Is AI Killing Trust in SaaS? (part 3 of 3)

AI can help you send more messages and create more content, respond to more customers, launch more campaigns, and really even handle more support queries. But scale is only useful when your scaling is true. And if your sales promise is wrong, AI scales the over promise. If your marketing claim is weak, AI makes it louder.

And if your service bot gives confident answers, but they're wrong, your customer doesn't blame the bot, they blame you. Welcome to Breakthrough SaaS Growth with the Jasons, the podcast where we cut through the noise and share strategies that actually drive customer-led growth. Each week, we bring you straight-talking conversations on adoption, retention, expansion, all things AI, and everything in between. Let's get started.

Welcome back to another episode of Breakthrough SaaS Growth with the Jasons, where we talk about all things customer success, growth, leadership, AI, and real-world. I'm based here in London, joined as always by my partner in crime, Mr. Whitehead. Say hello, Jason.

Hello, Jason. This episode today is part three of our three-part series on how AI can set people up to fail. In part one, we talked about AI and manufactured credibility, so how AI can help people sound more prepared, more informed, and more credible than they actually are. Yeah.

And in part two, we took a look into customer success. We talked about polished misunderstanding, you know, how AI can create account summaries and QBRs and health scores and recommendations that look executive ready, but really miss the story, miss the point. Exactly. And this last one today, we're going to really talk about more around sales, marketing and service, because this is where the risk can get even bigger.

AI doesn't help just one person sound more polished. It can help the whole company scale messages, claims and customer interactions really quickly. Yeah. And that's brilliant when the underlying content is accurate.

But if the claim is wrong or the promise is inflated or the answer is unreliable, AI doesn't solve it. It just spreads it faster. Absolutely. So let's start with the sales side of things, because this really isn't a new problem.

Our sales teams have always had to manage the line between making a compelling promise and over-promising. There's been a fine balance. Oh, it's usually fiction. And every CS leader listening to this has probably had that moment when they've looked at the customer or handover document or had that meeting and just said, who on the earth promised this?

Absolutely. And we've all seen it. And it's usually followed by a quiet sigh, a true Brit, a strong cup of tea, or maybe a cup of coffee. Or bourbon.

What AI does is changes the speed and I'd say the volume of it. A sales rep now can generate highly tailored outreach discovery follow-ups, proposal language, business cases, and exec summaries than ever before. They can do it so much faster. That can be really, really helpful, but it can also make weak qualification or loose promises sound much more credible.

Yeah, absolutely. Can you give me an example? Yeah. So let's say AI helps write a proposal that says we can cut onboarding time in half, or our platform eliminates manual customer follow-up, or you'll get predictive churn insight across your full customer base.

Now those might be directionally attractive claims, they might even be true in some situations, but can we, can our sales team defend them? Can our product team prove them? And can our post sales and customer success teams actually deliver them? can the customer actually achieve the outcome with their current data, team maturity and process reality?

And that's where the problem can start. Yeah, absolutely. You know, because the AI generated wording, it sounds strong and it sounds confident, almost arrogant sometimes and commercially useful. And, you know, my wife often says to me, she's like, I'm often wrong, but never doubtful, which is so true.

And I think that's really the case with AI, that it could be making a promise the organization can't actually stand behind. Exactly that. And I love that real world example there of you and the other half, Jason. And it's a dangerous position to be in, not for you and your wife, but for customers.

Because the customer doesn't care that the sentence was AI assisted. They heard a promise. That helped them then build expectations around it. And when onboarding starts, customer success or whatever we want to call that function inherits that gap.

Yeah. And that's the real commercial problem right there. You know, when the deal closes and you're starting to transfer the relationship from sales over and build trust and get going so you can be the trusted advisor, but you're starting with a trust debt. And that's the key phrase here, a trust debt.

And in SaaS, that debt always becomes due. Sometimes in onboarding, sometimes in the first QBR or first value review, sometimes at renewal, but it does always come true. Yeah. So let's move into marketing, though.

