A Brand is Trust

Jasmine and Matt open with the line from Matt's latest piece that neither of them can shake, then trace a single argument from an unexpected award to the heart of what marketers owe their audience in the AI era. The conversation moves from why brand is trust, through a practical method for keeping AI output factually clean, to discernment and synthetic customers as the human checks that protect a reputation.

Welcome

Speaker: Jasmine [00:02]

Hello, hello, Matt.

Speaker: Matt [00:04]

Hey Jasmine, how you doing?

Speaker: Jasmine [00:06]

Very well, and yourself?

Speaker: Matt [00:08]

I'm very good, thank you. Well, some would argue that I'm not good, but you know, sometimes I misbehave.

Speaker: Jasmine [00:13]

I would say that you're a good man. And I'm always learning from you. So there you have it. Okay, let's do that today. Your latest piece opens with a sentence I haven't been able to get out of my head. Somewhere in your marketing stack right now, an AI is publishing faster than anyone can check it.

Speaker: Matt [00:17]

Well, thank you. Thank you.

The CX50 recognition, and burying the good news

Speaker: Jasmine [00:42]

And the argument you build from there is a quiet one. It's about trust and how fast you can lose it without noticing. But here's where I want to start. The piece only exists because you've just been named on the Marketing Week CX50 list, congrats, as one of the top 10 life sciences marketers in the UK. And true to form, you buried that news somewhere near the bottom of the page. So share with us, how did you actually end up there?

Speaker: Matt [01:21]

Well, I'm not entirely sure, but also I should note that on the CX50 list, I am buried at the bottom of the page. That's partly because they're alphabetical, but also because all the others, you know, are people from much, much bigger organizations. So I'm not quite there, I'm still not quite sure it feels real that I'm amongst such exalted luminaries. I don't quite know how I got on the list, but I do know roughly how the list is built.

It's not a popularity vote, which was the first thing I checked. Marketing Week run it with Cognizant and Google Cloud, and for the CX50 they rate 50 people most highly across five different segments for customer experience. So there's 10 names in each. Life Sciences is just one of those sectors, and so I'm one of ten. The way they get there is part nomination, part research. They take nominations from their own network and Cognizant, then weigh that against independent measures of which organizations are doing customer experience properly. And then the final cut gets judged against three things: impact, innovation, and influence over the past year and across a career. So somewhere in all of that, my name surfaced, and quite how it stayed on the list, I don't know.

But the process, from what I gather, is more rigorous than I expected, which made it a little bit harder to write off as a mistake.

Why disbelief was the first instinct

Speaker: Jasmine [02:57]

So I know from a previous conversation that you and I had about this nomination and award, your first instinct was disbelief. Share with us why.

Speaker: Matt [03:11]

Yeah, well I got an email from somebody I'd never heard from before, and I thought it was maybe a phishing scam. I get a lot of emails. Most of them are from people wanting to try to sell me stuff or, you know, trying to scam me out of money. But it said that I'd been selected for the award, and it asked me for nothing in return other than fifteen minutes of my time for an interview. And so I figured, what does it hurt to reply and give them a link to a meeting booking link? Which means that everything I shared is publicly available anyway.

Speaker: Jasmine [03:53]

Yeah, so it was a little bit of, okay, I have a bit of extra time, let's go on faith and see what they have to say.

Speaker: Matt [04:03]

Yeah. I mean, I hadn't entered anything. I certainly wasn't expecting it, and I really didn't think I was the sort of person who'd end up on this list. But it turned out to be real. You know, Marketing Week published the list last week and, yeah, very, very honoured to be part of it.

Speaker: Jasmine [04:28]

So there's something quite fitting about a person whose entire argument is check before you believe it, refusing to believe his own good news until he checked it a whole bunch of times.

Speaker: Matt [04:43]

Well, yeah, I checked it, I had conversations with them, I connected with all the people that were on the email thread, because it wasn't just a lone sender. And even after that, and after connecting with people from Haymarket Media, who are the publishers of Marketing Week, I still didn't believe it until I saw it in print. Well, at least in digital print, as it were.

Brand equals trust

Speaker: Jasmine [05:13]

Okay, so what did you tell them? What was the conversation with them?

