Latest / Omni Talk Retail / AWS's David Dorf On Why LLMs Have Peaked & What's Coming Next In 2026

Transcript

AWS's David Dorf On Why LLMs Have Peaked & What's Coming Next In 2026 · Omni Talk Retail

• 0:00
  Foreign.
• 0:08
  Joining
• 0:08
  us
• 0:08
  once
• 0:08
  again
• 0:09
  for
• 0:09
  five
• 0:09
  insightful
• 0:10
  minutes
• 0:10
  is
• 0:10
  one
• 0:10
  of
• 0:11
  our
• 0:11
  favorite
• 0:11
  recurring
• 0:12
  guests,
• 0:12
  David
• 0:13
  Dorff.
• 0:13
  David
• 0:14
  is
• 0:14
  the
• 0:14
  head
• 0:14
  of
• 0:14
  Retail
• 0:15
  Industry
• 0:15
  Solutions
• 0:15
  at
• 0:16
  aws
• 0:16
  and
• 0:16
  today
• 0:17
  he
• 0:17
  is
• 0:17
  going
• 0:17
  to
• 0:17
  regale
• 0:18
  us
• 0:18
  with
• 0:18
  his
• 0:18
  AI
• 0:19
  predictions
• 0:19
  for
• 0:20
  2026.
• 0:21
  And
• 0:21
  if
• 0:21
  we
• 0:21
  go
• 0:21
  a
• 0:22
  little
• 0:22
  long
• 0:22
  today,
• 0:23
  I
• 0:23
  will
• 0:23
  take
• 0:23
  full
• 0:24
  responsibility
• 0:24
  because
• 0:25
  you
• 0:25
  will
• 0:25
  want
• 0:25
  to
• 0:25
  hear
• 0:26
  what
• 0:26
  what
• 0:26
  David
• 0:27
  has
• 0:27
  to
• 0:27
  say.
• 0:28
  David,
• 0:28
  welcome
• 0:28
  back
• 0:29
  to
• 0:29
  Omni
• 0:29
  Talk
• 0:29
  and
• 0:29
  let's
• 0:30
  get
• 0:30
  started
• 0:30
  with
• 0:30
  this.
• 0:31
  You
• 0:31
  said
• 0:31
  in
• 0:32
  a
• 0:32
  recent
• 0:32
  article
• 0:32
  that
• 0:33
  2025
• 0:34
  was
• 0:34
  the
• 0:34
  year
• 0:35
  of
• 0:35
  LLM
• 0:36
  overhype.
• 0:37
  To
• 0:37
  paraphrase,
• 0:38
  that's
• 0:38
  a
• 0:38
  bold
• 0:39
  statement
• 0:39
  given
• 0:39
  all
• 0:40
  the
• 0:40
  noise
• 0:40
  in
• 0:40
  the
• 0:40
  market,
• 0:40
  and
• 0:41
  also
• 0:41
  one
• 0:41
  that
• 0:41
  I
• 0:42
  happen
• 0:42
  to
• 0:42
  agree
• 0:42
  with
• 0:42
  for
• 0:42
  those
• 0:43
  that
• 0:43
  know
• 0:43
  me
• 0:43
  well.
• 0:43
  So
• 0:44
  what
• 0:44
  makes
• 0:44
  you
• 0:44
  think
• 0:44
  we've
• 0:44
  hit
• 0:45
  a
• 0:45
  plateau
• 0:45
  and
• 0:45
  where
• 0:46
  do
• 0:46
  you
• 0:46
  see
• 0:46
  real
• 0:47
  gains
• 0:47
  happening
• 0:47
  from
• 0:47
  here
• 0:48
  on
• 0:48
  out?
• 0:48
  So
• 0:49
  it
• 0:50
  was
• 0:50
  just
• 0:50
  a
• 0:50
  couple
• 0:50
  years
• 0:50
  ago
• 0:51
  people
• 0:51
  were
• 0:51
  talking
• 0:51
  about
• 0:52
  artificial
• 0:53
  general
• 0:53
  intelligence
• 0:54
  being
• 0:54
  right
• 0:54
  around
• 0:54
  the
• 0:54
  corner
• 0:55
  and
• 0:55
  I
• 0:55
  think
• 0:55
  we
• 0:55
  got
• 0:55
  a
• 0:56
  little
• 0:56
  bit
• 0:56
  over
• 0:56
  our
• 0:56
  skis
• 0:57
  on
• 0:57
  that.
• 0:58
  LLMs
• 0:58
  are
• 0:59
  tremendously
• 0:59
  powerful,
• 1:00
  but
• 1:01
  the
• 1:01
  idea
• 1:01
  that
• 1:01
  we
• 1:01
  can
• 1:02
  continue
• 1:02
  to
• 1:02
  just
• 1:02
  scale
• 1:02
  up
• 1:03
  the
• 1:03
  data
• 1:03
  doesn't
• 1:04
  actually
• 1:04
  lead
• 1:04
  to
• 1:05
  better
• 1:05
  results.
• 1:06
  So
• 1:06
  one
• 1:06
  of
• 1:07
  the
• 1:07
  things
• 1:07
  is
• 1:07
  that
• 1:07
  LLMs
• 1:08
  will
• 1:08
  incrementally
• 1:09
  improve,
• 1:10
  but
• 1:10
  there
• 1:10
  won't
• 1:11
  be
• 1:11
  exponential
• 1:12
  gains
• 1:12
  with
• 1:13
  the
• 1:13
  current
• 1:13
  technology.