Where do you see AI creating risks there? I think today, more than ever, our marketing teams are under a huge, huge pressure. More campaigns, more competitors, more content, more channels, more personalization, and a need for more personalization, more proof points, and more demand generation. So, of course, AI is very, very, very attractive.

It can draft landing pages, email sequence, LinkedIn posts, comparison pages, webinar copy, nurture campaigns, case study summaries, et cetera, et cetera. But that's not the issue. The issue is when AI makes everything sound stronger than the evidence really supports. You know, and that's a big one.

I mean, how many times have you seen AI hallucinating or just making shit up? I mean, AI loves competent language and it loves to be confident. And it really does. It will happily produce.

We were talking about this earlier on. It says, yes, it will happily produce phrases like transform your customer journey, eliminate churn, achieve instant time to value, unlock effortless expansion. And I'm reading a list here. Yeah.

Automate your customer experience and you read it and think this sounds really good. And as a customer, it's very compelling. Right. It doesn't mean anything.

And more importantly, how true is it, if at all? Yeah. And can the business actually defend it? I mean, exactly that.

And that needs to become a standard review question not does this copy the AI generated sound good but can we defend this claim Do our customers actually experience this Can our product and our product team prove this Can our sales team explain it without exaggerating? Can our post-sales and customer success teams actually deliver it once the contract's signed? Because otherwise, marketing creates demand with one story, and the customer experiences something completely different, And that's not where we want to be.

Yeah. And I think that's really where, you know, companies confuse attention with trust. You know, you need both. You do.

And I think so, right. That confusion is real. AI can help you get the attention. But attention without truth is a very, very, very expensive way to create future disappointment.

And we've all seen this. The website promises transformation. The sales deck promises value. The onboarding process delivers admin.

and then everyone wonders why is the customer not excited yeah and that and that one hurts because it's true and it is it is true it's not always malicious and most companies are not trying to mislead customers the vast majority of companies are not but they're under pressure to sound bigger faster sharper and more differentiated and ai makes that easier and this is exactly why this review process needs to get stronger, has to get stronger even. Oh, absolutely. But let's clear this back to customer success, because sales and marketing over promises, those promises don't stay in sales and marketing.

They follow you. They do, absolutely. And they land straight away in post-sales. The customer arrives with expectations created upstream.

They expected faster onboarding. They expected automation to be easier. They expect value to show up quicker and they expect the product to fit their specific situation. And then the CSM or whoever, the accountant, navigate that gap.

Yeah, and that's tough. I mean, how often have you spoken with CSMs and like, you know, the expectation the customer have that we help them create is off the mark. And this really creates a lot of tension before the relationship even starts. Exactly that.

And that tension is critical. A customer might say, but your website said this. Your sales team told us this. Your proposal said we'd be live in four weeks.

The business case showed this level of ROI. The demo made it look simple. Now, the CSM is not starting from a neutral position. They're starting from a correction, and that's a bad place to start.

Yeah. And, you know, I think, you know, the CSM always starts from whatever expectations sales set. But when AI is helping sales and marketing set an accurate one, that's where you really get into trouble. And, you know, instead of helping the customer move towards value, you know, first the CSM has to spend time unwinding the overpromise while establishing trust and credibility to move forward.

And that's not a great position to be in as a business or for an individual starting to build those relationships. And it's wasted trust. Yeah. And it's also wasted time.

The CSM, CS team have to reset expectations. implementation has to explain constraints product can get pulled into awkward conversations our commercial teams can get defensive and most importantly the customer gets frustrated and all because that original claim sounded better than the operating reality yeah you know and i can make that worse because it can generate a lot of those claims very quickly you know which is a real challenge it is i think that's the key point this this is not just about one bad slide or one clumsy email.

It's about scale. AI can produce hundreds of messages very quickly. It can produce ads, landing pages, full websites, proposals, and follow-ups. And if those are aligned to reality, fantastic.

But if not, what you're doing is automating future friction. Yeah. And who doesn't love that? But Jason, let's pivot a little bit here and talk about service, you know, because this is probably where customers feel the impact most directly.