Speaker: Matt [05:20]

Yeah, so I told them a truth that seems all too easily forgotten these days. You know, in the race toward greater speed and efficiency through AI, people tend to forget that brand is essentially trust. And if a brand equals trust, then you really shouldn't undermine it by putting out incorrect information.

Speaker: Jasmine [05:44]

I like that. You know, you and I are both scientists, so we think in a formula kind of way. So what I'm envisaging is a simple formula of brand equals trust, but that almost seems too simple to build a reputation on.

Speaker: Matt [06:06]

Well, it is, but that's the whole point of brand, isn't it? Warren Buffett once said that it takes a lifetime to build a reputation and just five minutes to destroy it. And I think it's really important that, in our rush towards finding greater efficiency and the ability to do more, we don't erode the trust that we've built up through hard work and years and years of putting the customer at the centre, while we're chasing efficiency.

What erosion looks like when you are caught wrong

Speaker: Jasmine [06:47]

So help our listeners by painting a picture for what that erosion could look like. What does being caught wrong look like, and what does it say to the recipient, to your target audience?

Speaker: Matt [07:06]

Yeah, so it often looks small while it's happening. It's a regulated claim that's drifted a few words off label. A figure that was right in the past, but it hasn't been updated, and so you've got that inconsistency. The numbers sitting somewhere, but you haven't carried that forward. Maybe it's a confident line about a competitor that doesn't hold up anymore. None of it feels dangerous when you see the draft. But all of it gets expensive the moment a scientist reads the source and it doesn't match.

Speaker: Jasmine [07:43]

So the damage is cumulative. [reconstructed, please verify]

Speaker: Matt [07:47]

Yeah, it really is. And I think it's also one of those areas where it's that slow erosion that happens really, really quickly. One mistake, people forgive. But if they start seeing sloppy mistakes time and time again, all of a sudden they really do start to question. And this is the part that generic marketing advice often misses. A consumer brand can carry a bit of exaggeration. But the people that our clients sell to, they earned doctorates by not taking things on face. They were trained to test and challenge.

Let me give you a concrete example of a project we worked on together, because this obsession of trying to make sure that everything is factually correct is really, really important. I set up two separate AI projects using exactly the same sets of references to help me create a white paper. I did that in two different models, so I had one in ChatGPT, one in Claude. And after working with Claude to architect the structure of the white paper, figure out the best claims and the best structure, and generate the text, I took that into the other model and got it to fact check. Then I went backwards and forwards, taking the feedback back into Claude, getting it to make edits, taking it back in, making sure we weren't overstating claims, we weren't understating things, that basically everything was as true as it possibly could be.

The claim-by-claim fact-check table

Speaker: Matt [09:49]

But then I did something that I don't know a lot of people have thought of yet. I got the AI to build me a table that had one row per claim. Each row carried the claim itself, where it appears in the white paper, the reference behind it, and the exact spot in that reference where the claim sits, so that for any sentence with a claim in it, I could point to chapter and verse. And then I could go through line by line and easily check every single statement that was made.

So while I'd had two AIs check everything for me, I also put that final pass in of making sure that every claim was correct. What was great was that not only did we get very, very minimal changes to the bulk of the white paper, really the only change came in the conclusion, the future-looking statements where we didn't have any evidence to support. So that was a process that really didn't take too much time. For that sort of process, you could easily build a workflow to run it. If you're doing that all day, every day, you get to a point where you're building something that's then really easy for a human to check.

Speaker: Jasmine [11:01]

I really like that example because, from a product marketing perspective, where I've spent a lot of my time, that is the equivalent of value mapping each of your claims to make sure that the claim ties back to the evidence and has a clear benefit that is evidence-based. And you're clearly using that approach also for outbound marketing.

Catching calculation errors before regulatory sees them

Speaker: Matt [11:34]

Yeah. I did something in a fun way, and you talk about outbound marketing, but there was another example where I was creating a blog from a white paper for somebody and actually identified that in the white paper there appeared to be some calculation errors. Because of using the AI and getting it to go through this process, it actually flagged, and had a bit of a hissy fit about, I will say, the fact that some of the maths didn't appear to add up. Now I didn't have access to all of the source material, but that was then something I could easily flag internally to the organization so they could work through the process and then alter the language in the white paper itself, push that back through regulatory before the blogs went through. So I think this sort of approach of an augmented human working on things is really, really important.