• 1:14
  There
• 1:14
  needs
• 1:15
  to
• 1:15
  be
• 1:15
  a
• 1:15
  new
• 1:16
  architecture
• 1:16
  like
• 1:17
  the
• 1:17
  Google
• 1:17
  Transformer
• 1:18
  that
• 1:18
  started
• 1:18
  the
• 1:18
  whole
• 1:19
  LLM
• 1:19
  thing.
• 1:20
  Something
• 1:21
  like
• 1:21
  maybe
• 1:21
  what's
• 1:21
  going
• 1:22
  on
• 1:22
  with
• 1:22
  world
• 1:22
  models.
• 1:23
  I'm
• 1:23
  not
• 1:23
  a
• 1:23
  scientist
• 1:24
  so
• 1:24
  I
• 1:24
  don't
• 1:25
  go
• 1:25
  that
• 1:25
  deep,
• 1:25
  but
• 1:25
  there's
• 1:26
  some
• 1:26
  cool
• 1:26
  stuff
• 1:26
  brewing
• 1:27
  out
• 1:27
  there
• 1:27
  that
• 1:27
  could
• 1:27
  help
• 1:27
  us
• 1:28
  in
• 1:28
  the
• 1:28
  long
• 1:28
  run.
• 1:29
  But
• 1:29
  in
• 1:29
  the
• 1:29
  meantime
• 1:30
  I
• 1:30
  think
• 1:31
  it's
• 1:31
  better
• 1:32
  to
• 1:32
  focus
• 1:32
  on
• 1:32
  answers
• 1:33
  at
• 1:33
  lower
• 1:33
  costs.
• 1:34
  And
• 1:35
  so
• 1:35
  we
• 1:35
  need
• 1:35
  to
• 1:35
  think
• 1:35
  about
• 1:36
  things
• 1:36
  like
• 1:36
  domain
• 1:37
  specific
• 1:37
  models.
• 1:38
  So
• 1:38
  these
• 1:38
  are
• 1:39
  models
• 1:39
  that
• 1:39
  have
• 1:39
  a
• 1:40
  special
• 1:40
  purpose.
• 1:41
  We've
• 1:41
  already
• 1:41
  seen
• 1:41
  some
• 1:42
  of
• 1:42
  these
• 1:42
  in
• 1:42
  like
• 1:43
  finance
• 1:43
  and
• 1:43
  healthcare
• 1:44
  and
• 1:44
  I'm
• 1:45
  waiting
• 1:45
  for
• 1:45
  some,
• 1:46
  some
• 1:46
  of
• 1:46
  these
• 1:46
  to
• 1:46
  come
• 1:46
  out
• 1:47
  for
• 1:47
  retail.
• 1:48
  One
• 1:48
  of
• 1:48
  the
• 1:48
  things
• 1:48
  that
• 1:49
  we
• 1:49
  just
• 1:49
  announced
• 1:49
  at
• 1:49
  our
• 1:50
  RE
• 1:50
  Invent
• 1:50
  conference
• 1:50
  last
• 1:51
  week
• 1:51
  was
• 1:51
  this
• 1:52
  product
• 1:52
  called
• 1:52
  novaforge.
• 1:54
  It's
• 1:54
  pretty
• 1:54
  unique.
• 1:54
  We
• 1:55
  take
• 1:55
  a
• 1:56
  half
• 1:56
  trained
• 1:56
  frontier
• 1:57
  model
• 1:58
  and
• 1:58
  hand
• 1:58
  it
• 1:58
  over
• 1:58
  to
• 1:59
  the
• 1:59
  customer
• 1:59
  and
• 1:59
  then
• 1:59
  the
• 1:59
  customer
• 2:00
  can
• 2:00
  finish
• 2:00
  off
• 2:01
  the
• 2:01
  training
• 2:01
  with
• 2:01
  all
• 2:01
  their
• 2:02
  specific
• 2:02
  data.
• 2:03
  So
• 2:03
  think
• 2:03
  about
• 2:04
  using
• 2:04
  that
• 2:04
  for
• 2:04
  something
• 2:05
  like
• 2:05
  a
• 2:05
  product
• 2:05
  catalog
• 2:06
  where
• 2:06
  you
• 2:06
  have
• 2:06
  a
• 2:06
  really
• 2:07
  tight
• 2:07
  LLM
• 2:08
  that
• 2:08
  really
• 2:08
  knows
• 2:09
  product
• 2:09
  catalog
• 2:10
  super
• 2:10
  super
• 2:10
  well,
• 2:11
  knows
• 2:11
  your
• 2:11
  brand,
• 2:11
  all
• 2:11
  that
• 2:12
  good
• 2:12
  stuff.
• 2:12
  That's
• 2:13
  kind
• 2:13
  of
• 2:13
  where
• 2:13
  I
• 2:13
  think
• 2:13
  we're
• 2:14
  heading
• 2:14
  for
• 2:15
  the
• 2:15
  next
• 2:15
  thing.
• 2:16
  David,
• 2:16
  we
• 2:16
  saw
• 2:16
  this
• 2:17
  flood
• 2:17
  of
• 2:17
  agent
• 2:18
  e
• 2:18
  commerce
• 2:18
  announcements
• 2:19
  at
• 2:19
  the
• 2:19
  end
• 2:19
  of
• 2:19
  2025.
