TRANSKRİPT 🔗
WEBVTT
00:00:00.000 --> 00:00:04.687
Welcome to Shallow DeepDive, where we take
a deep-ish dive into the shallow world of
00:00:04.767 --> 00:00:06.673
branding, advertising and creative.
00:00:06.753 --> 00:00:07.524
I'm Joe Qualls.
00:00:07.604 --> 00:00:12.178
Today, we're covering why we've become an
OpenAI Select Partner, and what we think
00:00:12.258 --> 00:00:13.937
that actually obliges us to do.
00:00:14.017 --> 00:00:18.647
Let's get the announcement out of the way
first, because it's the least interesting
00:00:18.727 --> 00:00:19.439
part of this.
00:00:19.519 --> 00:00:23.355
Qualls has been named a Select Partner in
the OpenAI Partner Network.
00:00:23.435 --> 00:00:24.263
That's the news.
00:00:24.343 --> 00:00:28.746
It took one sentence, and if you came here
for the sentence, you're free to go.
00:00:28.826 --> 00:00:33.002
Now the awkward bit, which is the reason
we wanted to talk about it at all.
00:00:33.082 --> 00:00:35.726
Every agency in the country has a wall of
these.
00:00:35.806 --> 00:00:40.380
Partner badges, certifications,
accredited-this and authorised-that,
00:00:40.460 --> 00:00:43.444
tidy grey row down near the footer where
nobody looks.
00:00:43.524 --> 00:00:45.430
They're meant to signal capability.
00:00:45.510 --> 00:00:48.211
Mostly they signal that someone filled in
a form.
00:00:48.291 --> 00:00:52.808
And we're aware that adding another one to
the wall is a strange thing to make an
00:00:52.888 --> 00:00:57.291
episode about — that the honest reaction
to most partnership announcements is a
00:00:57.371 --> 00:00:58.653
polite nod and a scroll.
00:00:58.733 --> 00:01:03.136
So we should say what this one is for,
because we didn't take it for the row of
00:01:03.216 --> 00:01:03.928
logos.
00:01:04.008 --> 00:01:08.582
Back in episode seven we said an agency's
position on AI should be boringly clear.
00:01:08.662 --> 00:01:10.795
Not evangelical, not squeamish — clear.
00:01:10.875 --> 00:01:15.562
Clients shouldn't have to work out where
you stand by reading between the lines of
00:01:15.642 --> 00:01:16.354
proposal.
00:01:16.434 --> 00:01:21.121
They should be able to ask, and get a
plain answer, and know that the answer
00:01:21.201 --> 00:01:22.256
the same next month.
00:01:22.336 --> 00:01:24.185
We meant it, and it's easy to say.
00:01:24.265 --> 00:01:28.498
The trouble with a position stated in the
abstract is that it costs nothing.
00:01:28.578 --> 00:01:30.371
Anyone can announce a philosophy.
00:01:30.451 --> 00:01:35.081
It only starts to mean something when you
attach yourself to something specific and
00:01:35.161 --> 00:01:36.046
let people check.
00:01:36.126 --> 00:01:38.486
That's the first thing this is: a
specific.
00:01:38.566 --> 00:01:42.288
We've stopped describing our position and
started demonstrating it.
00:01:42.368 --> 00:01:46.715
You can go and look at what we've
committed to, and if the work doesn't
00:01:46.795 --> 00:01:50.517
that's now visible in a way that a
paragraph on a website never is.
00:01:50.597 --> 00:01:53.695
The second thing is more uncomfortable,
and more useful.
00:01:53.775 --> 00:01:56.702
A partnership is an obligation before it's
a benefit.
00:01:56.782 --> 00:02:01.242
The programme gives you access to
enablement, technical resources, support —
00:02:01.322 --> 00:02:05.669
machinery that helps you actually deploy
this stuff rather than gesture at it.
00:02:05.749 --> 00:02:09.300
Which sounds like a gift, and is, but it
also removes an excuse.
00:02:09.380 --> 00:02:13.784
Once you've got the training and the
access and the direct line, you can't hide
00:02:13.864 --> 00:02:14.862
behind not knowing.
00:02:14.942 --> 00:02:17.813
You've said, in public, that you'll be
good at this.
00:02:17.893 --> 00:02:19.004
Not interested in it.
00:02:19.084 --> 00:02:20.196
Not excited about it.
00:02:20.276 --> 00:02:20.989
Good at it.
00:02:21.069 --> 00:02:24.677
That's a much higher bar and a much
shorter distance to fall off.
00:02:24.757 --> 00:02:26.607
We took it partly for that reason.
00:02:26.687 --> 00:02:31.374
It's easy to be vaguely enthusiastic about
AI from a safe distance, and the industry
00:02:31.454 --> 00:02:36.140
is full of people doing exactly that —
comfortable at the level of the keynote,
00:02:36.220 --> 00:02:37.616
at the level of the build.
00:02:37.696 --> 00:02:42.156
Putting your name against a specific set
of tools is a way of forcing the second
00:02:42.236 --> 00:02:42.948
thing.
00:02:43.028 --> 00:02:44.367
It's a commitment device.
00:02:44.447 --> 00:02:45.672
We wanted the pressure.
00:02:45.752 --> 00:02:47.148
So that's why we said yes.
00:02:47.228 --> 00:02:51.688
The more practical question is what
changes for the people we work with, and
00:02:51.768 --> 00:02:56.284
honest answer is: less than you'd expect,
and the parts that don't change are the
00:02:56.364 --> 00:02:57.135
important ones.
00:02:57.215 --> 00:02:58.895
The brief still has to be good.
