BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
DTSTART:20270314T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20261101T010000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260914T220159Z
UID:695EB086-D03B-451D-A974-E40D2613613C
DTSTART;TZID=America/New_York:20261208T180000
DTEND;TZID=America/New_York:20261208T200000
DESCRIPTION:[Shankar_137x140]\nShankar Krishna	Voice AI has crossed a thres
 hold\, and the shift changes where engineering effort should go. Systems t
 hat once only transcribed speech now answer customers\, resolve disputes\,
  and take actions across enterprise systems without a person in the loop. 
 Yet most teams still pour their effort into the speech model\, which has b
 ecome a component you buy from a few providers. The real engineering\, and
  the real risk\, lives in the agent layer beneath the voice. That layer is
  a loop that reasons\, calls tools\, and acts under a hard latency budget 
 measured in milliseconds. Drawing on years of building voice AI at enterpr
 ise scale\, I will cover the anatomy of a modern voice agent. I will show 
 where latency is spent across a single turn\, and the failure modes that c
 atch teams late\, including barge in and accent drift. I will close with a
  practical playbook for what to own\, what to buy\, how to evaluate from r
 eal calls\, and how to design the handoff to humans. Attendees will be abl
 e to judge where voice agents belong\, and where they do not.\n\nCo-sponso
 red by: IEEE Boston AI Group\n\nSpeaker(s): Shankar Krishnan\, \n\nRoom: N
 ERD/1250\, Abigail Adams\, Microsoft NERD Center\, Kendall Square\, 1 Memo
 rial Drive\, Cambridge\, Massachusetts\, United States\, 02142\, Virtual: 
 https://events.vtools.ieee.org/m/575329
LOCATION:Room: NERD/1250\, Abigail Adams\, Microsoft NERD Center\, Kendall 
 Square\, 1 Memorial Drive\, Cambridge\, Massachusetts\, United States\, 02
 142\, Virtual: https://events.vtools.ieee.org/m/575329
ORGANIZER:info@voptronix.com
SEQUENCE:72
SUMMARY:The Hard Part of Voice AI Is Not the Voice: Engineering Voice Agent
 s That Take Action
URL;VALUE=URI:https://events.vtools.ieee.org/m/575329
X-ALT-DESC:Description: <br /><table style="border-collapse: collapse\; wid
 th: 100.119%\; border-width: 0px\;" border="1"><colgroup><col style="width
 : 19.4411%\;"><col style="width: 80.5589%\;"></colgroup>\n<tbody>\n<tr>\n<
 td style="text-align: center\; border-width: 0px\;"><img src="https://even
 ts.vtools.ieee.org/vtools_ui/media/display/0611d2af-2df7-4548-a2be-24a0fde
 a65ea" alt="Shankar_137x140" width="137" height="140"><br><strong>Shankar 
 Krishna</strong></td>\n<td style="border-width: 0px\;">Voice AI has crosse
 d a threshold\, and the shift changes where engineering effort should go. 
 Systems that once only transcribed speech now answer customers\, resolve d
 isputes\, and take actions across enterprise systems without a person in t
 he loop. Yet most teams still pour their effort into the speech model\, wh
 ich has become a component you buy from a few providers. The real engineer
 ing\, and the real risk\, lives in the agent layer beneath the voice. That
  layer is a loop that reasons\, calls tools\, and acts under a hard latenc
 y budget measured in milliseconds. Drawing on years of building voice AI a
 t enterprise scale\, I will cover the anatomy of a modern voice agent. I w
 ill show where latency is spent across a single turn\, and the failure mod
 es that catch teams late\, including barge in and accent drift. I will clo
 se with a practical playbook for what to own\, what to buy\, how to evalua
 te from real calls\, and how to design the handoff to humans. Attendees wi
 ll be able to judge where voice agents belong\, and where they do not.</td
 >\n</tr>\n</tbody>\n</table>\n<p>&nbsp\;</p>
END:VEVENT
END:VCALENDAR