I think it is. And I think service is where this gets really, really important. Yeah. It's where the phrase confident wrong answer becomes very real.

A service bot can sound helpful. It can respond instantly. It can use the right friendly language. It can summarize your policy.

It can guide customers through steps. But if it gives the wrong answer, as many of them still do today, that experience becomes far worse than waiting for a human. Yeah. And, you know, when the customer may act on that answer, you know, they have trouble.

And I know in my case, when I've gotten the wrong answer or the bots been there, I get pretty pissed off at the company that they didn't quite train it right. That they said this was OK instead of actually providing the information I need. It is. And people have, you know, organizations have invested in this tech trying to make these bots better.

Yeah. It's still a long way to go. But it's the risk. If a bot, an AI bot gives us the wrong answer about a basic feature, that's annoying.

but he gives the wrong answer about billing cancellation team terms security compliance data handling or contract rates that gets really serious and i i've seen that today with a a broker in the uk where a message an automated message was next to meaningless it then said you can contact us between 9 and 5 30 and then they're back to me they said we should phone us up you know i couldn't phone you up at nine o'clock at night and your message was wrong. If your message has said something very simple instead of just a system error.

So you've got to think about these. And I think the customer will make decisions based on the answer they get. And then a human has to step in and walk it back. And nothing destroys confidence quite like, sorry, the bot's got it wrong.

Yeah, I love that phrase. Or it's a system problem. Oh, right, right. Yeah, it's never us.

That's not exactly a trust-building sentence that really gets me to want to renew and buy more. Absolutely. And this, I think a lot of people aren't thinking about this. And the customer doesn't separate that bot or automation or AI from the company.

They don't say the AI model misunderstood the knowledge base. They say, your company, you, the big you, told me this. And that's the bit that companies really need to get clear on. AI is not a separate experience.

It's part of your customer experience. Right. You know, I was at a meetup yesterday and we were talking about who is accountable. And basically you know as Americans who do you sue when the AI is wrong Do you sue the developer Do you sue the content that in there Do you see the people that are supposed to be the human in the loop And it a very valid question but it all comes down to this changes accountability It does completely.

And we had a conversation with a guest not too long ago about this. It is that trust and accountability is critical. If the bot says it, the company said it, and that's how the customer experiences it. So you need very clear rules around what AI can answer, what it can suggest and what it must escalate to a human and how quickly it can escalate that.

Yeah, I mean, that's so important, which kind of leads to the next next piece here. You know, what would you put into the high risk category and where should AI have strict boundaries? I think I'd start with anything that involves commercial reality, some money, anything that involves rights, risk or commitments. So pricing, discounts, refunds, cancellation terms, contract language, security, compliance, data privacy, roadmap, product limitation, service level commitments and performance guarantees.

These today are areas where AI needs to be very, very careful. And in some cases, I'd almost go, it shouldn't be making any commitment at all. Yeah, I agree with that. It shouldn't be able to commit you to it, you know, but it can still help.

it on. It can, absolutely. It can help route the customer. It can gather information.

It can explain where to find the approved real policy. And it can summarize knowledge-based content if that content is current and approved. Another big question. And it can say, I can help with general guidance, but this needs to be confirmed by a team, a much more powerful message.

And I think that's much better than confidently, confidently, sorry, making something up. You know, I like that. And I think that's really important. Like, here's the direction, here's the stuff.

And I think people, people understand, you know, we need further approval, we need further review. You know, I see things all the time, like, here's the direction, but consult your lawyer, consult your doctor, those sorts of things. You know, so, you know, I don't think what we're talking here is anti AI. I think it's more pro boundary.

It is completely. And we're still early days. And I think the point is not, let's keep AI away from our customers. That's not realistic.

What we're trying to do is make sure that experience is designed properly. Know where and what AI can resolve safely and where it should assist and where it should escalate. They're really the key things. I like what we're saying, but we're not trying to keep it away from customers because as customers, I go to AI every day and I'm using it.