• 2:21
  I'm
• 2:21
  curious
• 2:21
  what
• 2:21
  your
• 2:22
  take
• 2:22
  was
• 2:22
  on
• 2:22
  all
• 2:23
  the
• 2:23
  announcements
• 2:23
  and
• 2:23
  what
• 2:24
  needs
• 2:24
  to
• 2:24
  happen
• 2:24
  for
• 2:25
  these
• 2:25
  answer
• 2:25
  engines
• 2:26
  to
• 2:26
  really
• 2:27
  become
• 2:27
  a
• 2:27
  meaningful
• 2:28
  channel
• 2:29
  for
• 2:29
  retailers
• 2:30
  rather
• 2:30
  than
• 2:30
  just
• 2:31
  another
• 2:31
  place
• 2:32
  that
• 2:32
  demand
• 2:32
  shifts
• 2:33
  over
• 2:33
  to.
• 2:33
  First
• 2:34
  of
• 2:34
  all,
• 2:34
  I
• 2:34
  think
• 2:34
  Flood
• 2:35
  is
• 2:35
  being
• 2:35
  kind.
• 2:35
  It
• 2:35
  was
• 2:36
  more
• 2:36
  of
• 2:36
  a
• 2:36
  tsunami,
• 2:36
  actually.
• 2:37
  Yeah.
• 2:37
  Wow.
• 2:39
  I
• 2:39
  was.
• 2:39
  I
• 2:40
  was
• 2:40
  just
• 2:40
  trying
• 2:40
  to
• 2:40
  keep
• 2:40
  track
• 2:41
  of
• 2:41
  all
• 2:41
  the
• 2:41
  things
• 2:41
  that
• 2:41
  were
• 2:42
  going
• 2:42
  on.
• 2:42
  There
• 2:42
  was
• 2:42
  so
• 2:43
  much
• 2:43
  great
• 2:43
  innovation.
• 2:44
  It
• 2:44
  was
• 2:44
  really
• 2:44
  great.
• 2:45
  But
• 2:45
  I
• 2:45
  would
• 2:46
  sum
• 2:46
  up
• 2:46
  things
• 2:46
  by
• 2:47
  saying
• 2:48
  I
• 2:48
  think
• 2:48
  it
• 2:48
  was
• 2:48
  fantastic
• 2:49
  progress,
• 2:49
  but
• 2:50
  let's
• 2:50
  put
• 2:50
  the
• 2:50
  cake
• 2:51
  back
• 2:51
  in
• 2:51
  the
• 2:51
  oven
• 2:51
  for
• 2:52
  a
• 2:52
  little
• 2:52
  longer.
• 2:53
  Some
• 2:53
  of
• 2:53
  the
• 2:53
  usages
• 2:54
  were
• 2:54
  a
• 2:54
  little
• 2:54
  bit
• 2:55
  clumsy,
• 2:56
  but
• 2:56
  kudos
• 2:56
  to
• 2:57
  Walmart
• 2:57
  and
• 2:58
  Target
• 2:58
  for
• 2:58
  getting
• 2:59
  their
• 2:59
  implementations
• 2:59
  done
• 3:00
  with
• 3:00
  Peak.
• 3:01
  We
• 3:01
  three
• 3:01
  have
• 3:02
  all
• 3:02
  been
• 3:02
  part
• 3:02
  of
• 3:02
  retail
• 3:03
  and
• 3:03
  we
• 3:04
  know
• 3:04
  November
• 3:04
  is
• 3:04
  a
• 3:05
  crushing
• 3:05
  month
• 3:06
  for
• 3:06
  them
• 3:06
  to
• 3:06
  get
• 3:07
  those
• 3:07
  things
• 3:07
  done.
• 3:07
  I
• 3:08
  really
• 3:08
  know
• 3:08
  that
• 3:08
  was
• 3:08
  tough.
• 3:09
  But
• 3:09
  I
• 3:10
  do
• 3:10
  think
• 3:10
  that
• 3:10
  agenic
• 3:11
  commerce
• 3:11
  is
• 3:11
  going
• 3:11
  to
• 3:12
  continue
• 3:12
  to
• 3:12
  make
• 3:12
  inroads.
• 3:13
  But
• 3:13
  digital
• 3:14
  and
• 3:14
  physical
• 3:14
  commerce
• 3:15
  are
• 3:15
  certainly
• 3:15
  not
• 3:16
  going
• 3:16
  anywhere.
• 3:17
  And
• 3:17
  yes,
• 3:17
  there
• 3:18
  will
• 3:18
  someday
• 3:18
  be
• 3:18
  an
• 3:19
  online
• 3:19
  retailer
• 3:19
  that
• 3:20
  has
• 3:20
  no
• 3:20
  website
• 3:21
  and
• 3:21
  just
• 3:21
  uses
• 3:22
  MCP
• 3:22
  servers
• 3:23
  to
• 3:23
  sell
• 3:23
  things,
• 3:24
  but
• 3:24
  that's
• 3:24
  an
• 3:24
  exception.
• 3:25
  Most
• 3:25
  are
• 3:26
  going
• 3:26
  to
• 3:26
  really
• 3:26
  want
• 3:26
  to
• 3:26
  cater
• 3:27
  to
• 3:27
  both
• 3:27
  humans
• 3:28
  and
• 3:28
  agents.
• 3:28
  And
• 3:29
  by
• 3:29
  the
• 3:29
  way,
• 3:29
  bots
• 3:30
  now
• 3:30
  outnumber
• 3:31
  the
• 3:31
  humans
• 3:31
  on
• 3:32
  the
• 3:32
  Internet.
• 3:32
  As
• 3:32
  of
• 3:33
  last
• 3:33
  month,
• 3:33
  they
• 3:33
  crossed
• 3:34
  the
• 3:34
  line.
