSM system
- The caller is addressed by first name automatically
- The knowledge base (Desk 365 in this example) is searched for the caller’s question
- The relevant article is read back to the caller
- The caller confirms resolution, and the ticket is auto-logged and closed with the KB article attached
The same voice agent handles natural language well enough that one frontline agent can route across sales, support, and accounting based on intent. No need to build a separate agent per department, train pathways inside a single agent.
Authentication guardrails
Before disclosing anything related to PII, the caller has to be verified. Otherwise anyone calling your main number could request ticket updates by guessing IDs. The flow:
- Caller is identified via CRM and mobile number lookup
- A one-time passcode is sent via SMS
- The caller reads the code back to the AI agent
- On verification, the caller can update, log, or close cases
- Ticket ownership is cross-referenced. If the caller asks about a ticket they don’t own, the agent denies the request and lists their actual open tickets instead
AI-to-human handover with full context
When transfer becomes necessary, the AI agent hands the human everything it collected:
- Caller sentiment (happy, frustrated, angry, neutral)
- Call summary
- Full transcript between the AI agent and the caller
- Relevant ticket numbers
- Call history
The AI also updates the ticket before transfer, finds the right department, checks Teams presence for agent availability, and routes to a free agent. The human picks up with full context instead of asking the caller to start over.
Post-call workflows and reporting
Once you have full call context in structured form, the post-call workflow compounds:
- Post-call wrap-up and surveys logged automatically
- Resolved support calls converted into outbound sales leads when relevant
- Outbound meetings booked directly from the wrap-up flow
- Dashboard visibility into calls solved by AI versus transferred
- Agent scoring based on sentiment outcomes (positive points for happy resolutions, negative for angry ones)
- Customer feedback aggregated automatically
When the local knowledge base lacks an answer, the agent can fall back to verified external sources like learn.microsoft.com and read back possible resolutions.
Queue elimination versus reduction
Even if a voice agent resolves 80% of calls, high-volume environments will still need queues. In lower-volume scenarios, the queue disappears entirely. The voice bot hunts available agents directly via Teams presence and routes the call.
Orchestration is the missing layer
The point of the session wasn’t to pitch a product. It was to show the art of the possible. The industry is moving toward automated workflows that bring efficiency forward, and voice is one of the last places this hasn’t fully happened yet.
There’s no shortage of tools. Voice AI platforms like LiveKit, CRMs, ticketing systems, knowledge bases. All available, most of them excellent. What’s missing is the orchestration layer that ties them together. Whether you use Power Automate, the broader Power Platform, N8N, or another tool, orchestration is what makes the call flow contextual instead of disconnected.
An audience question in the Q&A raised why more customers aren’t using Power Automate already. The answer is education. Most teams meet the tool through trivial demos and never see what it does at scale. Once they encounter workflows like the ones above, adoption follows.
Thanks to Comms vNext for hosting, and to Mark Vale and Brandon Bradley for the work that went into building, demoing, and breaking this session down on stage.
Workflows belong in the call flow, not next to it
Momentum brings it all together. From Microsoft Teams Phone deployments and call flow design to Power Automate orchestration and voice AI integrations, we connect the systems that hold your data to the call where it’s needed. One provider, one bill, one team.
