The Role of AI in Microsoft Teams Phone and Modern Work, Real-world Thoughts with Rohan Milne

Prefer audio?

Who should listen: Microsoft 365 and voice service owners planning Teams Phone migrations off legacy PBXs, MSPs and telcos building Operator Connect or direct routing practices, and anyone evaluating Teams contact centre options or setting realistic expectations for AI in voice workloads.

Guest: Rohan Milne — CEO and Founder, Switch Connect

Rohan Milne is CEO and founder of Switch Connect, an Australian Operator Connect provider that also runs a wholesale/distribution business and consults to MSPs adding Teams Phone. He and Tom Arbuthnot cover shared calling's impact on carrier margin models, why voice expertise still decides Teams Phone project outcomes, the reality behind "Teams certified" contact centres, and where AI genuinely earns its keep versus where the hype outruns it.

Many thanks to Ribbon, sponsor of this episode.

Key insights

• Shared calling breaks the carrier per-user-per-month economics that many Operator Connect business cases are built on. Carriers that only deliver the voice component have nowhere else to recover the margin, so those selling per-channel from day one (as Switch Connect has) see it as a natural fit — particularly in education and large enterprise where handsets need dial tone but not a DID. ▶ 2:02
• Adding dial tone is no longer the differentiator: a Ribbon SBC Swe Lite with trial licensing, a wizard and a YouTube video gets direct routing live in about half an hour with no voice background. The value now sits in migrating complex use cases off Call Manager, Mitel and Avaya, which requires real call-flow knowledge. ▶ 7:07
• Requirements language is the biggest source of scope churn. One project was re-quoted eight times in a week across four platforms because the customer's "attendant console" turned out to be a wallboard showing calls in queue — reporting and analytics, not an operator console. Same pattern with "I need a contact centre" five or six times a week. ▶ 9:41
• Frame Teams as a collaboration platform with a PBX overlay, whereas Webex and similar are a PBX with a collaboration overlay. Customers listing 5,000 legacy feature sets typically use about nine of them, all of which Teams covers. A practical trick: turn a call queue into a hunt group by uploading a ringtone as the music-on-hold file, so callers do not think they have been parked. ▶ 11:15
• Of the 24 Teams-certified contact centres, roughly 18 fall straight off the list when a tender specifies "Teams native". SIP-integrated products like Genesys cannot be sold in an Operator Connect motion — and arguably placing them in front of Teams breaches the wording of an OC contract — plus the end-user experience is poor for the three large Genesys customers Switch Connect supports. ▶ 13:46
• Teams is not right for every persona. Shared devices are where it falls down — nurses on a ward with cordless handsets, or a GP practice that parks a call on line two and pages the doctor with no BLF or park buttons on the phone. Hybrid Teams plus UCaaS deployments are the answer in healthcare, with Teams for doctors, admins and specialists. ▶ 16:20
• Australia is the most densely E5-licensed country globally, yet Teams calling adoption inside Queensland government's five-year E5 agreement sits at around 12%. The blockers are contractual and political, not technical: bundled internet/SIP contracts, carriers earning more from SIP trunking than Teams, and internal empires of ten people who justify their roles on PBX black magic. Switch Connect's answer is Operator Connect with bring-your-own-carriage so the incumbent Telstra SIP still carries the calls. ▶ 19:24
• Real-time call QA and sentiment analytics pay for themselves twice: they replace the 10% of contact centre headcount doing manual one-in-ten call sampling and flag compliance failures (e.g. Australia's RG236 identity verification obligation) instantly, and keyword spotting on help-desk calls turns unmet customer intent into sales leads for the account manager. ▶ 24:30
• AI today is far stronger at summarising existing context than generating content. ChatGPT-written marketing copy read three days later often does not hold up, and a construction test found every plastering step correct but four in the wrong order — technically right, practically useless. Bing's browser AI is recommended for knowledge-base articles because it back-links to learn.microsoft.com so the source can be vetted. ▶ 28:06

Insights summarised by AI from the episode transcript, reviewed by the Empowering.Cloud team.

Listen: Apple Podcasts · Spotify · Other platforms