• 3:35
  So
• 3:36
  just
• 3:36
  as
• 3:37
  it's
• 3:37
  really
• 3:37
  important
• 3:37
  for
• 3:37
  your
• 3:38
  products
• 3:38
  to
• 3:38
  be
• 3:38
  listed
• 3:38
  in
• 3:39
  search
• 3:39
  results,
• 3:40
  you'll
• 3:40
  also
• 3:40
  want
• 3:40
  to
• 3:40
  be
• 3:41
  listed
• 3:41
  in
• 3:41
  answer
• 3:41
  engines.
• 3:42
  And
• 3:42
  the
• 3:43
  key
• 3:43
  question
• 3:43
  for
• 3:44
  us
• 3:44
  is,
• 3:44
  did
• 3:45
  the
• 3:45
  answer
• 3:45
  engine
• 3:46
  get
• 3:46
  her
• 3:46
  to
• 3:46
  buy
• 3:47
  when
• 3:47
  she
• 3:47
  would
• 3:47
  not
• 3:48
  have
• 3:48
  otherwise?
• 3:49
  And
• 3:50
  today
• 3:50
  I
• 3:50
  would
• 3:51
  say
• 3:51
  no.
• 3:51
  Answer
• 3:52
  engines
• 3:52
  are
• 3:52
  substantially
• 3:53
  similar
• 3:53
  to
• 3:54
  search
• 3:54
  engines,
• 3:54
  and
• 3:55
  that's
• 3:55
  because
• 3:55
  of
• 3:55
  a
• 3:55
  disjointed
• 3:56
  experience.
• 3:57
  But
• 3:58
  if
• 3:58
  you
• 3:58
  look
• 3:58
  at
• 3:59
  Amazon,
• 3:59
  for
• 3:59
  example,
• 4:00
  they're
• 4:01
  seeing
• 4:01
  an
• 4:01
  uplift
• 4:01
  in
• 4:02
  sales
• 4:02
  from
• 4:02
  shoppers
• 4:03
  that
• 4:03
  use
• 4:03
  Rufus
• 4:04
  over
• 4:04
  shoppers
• 4:05
  that
• 4:05
  don't.
• 4:06
  So
• 4:06
  I
• 4:07
  kind
• 4:07
  of
• 4:07
  feel
• 4:07
  like
• 4:07
  if
• 4:07
  you
• 4:08
  get
• 4:08
  the
• 4:08
  experience
• 4:09
  right,
• 4:09
  then
• 4:09
  it
• 4:09
  could
• 4:10
  deliver
• 4:10
  incremental
• 4:11
  sales.
• 4:11
  Got
• 4:11
  it,
• 4:11
  David.
• 4:12
  And
• 4:12
  that
• 4:12
  brings
• 4:12
  up.
• 4:12
  Actually
• 4:13
  brings
• 4:13
  up
• 4:13
  the
• 4:13
  next
• 4:13
  question
• 4:13
  I
• 4:14
  was
• 4:14
  going
• 4:14
  to
• 4:14
  ask
• 4:14
  you
• 4:14
  because,
• 4:14
  you
• 4:15
  know,
• 4:15
  sometimes
• 4:15
  when
• 4:15
  I
• 4:15
  sit
• 4:16
  back
• 4:16
  and
• 4:16
  I
• 4:16
  think
• 4:16
  about
• 4:16
  everything
• 4:17
  that's
• 4:17
  been
• 4:17
  going
• 4:17
  on,
• 4:17
  I
• 4:18
  just
• 4:18
  wonder
• 4:18
  if
• 4:18
  we're
• 4:18
  just
• 4:18
  reimagining
• 4:20
  search
• 4:21
  first
• 4:21
  and
• 4:21
  foremost.
• 4:21
  Which
• 4:22
  brings
• 4:22
  up
• 4:22
  a
• 4:22
  natural
• 4:23
  tension
• 4:23
  to
• 4:23
  me
• 4:23
  in
• 4:24
  my
• 4:24
  mind.
• 4:24
  Like,
• 4:24
  there's
• 4:25
  always
• 4:25
  a
• 4:25
  tension
• 4:25
  in
• 4:25
  search
• 4:26
  between
• 4:26
  advertising
• 4:27
  and
• 4:27
  trust.
• 4:28
  So
• 4:29
  how
• 4:29
  do
• 4:29
  you
• 4:29
  predict
• 4:30
  this
• 4:30
  reimagining
• 4:31
  of
• 4:31
  search,
• 4:32
  so
• 4:32
  to
• 4:32
  speak,
• 4:32
  will
• 4:32
  evolve?
• 4:33
  And
• 4:33
  how
• 4:33
  should
• 4:34
  retailers
• 4:34
  think
• 4:34
  about
• 4:35
  this
• 4:35
  tension
• 4:36
  between
• 4:36
  profiting
• 4:37
  from
• 4:37
  but
• 4:37
  also
• 4:37
  figuring
• 4:38
  out
• 4:38
  how
• 4:38
  to
• 4:38
  participate
• 4:39
  with
• 4:40
  the
• 4:40
  answer
• 4:40
  engines?
• 4:41
  Yeah,
• 4:41
  it's
• 4:41
  very
• 4:42
  expensive
• 4:42
  to
• 4:42
  run
• 4:43
  an
• 4:43
  answer
• 4:43
  engine,
• 4:43
  and
• 4:44
  most
• 4:44
  answer
• 4:44
  engines
• 4:45
  can't
• 4:45
  survive
• 4:45
  on
• 4:46
  consumer
• 4:47
  subscriptions
• 4:47
  alone.
• 4:48
  So
• 4:48
  advertising
• 4:49
  is
• 4:49
  coming
• 4:50
  and
• 4:51
  sponsored
• 4:51
  prompts
• 4:52
  are
• 4:52
  the
• 4:52
  first
• 4:52
  shot
• 4:53
  that
• 4:53
  I've
• 4:53
  seen.