Faster is easy, dependable is the harder bar

Speaker: Jasmine [12:35]

So here's the line I keep coming back to. You don't just need it to be faster, it being AI, you need it to be dependable. Why is that the harder bar? Faster sounds harder. Faster is a fight every single day, given how quickly AI is changing.

Speaker: Matt [12:58]

Yeah, I mean faster is easy to measure. That's why everyone obsesses over it. But fast and dependable is really tough to get to. Dependable is tough to register on a chart, but you can't fake it, and that's exactly why it's the stronger position. Any competitor can match your publishing cadence if all you're looking for is to publish more blogs. You can absolutely dominate the number of blogs created in a sector. But if nobody's spending any time reading them, then they're not going to have any SEO value. In fact, they could have negative SEO value because people are going to be looking at them and bouncing away. And they're certainly not going to have any AI value if they're not adding value. So there's a big challenge there. In terms of the customers of our clients, making sure that everything is correct and dependable really is their reputations on the line, and it's massively important.

Discernment and the human warranty

Speaker: Jasmine [14:08]

That makes a lot of sense. And one of the words that I'm seeing used a lot more when it comes to the human in the loop with an AI process, with the AI output, is discernment. What you sometimes call the human warranty. Can you tell us a bit more about that?

Speaker: Matt [14:32]

Yeah, so it comes down to needing to understand what good looks like. And there is a level of experience needed to do that. But it's very easy just to pump things out. You could just easily go into ChatGPT or Claude, find a lot of things that are being spoken about, and then just push them out one after the other. And you could do that agentically. So you'd be able to very, very easily create a blog every ten minutes if you wanted to, or faster. Publish them, keep that publishing schedule going. Does it add any value? No. Are they going to be any good? No. Are people going to read them? No. It's going to do your brand damage, because they're just going to see this noise and they go, well, this stuff's rubbish.

So understanding what is good, and what good looks like, there's something very human in it. There's a gut instinct, a feeling. This is something I'm proud to put my name against. That feeling, I think, is the one that we have to chase as marketers in the AI era. Just because we might be using AI to augment what we do doesn't mean we should lose personal pride in the output. In fact, I think we should spend more time trying to get to a point where we can be really proud of what it does. And that's something that I really chase every day.

Synthetic customers as a different lens

Speaker: Jasmine [16:04]

Yeah, I think for me, discernment means applying your life experience in a particular area or in a particular role, not only to fact check, but to understand how your target audience is going to react to your deliverable. And I think this is where the synthetic customer platforms that you build are of extreme value. Can you tell us more about that? [reconstructed, please verify]

Speaker: Matt [16:37]

Yeah, so it can be really hard to walk in a customer's shoes. If my customer's wearing a size five pair of shoes and I'm a size twelve, it's really hard to spend much time in those shoes. I can try all I like, but I'm not going to be able to walk very far in them. I'm probably going to twist an ankle, particularly if they're high heels. It's just not going to be a good fit. So a synthetic customer can help you to look through the world in their eyes. Now is it going to be a perfect representation of your customer? No. But is it going to give you the lens to be able to look at it through a different perspective and ask, how would this customer react? And if you've built them with enough grounding evidence, it can really help shift the perspective away from, well, this is what I think, towards, well, maybe this is where we need to move to. And that shift can be incredibly helpful. Does it still need a marketer's discernment? Absolutely. Do we still need to make sure that it still feels good and we're proud of it? Yeah. But it's an extra test. And so each of these steps adds a step. But in going through these steps in a different way, it actually increases the quality of what we're able to create.

Speaker: Jasmine [17:57]

And more often than not, I found in using synthetic customers, it will raise issues with your output that you hadn't thought about in one instance, or that you had thought about but it sort of fell out of memory and you just kept pushing forward. So I really like the idea that it is an extra check, and it's certainly a worthwhile extra check because these synthetic customers are grounded in real voice of customer, grounded in brand guidelines, and so much more.

