Teams Insider

Is AI the new UC? Enterprise Connect thoughts with Kevin Kieller

Kevin Kieller, co-founder and lead analyst at enableUC and lead of the BC Strategies team, joins Tom Arbuthnot straight after Enterprise Connect 2024 to unpack whether AI is the industry's next big pivot after PBX and UC. The pair cover the terminology IT pros now need, where Copilot's grounding dif

Prefer audio?

Who should listen: Microsoft 365 and Teams service owners building the business case for Copilot, plus UC consultants and adoption leads who need to explain the differences between Copilot surfaces and set realistic expectations with end users.

Guest: Kevin Kieller — Co-Founder and Lead Analyst, enableUC

Many thanks to Pure IP, sponsor of this episode.

Key insights

AI dominated Enterprise Connect 2024 to the point that Kieller counted 73 of the 171 sessions with AI in the session description — if a vendor had no AI story, it was a problem. ▶ 2:36
AI is repeating the OCS/Lync pattern: just as Exchange and AD admins were forced to learn jitter, packet loss and round-trip time, IT pros now need working knowledge of transformer models, probabilistic output, temperature, grounding and prompt engineering — not data science, but enough to know how the car runs. ▶ 5:41
Meeting and call recap are grounded only on the meeting transcript today, so two things have to go right — speech-to-text then summarisation. Chat isn't yet included, which matters because decisions and actions are often cemented in the meeting chat; with chat there's only one failure point because the user typed it. ▶ 7:44
The Copilot chat surface in Teams is currently the only genuine AI assistant across the UCaaS platforms, because it visibly reaches into email, chat, previous meeting transcripts and the graph — whereas in-meeting or in-chat Copilot sees only that isolated context. Users assume all the Copilots behave identically, and they don't. ▶ 10:46
AI-generated meeting summaries can't be edited by design, while every other vendor allows it. Kieller expects Microsoft to relent and to track who corrected a summary, so the editor becomes accountable for it in the same way as human-sent minutes. ▶ 14:53
Frame the cost as $360 per user per year on an annual commit rather than $30 a month, then budget potentially the same again for adoption and training — without it, most organisations won't see a return. The new Copilot dashboard analytics let you spot users who never run a summary and reassign their licence. ▶ 15:54
A practical prompt from a senior IT leader with overlapping meetings: ask the transcript "was there any conflict in the meeting?" — if no, move on; if yes, ask Copilot for a table of the contentious items with the pros and cons of each position before deciding whether to intervene. ▶ 17:58
AI is pulling the market back towards suites and away from best-of-breed, because broader data access makes a better assistant. Kieller's ask of vendors for 2025 is quantitative usage data and real use cases rather than "my summaries are better than yours" — you can't fairly test a probabilistic model when speech-to-text is skewed by accent, gender and room device. ▶ 21:04
Contact centre is the early indicator for AI value because every agent metric is already tracked; UC owners should look at first-call resolution and average handle time and work out which equivalents map to knowledge workers. ▶ 23:07