• 4:53
  I
• 4:54
  know
• 4:54
  Amazon's
• 4:54
  doing
• 4:54
  that
• 4:55
  and
• 4:55
  Walmart's
• 4:56
  a
• 4:56
  fast
• 4:56
  follower
• 4:56
  in
• 4:56
  that
• 4:56
  regard.
• 4:58
  Ads
• 4:58
  are
• 4:59
  coming
• 4:59
  and
• 4:59
  it's
• 5:00
  going
• 5:00
  to
• 5:00
  result
• 5:00
  in
• 5:00
  a
• 5:01
  loss
• 5:01
  of
• 5:01
  trust
• 5:02
  for
• 5:02
  some
• 5:02
  savvy
• 5:03
  consumers.
• 5:03
  But
• 5:04
  we've
• 5:04
  been
• 5:04
  dealing
• 5:04
  that
• 5:04
  for
• 5:04
  a
• 5:05
  while
• 5:05
  and
• 5:05
  as
• 5:05
  you
• 5:05
  mentioned,
• 5:05
  Chris,
• 5:06
  this
• 5:06
  is
• 5:06
  an
• 5:06
  echo
• 5:06
  of
• 5:07
  se.
• 5:08
  It's
• 5:08
  not
• 5:08
  just
• 5:09
  the
• 5:09
  ads.
• 5:09
  Also,
• 5:10
  it's
• 5:10
  answer
• 5:11
  engine
• 5:11
  optimization.
• 5:12
  Companies
• 5:12
  figuring
• 5:13
  out
• 5:13
  how
• 5:13
  to
• 5:13
  manipulate
• 5:14
  LLMs
• 5:14
  as
• 5:15
  well.
• 5:15
  Right.
• 5:16
  Just
• 5:16
  like
• 5:16
  they
• 5:16
  did
• 5:16
  with
• 5:16
  search
• 5:17
  engines.
• 5:17
  So
• 5:18
  if
• 5:18
  you
• 5:18
  think
• 5:18
  about
• 5:18
  it,
• 5:18
  ChatGPT
• 5:19
  is
• 5:20
  notoriously
• 5:21
  known
• 5:22
  for
• 5:22
  loving
• 5:22
  top
• 5:22
  10
• 5:23
  lists.
• 5:23
  So
• 5:24
  retailers
• 5:24
  have
• 5:24
  figured
• 5:24
  out,
• 5:25
  hey,
• 5:25
  if
• 5:25
  I
• 5:25
  publish
• 5:25
  a
• 5:25
  bunch
• 5:26
  of
• 5:26
  blogs
• 5:26
  with
• 5:26
  top
• 5:26
  10
• 5:27
  lists,
• 5:27
  I
• 5:27
  put
• 5:27
  my
• 5:27
  products
• 5:27
  at
• 5:28
  the
• 5:28
  top.
• 5:29
  More
• 5:29
  chance
• 5:29
  that
• 5:30
  I'll
• 5:30
  be
• 5:30
  mentioned
• 5:30
  by
• 5:31
  an
• 5:31
  answer
• 5:31
  engine.
• 5:31
  So
• 5:32
  there's
• 5:32
  already
• 5:32
  a
• 5:33
  lot
• 5:33
  of
• 5:33
  manipulation
• 5:34
  going
• 5:34
  on.
• 5:34
  But
• 5:34
  there's.
• 5:35
  But
• 5:35
  it's
• 5:35
  not
• 5:35
  new.
• 5:36
  We've
• 5:36
  dealt
• 5:36
  with
• 5:36
  this
• 5:37
  with
• 5:37
  search
• 5:37
  engines
• 5:38
  and
• 5:38
  consumers
• 5:39
  will
• 5:39
  figure
• 5:39
  it
• 5:39
  out.
• 5:39
  This
• 5:40
  is
• 5:40
  an
• 5:40
  ongoing
• 5:40
  battle
• 5:41
  that
• 5:41
  dates
• 5:42
  back
• 5:42
  to
• 5:42
  Mad
• 5:42
  Men,
• 5:43
  right?
• 5:44
  Yeah.
• 5:44
  So,
• 5:44
  David,
• 5:45
  how
• 5:45
  optimistic
• 5:46
  then
• 5:46
  are
• 5:46
  you
• 5:46
  about
• 5:47
  shopping
• 5:47
  agents
• 5:48
  actually
• 5:48
  delivering
• 5:48
  time
• 5:49
  savings
• 5:49
  and
• 5:50
  how
• 5:50
  do
• 5:50
  you
• 5:50
  envision
• 5:51
  them
• 5:51
  giving
• 5:52
  or
• 5:52
  gaining
• 5:53
  consumer
• 5:53
  trust,
• 5:53
  really,
• 5:54
  and
• 5:54
  consumer
• 5:55
  adoption,
• 5:55
  if
• 5:56
  ever?
• 5:56
  That's
• 5:56
  where
• 5:57
  we
• 5:57
  diverge
• 5:57
  a
• 5:57
  little
• 5:58
  bit
• 5:58
  from
• 5:58
  answer
• 5:58
  engines
• 5:59
  to
• 6:00
  kind
• 6:00
  of
• 6:00
  standalone
• 6:00
  shopping
• 6:01
  agents.