The check bit, making sure the message lands as intended

Speaker: Matt [18:39]

Yeah, and I think one of the things there as well is that it's not just an extra check. Sometimes it can be a check that, you know, so often we communicate and we believe that what we think we've communicated is what's received. But if an AI is receiving something that's different to what we're communicating, maybe our own lenses that we put on, our own belief that we've communicated something clearly, is somewhat distorted, because we know what we meant, and maybe that's not quite what we communicated. And so I think that shift is a really important one to get through.

And, you know, we don't often in human communication do what computers do, which is to have that check bit at the end of a block of numbers, a block of code. So you'd have the binary code, and then the last one is the sum of those as a check bit. And if that doesn't match, you know that there's been a problem in the communication. I think we need to try to use all the tools that we have these days to check our outputs more thoroughly than ever before. Because people are going to be using AI more and more to review our content, to look at our content. And so we have to be clear that, one, the AI perceives the content in the way that we want it to be perceived, so that when they do rewrite it, it gives the right message. And secondly, when people arrive on our pages and look at our messaging, we need to make sure that they clearly see exactly what we intend them to. And that we're not creating that, well, here's some nice marketing speak, we know what we're saying, but it's the view from the inside out.

Where to read the blog, and closing thoughts

Speaker: Jasmine [20:27]

Yeah. The power of clear communication cannot be overstated, especially in an AI world now. So I'd be remiss not to remind everybody that your full blog, "A brand is trust, and AI slop can break it fast", can be found on the strivenn.com website under Thinking, and the link will be in the show notes.

Speaker: Matt [20:57]

Thank you. And if the hard part for you is the bit we just described, building something that holds your buyers' real objections so your team can argue with them before anything goes out, that's what we do. We run growth consultations at no charge. Just visit us at strivenn.com and book in for a meeting.

Speaker: Jasmine [21:17]

And there's no time like now to remember your buyer is already checking. They're always checking. The job is to give them nothing to catch.

Speaker: Matt [21:28]

Your brand is trust. Look after it. The speed is worth having, but being dependable is the most important thing. And finally, it turns out that the spam folder occasionally does hold good news.

Speaker: Jasmine [21:41]

Yeah, exactly. Thanks so much for educating us on this really important topic of brand equalling trust. Thanks, Matt. And congratulations again on the CX50 award. Bye for now.

Speaker: Matt [21:50]

Thanks, Jasmine. Thank you.

Q&A

I publish a lot of AI-assisted content already. Where do I start if I want it to be dependable, not just fast?

Pick one live piece this week and run a single human pass with one person who knows the science. Pride is the test: would you put your name against it? Have them read the source behind each claim and confirm it matches. You do not need new software or budget. One reviewer, one document, one honest read is enough to find the drift before a scientist does.

How do I make fact-checking AI content practical at scale instead of a bottleneck?

Build the claim-by-claim table Matt describes. Ask your model to output one row per claim with the claim, where it sits in the draft, the reference, and the exact location in that reference. Then a human checks line by line in minutes rather than re-reading everything. Start with one white paper next week, save the prompt, and you have a repeatable workflow that makes the human pass fast and defensible.

We are a small team. Can we really afford a discernment step on every piece?

You cannot afford to skip it. One forgivable mistake is fine, but repeated sloppy errors are what erode trust with doctorate-level buyers. The cheapest version costs nothing: before anything ships, one experienced person asks whether they are proud to attach their name. That single question, applied to your next published piece, catches more reputational risk than any tool, and it takes minutes.

What is a synthetic customer actually good for, and how do I try one cheaply?

It gives you a different lens, not a perfect customer. Its real value is showing the gap between what you meant and what was received. This week, take one landing page and ask a grounded buyer persona how they would react and what they would object to. Even a rough version, built on real voice of customer notes you already hold, will surface an objection you missed before the page goes live.

How do I check that AI reads my page the way I intend, now that buyers use AI to summarise us?

Treat the AI as your check bit. Paste your page into a model and ask it to summarise the core message and rewrite it for a buyer. If the rewrite drifts from what you meant, the message is unclear, not the model. Try this on one key page next week. Fixing the gap costs nothing and protects how you appear when a buyer's AI describes you for them.