When I'm prompting and understanding it, I understand the boundaries. But I do think in the support context too as well, So deflection is not the only metric that matters. You know, it's important, but it's not the only thing. That's a huge point, Jase.

It really is. A lot of service teams are measured on ticket deflection, ticket closure, cost reduction, and speed. And they all matter. But if you deflect the ticket and damage trust, you've lost.

You move the problem somewhere else, usually to the renewal conversation. Right. And let's be honest, you know, people, whether it was self-service or AI, people just want their problems answered quickly and easily. and they don't give a shit how it happens kind of thing.

Completely. But you know what? The more we're talking about this and some of the other conversations we have, you know, this really comes to me as more of a leadership issue as much as a technology issue. I think it is.

You're right. AI adoption is moving fast inside companies, sometimes faster than governance. Marketing's using it, sales using it, all the teams are using it. People are using it privately at home and using their own tools.

And suddenly you've got AI generated content touching prospects and customers in many different places. Yeah. And what I'm seeing is that, you know, this is happening, but no one's owning the overall risk. And organizations still aren't clear on, you know, who gets fired.

You know, kind of like the people were saying who gets sued if your AI says the wrong thing. People aren't really talking about who gets fired if this is delivering issues for us. This is where we need leadership to step in. The question is not just how do we use AI to move faster?

The better question is what we are now saying at scale might not work. What are we saying that we might not be able to defend with our customers? And that's really the leadership question. We need our leadership team to start looking at that.

Yeah, I like that because I think leaders are a lot of them are like, AI is great. It's going to save us money. It's going to be our growth director, all this great stuff, which it absolutely can be. But there are those strategic questions in there because it does force your business not just to look at productivity, but also what are the claims it's making?

What's the impact on experience? And is this short term successful but long term going to harm us? I think that's so true. And I'd want leadership teams to review three different things.

One is customer facing claims. What are we promising in marketing, in sales, in onboarding, in CS and in support? Secondly, it's proof. Which claims are backed up by evidence, customer outcomes or product capability?

And thirdly, it's around ownership. Who is accountable when AI generated content reaches a customer? If the answer is everybody, then the answer is usually nobody. Exactly.

You know, from those of us who used to do the racy charge, you know, you can only have one person. And without that, you're just making fiction. And I think that's where problems start to get missed until the customer points them out. Yeah, and your example of racing matrices is a great example.

We've all seen them. And this situation is never, never ideal. Because once a customer points it out to you, you're already in that, how do we fix this mode? Yeah, absolutely.

And I think we should make this a little more practical. We've talked at a high level here. Jason, what do you think companies should do differently? I think the key thing is with a lot of these things, as we're starting off, keep it simple.

Yeah. First, create claim checks. Any AI generated material should be reviewed against these three questions. Can we prove it?

Can we deliver it? And would our customers agree to it with it? And if our answer is no, either soften it, qualify it or remove it. You know, that would help so much.

And at this event I went to the other day, we were talking about human in the loop. And I love the way the guy phrases like, I view AI as trying to parent a five-year-old child. You can't let it run wild. You have to guard rails around it.

And instead of human loop, he said, you know, you need to be the adult in the room. And I think that's really a great way to put it is that you need to have someone be the adult in the room here. And, you know, I don't see that happening a lot. I think that true And it is We talked about AI enabled customer success but it it a partner right now You are still there helping it training it and looking after it Yeah.

The second thing we'd look at is making sure that we've got escalation rules for service AI. Don't leave it vague. Define what AI can resolve, what it can explain and what it's got to escalate, especially as we said around billing, legal, compliance, security, and also customer specific commitments. Yeah, I think that's so true.

And managing those exceptions effectively is really important. So what else? I think the third thing is make sure we're connecting those pre-sales promises to post-sales reality. If sales and marketing are using AI to generate business cases, there's white papers, proposals.

We need to make sure the wider team has visibility, collaboration, break down those silos, because ultimately it's the post-sales team that are responsible for delivering much or all of what's been promised. Absolutely. I'd love an AI bot that can really come in and call bullshit in real time during the sales discussions, whether it's as you're having that conversation, it's going off the transcript or in the proposal that they could say unsupported, unsupported. But that's my dream.