• 6:02
  And
• 6:02
  I
• 6:03
  definitely
• 6:03
  predict
• 6:03
  that
• 6:03
  third
• 6:04
  parties
• 6:04
  are
• 6:04
  going
• 6:04
  to
• 6:04
  start
• 6:04
  offering
• 6:05
  the
• 6:05
  shopping
• 6:05
  agent
• 6:06
  subscriptions
• 6:07
  that
• 6:07
  are
• 6:07
  going
• 6:07
  to
• 6:07
  automate
• 6:08
  some
• 6:08
  select
• 6:09
  shopping
• 6:10
  tasks.
• 6:10
  And
• 6:11
  I
• 6:11
  think
• 6:11
  it's
• 6:11
  going
• 6:11
  to
• 6:12
  start
• 6:12
  with
• 6:12
  delivery
• 6:12
  companies
• 6:13
  like
• 6:13
  Instacart
• 6:14
  and
• 6:14
  DoorDash,
• 6:14
  because
• 6:15
  they
• 6:15
  already
• 6:15
  look
• 6:16
  across
• 6:16
  multiple
• 6:16
  sources
• 6:17
  for
• 6:17
  their
• 6:17
  products.
• 6:18
  Right.
• 6:18
  And
• 6:18
  they
• 6:18
  already
• 6:19
  have
• 6:19
  the
• 6:19
  delivery
• 6:19
  infrastructure.
• 6:20
  So
• 6:21
  they
• 6:21
  might
• 6:21
  start
• 6:21
  with
• 6:21
  that
• 6:22
  answer
• 6:22
  engine
• 6:22
  integration
• 6:23
  like
• 6:23
  Instacart
• 6:24
  has
• 6:24
  done.
• 6:24
  But
• 6:25
  I
• 6:25
  think
• 6:25
  they
• 6:25
  will
• 6:26
  eventually
• 6:26
  offer
• 6:26
  their
• 6:27
  own
• 6:27
  agents
• 6:28
  that
• 6:28
  are
• 6:28
  personalized
• 6:29
  to
• 6:29
  your
• 6:29
  tastes,
• 6:30
  and
• 6:30
  that's
• 6:30
  the
• 6:30
  big
• 6:30
  value
• 6:31
  add.
• 6:32
  So
• 6:32
  they'll
• 6:32
  know
• 6:32
  your
• 6:32
  favorite
• 6:33
  restaurants,
• 6:34
  your
• 6:34
  favorite
• 6:35
  orders,
• 6:35
  the
• 6:36
  grocery
• 6:36
  items
• 6:36
  you
• 6:36
  typically
• 6:37
  buy.
• 6:37
  They'll
• 6:38
  find
• 6:38
  the
• 6:38
  best
• 6:38
  prices
• 6:39
  and
• 6:39
  they'll
• 6:39
  use
• 6:39
  coupons.
• 6:41
  That's
• 6:41
  where
• 6:41
  a
• 6:41
  lot
• 6:42
  of
• 6:42
  that
• 6:42
  time
• 6:42
  savings
• 6:43
  is
• 6:43
  going
• 6:43
  to
• 6:43
  come.
• 6:44
  I
• 6:44
  also
• 6:44
  think
• 6:45
  there's
• 6:45
  going
• 6:45
  to
• 6:45
  be
• 6:45
  more
• 6:46
  focus
• 6:46
  on
• 6:46
  outcomes
• 6:47
  over
• 6:47
  products.
• 6:49
  So
• 6:49
  you're
• 6:49
  going
• 6:49
  to
• 6:49
  say
• 6:50
  things
• 6:50
  like
• 6:51
  plan
• 6:51
  a
• 6:51
  birthday
• 6:52
  party
• 6:52
  with
• 6:52
  a
• 6:52
  pirate
• 6:53
  theme
• 6:53
  and
• 6:53
  shop
• 6:54
  for
• 6:54
  the
• 6:54
  items
• 6:54
  and
• 6:54
  it'll
• 6:55
  do
• 6:55
  that
• 6:55
  for
• 6:55
  you.
• 6:55
  Or
• 6:56
  I
• 6:56
  want
• 6:56
  a
• 6:56
  dinner
• 6:56
  party
• 6:56
  for
• 6:57
  eight
• 6:57
  with
• 6:57
  a
• 6:57
  French
• 6:58
  bistro
• 6:59
  theme
• 6:59
  or
• 6:59
  something
• 6:59
  like
• 7:00
  that,
• 7:00
  and
• 7:00
  the
• 7:00
  agent
• 7:01
  will
• 7:01
  go
• 7:01
  off
• 7:01
  and
• 7:02
  grab
• 7:02
  all
• 7:02
  those
• 7:02
  items
• 7:03
  and
• 7:03
  have
• 7:03
  them
• 7:03
  delivered
• 7:03
  to
• 7:03
  you.
• 7:04
  That's
• 7:04
  interesting.
• 7:04
  I
• 7:05
  just
• 7:05
  heard
• 7:05
  Today
• 7:05
  actually
• 7:05
  that
• 7:05
  UBS
• 7:06
  is
• 7:06
  predicting
• 7:07
  that
• 7:08
  grocery
• 7:08
  will
• 7:08
  be
• 7:08
  the
• 7:09
  first
• 7:09
  industry
• 7:09
  hit
• 7:10
  by
• 7:10
  agentic
• 7:11
  commerce,
• 7:11
  particularly
• 7:12
  from
• 7:12
  a
• 7:12
  commerce
• 7:12
  perspective,
• 7:13
  which,
• 7:13
  you
• 7:13
  know,
• 7:13
  is
• 7:14
  kind
• 7:14
  of
• 7:14
  the
• 7:14
  flip
• 7:14
  side
• 7:14
  of
• 7:14
  how
• 7:15
  we
• 7:15
  saw
• 7:15
  E
• 7:15
  commerce
• 7:16
  develop
• 7:16
  overall.