00:02:58.975 --> 00:03:03.548
We spent a whole episode on this — the
faster the tool, the better the brief
00:03:03.628 --> 00:03:06.783
to be — and the partnership makes it more
true, not less.
00:03:06.863 --> 00:03:10.131
A better model is a faster route to
whatever you asked for.
00:03:10.211 --> 00:03:14.614
If what you asked for was mush, it's now
mush at speed, and delivered with real
00:03:14.694 --> 00:03:15.406
confidence.
00:03:15.486 --> 00:03:18.698
Nothing in this arrangement lets anyone
skip the thinking.
00:03:18.778 --> 00:03:21.649
Strategy still comes first, or you pay for
it twice.
00:03:21.729 --> 00:03:23.805
Also a whole episode, also unaffected.
00:03:23.885 --> 00:03:28.459
If you don't know who you're for and what
you're claiming, no amount of capability
00:03:28.539 --> 00:03:29.251
rescues that.
00:03:29.331 --> 00:03:32.997
It just gets you to the wrong place with
better production values.
00:03:33.077 --> 00:03:35.493
And the point of view is still ours to
hold.
Load full transcript
00:03:35.573 --> 00:03:40.204
We said this about AI content and we'll
say it again here: the tool has no
00:03:40.284 --> 00:03:44.687
It can't tell you which true thing about
your business is the one worth saying,
00:03:44.767 --> 00:03:47.921
because it doesn't know what it would cost
you to say it.
00:03:48.001 --> 00:03:49.397
That judgement is the job.
00:03:49.477 --> 00:03:54.050
It was the job before, it's the job now,
and if anything it's more of the job than
00:03:54.130 --> 00:03:58.590
it used to be — because when the making
gets cheap, the deciding is what's left.
00:03:58.670 --> 00:04:03.357
Here's what genuinely does change, and
it's narrower than a press release would
00:04:03.437 --> 00:04:04.149
it.
00:04:04.229 --> 00:04:08.008
We can say yes to a category of work we'd
previously have hedged on.
00:04:08.088 --> 00:04:12.378
Not because we couldn't have attempted it,
but because attempting things on a
00:04:12.458 --> 00:04:17.145
client's budget is a bad way to learn, and
we'd rather have done the learning first.
00:04:17.225 --> 00:04:19.018
Our advice gets less speculative.
00:04:19.098 --> 00:04:23.728
There's a real difference between reading
about where a platform is going and being
00:04:23.808 --> 00:04:25.487
close enough to have some idea.
00:04:25.567 --> 00:04:29.856
Most of what agencies tell clients about
AI right now is inference dressed as
00:04:29.936 --> 00:04:30.649
insight.
00:04:30.729 --> 00:04:32.919
We'd like ours to be a bit less of that.
00:04:32.999 --> 00:04:37.629
And when we tell a client not to use it —
which we do, and will keep doing — that's
00:04:37.709 --> 00:04:40.239
now an informed no rather than a cautious
one.
00:04:40.319 --> 00:04:44.552
Those sound the same from the outside and
are completely different in value.
00:04:44.632 --> 00:04:47.617
A cautious no is just risk aversion with a
nice voice.
00:04:47.697 --> 00:04:51.305
An informed no comes with reasons, and the
reasons are checkable.
00:04:51.385 --> 00:04:55.845
Which brings us to the obvious objection,
and we'd rather raise it than have you
00:04:55.925 --> 00:04:56.810
raise it quietly.
00:04:56.890 --> 00:04:59.988
Isn't this just a vendor relationship with
a nicer name?
00:05:00.068 --> 00:05:04.528
Aren't you now, structurally, a person who
recommends the thing you're partnered
00:05:04.608 --> 00:05:05.320
with?
00:05:05.400 --> 00:05:06.512
It's a fair question.
00:05:06.592 --> 00:05:07.874
It's the right question.
00:05:07.954 --> 00:05:12.527
And there's really only one test that
answers it: whether we ever tell you not
00:05:12.607 --> 00:05:17.067
An agency that recommends the same
solution to every client isn't advising,
00:05:17.147 --> 00:05:17.860
distributing.
00:05:17.940 --> 00:05:22.570
If a partnership quietly turns the answer
to every brief into the same answer, then
00:05:22.650 --> 00:05:27.166
the badge has replaced the judgement and
you should stop paying us for judgement.
00:05:27.246 --> 00:05:31.933
So the way to hold us to this is simple,
and we're handing you the stick on
00:05:32.013 --> 00:05:32.898
Watch for the no.
00:05:32.978 --> 00:05:36.757
If it never comes, something has gone
wrong, and it won't be subtle.
00:05:36.837 --> 00:05:38.516
That's the whole of it, really.
00:05:38.596 --> 00:05:43.226
We're pleased about this — it's a good
thing, and we don't want to be so
00:05:43.306 --> 00:05:45.610
understated that it reads as indifference.
00:05:45.690 --> 00:05:50.377
But a partnership is a claim about what
you'll be capable of, made in advance and
00:05:50.457 --> 00:05:54.916
public, and the only thing that settles a
claim like that is the work that comes
00:05:54.996 --> 00:05:55.709
after it.
00:05:55.789 --> 00:05:58.263
The badge goes in the footer with the
others.
00:05:58.343 --> 00:06:01.270
What it's worth gets decided somewhere
else entirely.
00:06:01.350 --> 00:06:04.391
Thank you for indulging us in today's
Shallow DeepDive.
00:06:04.471 --> 00:06:06.661
We hope you'll join us for the next one.
00:06:06.741 --> 00:06:07.910
Until then, take care.