• 7:17
  So,
• 7:17
  so
• 7:17
  all
• 7:17
  right,
• 7:18
  so
• 7:18
  let's
• 7:18
  get
• 7:18
  you
• 7:18
  out
• 7:18
  here
• 7:18
  on
• 7:19
  this.
• 7:19
  We've
• 7:20
  talked
• 7:20
  a
• 7:20
  lot
• 7:20
  about
• 7:20
  the
• 7:20
  consumer
• 7:21
  facing
• 7:21
  side
• 7:21
  of
• 7:22
  generative
• 7:22
  AI
• 7:23
  already,
• 7:24
  but
• 7:24
  what's
• 7:24
  your
• 7:24
  take
• 7:24
  on
• 7:25
  how
• 7:25
  agents
• 7:25
  will
• 7:26
  transform
• 7:27
  internal
• 7:27
  retail
• 7:28
  operations
• 7:28
  in
• 7:28
  2026?
• 7:30
  Yeah,
• 7:30
  here's
• 7:30
  where
• 7:30
  I
• 7:31
  think
• 7:31
  the
• 7:31
  crystal
• 7:31
  ball
• 7:31
  gets
• 7:32
  a
• 7:32
  little
• 7:32
  clearer
• 7:32
  because
• 7:33
  you
• 7:33
  know,
• 7:33
  consumers
• 7:34
  are
• 7:34
  pretty
• 7:34
  fickle,
• 7:34
  but
• 7:35
  companies
• 7:35
  are
• 7:35
  all
• 7:35
  about
• 7:36
  increasing
• 7:36
  efficiency
• 7:37
  and
• 7:38
  that's
• 7:38
  kind
• 7:38
  of
• 7:38
  the
• 7:38
  sweet
• 7:39
  spot
• 7:39
  for
• 7:39
  AI.
• 7:40
  So
• 7:40
  retailers
• 7:40
  are
• 7:41
  definitely
• 7:41
  going
• 7:41
  to
• 7:41
  start
• 7:41
  deploying
• 7:42
  these
• 7:42
  agents
• 7:42
  that
• 7:43
  leverage
• 7:43
  reasoning
• 7:44
  to
• 7:44
  assist
• 7:44
  with
• 7:45
  things
• 7:45
  like
• 7:45
  merchandising,
• 7:46
  marketing
• 7:46
  and
• 7:46
  supply
• 7:47
  chain
• 7:47
  processes.
• 7:48
  We'll
• 7:49
  start
• 7:49
  with
• 7:49
  a
• 7:49
  lot
• 7:49
  of
• 7:49
  human
• 7:50
  in
• 7:50
  the
• 7:50
  loop
• 7:50
  as
• 7:50
  we
• 7:51
  gain
• 7:51
  confidence,
• 7:51
  which
• 7:52
  makes
• 7:52
  a
• 7:52
  lot
• 7:52
  of
• 7:52
  sense.
• 7:53
  And
• 7:54
  one
• 7:54
  example
• 7:54
  is
• 7:54
  last
• 7:55
  week
• 7:55
  at
• 7:55
  Re
• 7:55
  Invent
• 7:55
  we
• 7:56
  announced
• 7:56
  three
• 7:56
  New
• 7:56
  Frontier
• 7:57
  agents
• 7:58
  covering
• 7:58
  developers,
• 7:59
  security
• 8:00
  and
• 8:00
  cloud
• 8:00
  ops.
• 8:01
  And
• 8:01
  that's
• 8:02
  kind
• 8:02
  of
• 8:02
  the
• 8:02
  first
• 8:03
  start.
• 8:03
  I
• 8:03
  think
• 8:04
  we're
• 8:04
  going
• 8:04
  to
• 8:04
  start
• 8:04
  to
• 8:04
  see
• 8:05
  a
• 8:05
  lot
• 8:05
  of
• 8:05
  new
• 8:05
  agents
• 8:06
  and
• 8:06
  it's
• 8:06
  inevitable
• 8:07
  that
• 8:07
  they'll
• 8:07
  start
• 8:07
  with
• 8:08
  horizontal
• 8:08
  agents
• 8:09
  that
• 8:09
  work
• 8:09
  across
• 8:09
  lots
• 8:10
  of
• 8:10
  industries.
• 8:11
  But
• 8:11
  eventually
• 8:11
  we'll
• 8:11
  get
• 8:12
  to
• 8:12
  industry
• 8:12
  specific
• 8:13
  agents
• 8:13
  that
• 8:14
  can
• 8:14
  really
• 8:14
  help
• 8:14
  people
• 8:15
  in
• 8:15
  places
• 8:15
  like
• 8:15
  merchandising.
• 8:17
  And
• 8:17
  I
• 8:17
  wouldn't
• 8:17
  be
• 8:18
  surprised
• 8:18
  if
• 8:18
  eventually
• 8:19
  your
• 8:19
  SaaS
• 8:19
  subscription
• 8:20
  for
• 8:20
  some
• 8:21
  application
• 8:21
  comes
• 8:22
  with
• 8:22
  agents
• 8:22
  that
• 8:22
  help
• 8:23
  you
• 8:23
  run
• 8:23
  that
• 8:23
  application.
• 8:24
  That
• 8:25
  that
• 8:25
  could
• 8:25
  be
• 8:25
  coming
• 8:25
  up
• 8:25
  soon
• 8:26
  in
• 8:26
  the
• 8:26
  future
• 8:26
  as
• 8:27
  well.
• 8:27
  And
• 8:27
  then
• 8:28
  I
• 8:28
  also
• 8:28
  want
• 8:28
  to
• 8:28
  mention
• 8:28
  we
• 8:29
  just
• 8:29
  announced
• 8:29
  a
• 8:29
  partnership
• 8:30
  with
• 8:30
  Visa
• 8:31
  around
• 8:32
  agentic
• 8:32
  payments
• 8:33
  and
• 8:33
  I
• 8:33
  think,
• 8:34
  yeah,
• 8:34
  this
• 8:34
  works
• 8:35
  for
• 8:35
  B2B
• 8:35
  and
• 8:36
  B2C.
• 8:37
  I
• 8:37
  think
• 8:37
  B2B
• 8:37
  is
• 8:38
  going
• 8:38
  to
• 8:38
  be
• 8:38
  the
• 8:38
  big
• 8:38
  impact
• 8:39
  because
• 8:39
  I
• 8:39
  can
• 8:39
  imagine
• 8:40
  the
• 8:40
  whole
• 8:40
  purchase
• 8:41
  order,
• 8:41
  inventory,
• 8:42
  receiving,
• 8:42
  invoice,
• 8:43
  match
• 8:43
  payment,
• 8:44
  all
• 8:44
  that
• 8:44
  stuff
• 8:45
  can
• 8:45
  be
• 8:45
  easily
• 8:45
  handled
• 8:46
  by
• 8:46
  an
• 8:46
  agent
• 8:46
  that's
• 8:47
  checking
• 8:47
  all
• 8:47
  the
• 8:47
  policies,
• 8:48
  making
• 8:48
  sure
• 8:48
  everything's
• 8:49
  been
• 8:49
  done
• 8:49
  right
• 8:50
  and
• 8:50
  then
• 8:51
  making
• 8:51
  a
• 8:51
  payment
• 8:52
  without
• 8:52
  having
• 8:52
  to
• 8:52
  do
• 8:53
  checks
• 8:53
  and
• 8:53
  things
• 8:53
  like
• 8:54
  that.
• 8:55
  So
• 8:55
  I
• 8:55
  think,
• 8:55
  I
• 8:56
  think
• 8:56
  that's
• 8:56
  going
• 8:56
  to
• 8:56
  have
• 8:56
  a
• 8:56
  big
• 8:57
  impact
• 8:57
  on
• 8:57
  the
• 8:57
  back
• 8:58
  end
• 8:58
  as
• 8:58
  well.
• 8:58
  Amazing
• 8:59
  stuff.
• 8:59
  David,
• 8:59
  what
• 9:00
  are
• 9:00
  your
• 9:00
  plans
• 9:00
  for
• 9:00
  NRF
• 9:01
  before
• 9:01
  we
• 9:01
  let
• 9:01
  you
• 9:02
  really
• 9:02
  go?
• 9:02
  Yeah,
• 9:03
  so
• 9:03
  I'll
• 9:03
  be
• 9:04
  in
• 9:04
  the
• 9:04
  booth
• 9:04
  and
• 9:05
  hanging
• 9:05
  around
• 9:06
  on
• 9:06
  the
• 9:06
  floor.
• 9:06
  We're
• 9:06
  in
• 9:06
  booth
• 9:07
  4438
• 9:08
  this
• 9:08
  year
• 9:09
  and
• 9:09
  then
• 9:09
  also
• 9:10
  want
• 9:10
  to
• 9:10
  give
• 9:10
  a
• 9:10
  shout
• 9:10
  out
• 9:11
  to
• 9:11
  Super
• 9:11
  Saturday.
• 9:12
  So
• 9:12
  the
• 9:12
  retail
• 9:12
  ROI
• 9:13
  folks
• 9:13
  have
• 9:13
  their
• 9:15
  special
• 9:15
  conference
• 9:16
  on
• 9:16
  the
• 9:16
  Saturday
• 9:16
  before
• 9:17
  NRF
• 9:17
  that's
• 9:17
  free
• 9:17
  for
• 9:18
  retailers
• 9:18
  to
• 9:18
  attend.
• 9:19
  It's
• 9:19
  be
• 9:19
  great
• 9:19
  for
• 9:20
  people
• 9:20
  to
• 9:20
  go
• 9:20
  there.
• 9:20
  As
• 9:20
  well.
• 9:21
  I'm
• 9:21
  a
• 9:21
  big
• 9:21
  supporter.
• 9:22
  Did
• 9:22
  a
• 9:22
  trip
• 9:22
  to
• 9:22
  Honduras
• 9:23
  with
• 9:23
  the
• 9:23
  gang
• 9:24
  and
• 9:24
  handed
• 9:24
  out
• 9:24
  Tom's
• 9:25
  shoes
• 9:25
  once.
• 9:26
  And
• 9:26
  it's
• 9:26
  a
• 9:26
  it's
• 9:27
  a
• 9:27
  great
• 9:27
  time
• 9:28
  to
• 9:28
  really
• 9:28
  learn
• 9:29
  about
• 9:29
  retail
• 9:29
  tech,
• 9:30
  but
• 9:30
  also
• 9:30
  hear
• 9:30
  about
• 9:31
  some
• 9:31
  people
• 9:31
  that
• 9:32
  are
• 9:32
  less
• 9:32
  fortunate.
• 9:33
  Such
• 9:33
  a
• 9:33
  great
• 9:33
  capstone
• 9:34
  to
• 9:34
  2026
• 9:34
  for
• 9:35
  us,
• 9:35
  too.
• 9:35
  Really
• 9:35
  appreciate
• 9:36
  you
• 9:36
  coming
• 9:36
  on.