And I think what you're talking about here is really that old alignment problem. It's just got a new AI wrapper around it. As we're talking about this, that's exactly what I'm thinking. I mean, this is no different, really, to what we've seen before.

And we've talked about this before. AI is the next. It's a big evolution, but it's a big step. But it's that next thing in technology.

A lot of the way business works is still the same. So the problem is not new. what AI is doing today is making those consequences faster. The fourth thing I'd say is we've got to make sure we train our teams and people to not just verify, not sorry, not just generate, but to verify the flip that.

So not just generate, which I think a lot of the focus is on that now. Absolutely. It needs to be that human check. Most AI training today is focused on how to prompt better.

That's great. But we really need people. And I see this with my kids as well. How do you challenge the output?

Where did that claim come from? Where's the evidence to support it? What is the evidence? What's the context that's missing?

And again, we've talked about that before. That is critical. And I think this is where responsible AI use becomes real. Yeah, I think that's so important.

And also recognizing, I think there's a tendency that as it improves over time, you trust it more. But most folks forget that your AI responses can degrade over time. So you can't trust it in perpetuity. But Jason, I think you had one more.

So what's your fit? The last one is measure trust, not just speed. And that's harder to do, but it's still critical. If AI helps you respond faster, but customers trust you less, that's going backwards that's not progress right look at resolution quality customer sentiment escalation rates reopen tickets expectation research during onboarding things like that and these signals tell us whether ai is improving our customer experience or just making the machine busier yeah i think that's really important here you know so for me the big takeaway from this final episode of the three in the series is that ai can help sales marketing and service teams move faster.

But faster isn't always better. That's the same as every other team. It can help you scale and help you grow. But if it helps you scale accurate, useful, honest content and communication, that's really powerful.

If on the other hand, it helps you scale inflated claims, loose promises, unreliable answers, that's a trust problem. And customers don't experience that as an AI problem, as we said, they experience it as a problem with your company. Absolutely. And that's really the key there.

Okay. So to wrap this up, I think this has been a good series. In part one, we talked about manufactured credibility and the situation that we had when someone reached out to us and said they loved our episodes and then admitted they hadn't done anything. And in part two, we talked about polished misunderstanding and customer success.

And now in part three, we're really looking at bigger operational risk and scaling the wrong message and the problems that can have. And I think that's the thread that's run through this kind of mini series, if you want all three episodes. AI should help people be better prepared. It should make teams feel sharper.

It should help companies serve customers more efficiently, but it shouldn't be used to come up with fake preparation, inflate capability, and avoid judgment. Because in technology and SaaS, and I think in broader companies and commercial reality, Trust isn't a soft thing. It affects adoption. It affects retention.

It affects expansion. And ultimately, it affects commercial reality and revenue for us. Absolutely. Okay, so let's come to our breakthrough growth question.

And so here's our breakthrough question for today. Where is AI helping your company scale communication with prospects or customers today? And what is the one claim, promise, or answer you most need to verify before it reaches the market? I love that.

Look at your sales emails. look at your marketing pages, look at your proposals, look at your service bot responses, look at your onboarding promises. Pick one of those areas and ask, can we prove this? Can we deliver it?

And would the customer agree? Because AI can help you go faster, but if you're scaling the wrong message, faster means you're going to hit the wall sooner. Absolutely. All right.

Thanks for listening, everyone, to part three of this three-part series on how AI can set people up to fail. And if you found this useful, please follow Breakthrough SaaS Growth with the Jasons of on YouTube, LinkedIn, come to our webpage and sign up for our mailing list and share this series with someone in SaaS who's trying to use AI responsibly across the customer journey. Hi, this has been a great conversation. I think doing it across the three different episodes has made a lot of sense.

And if you as our listeners know anyone that would make an amazing guest for us or a sponsor that'd be a strong fit for the audience, do send them our way. And until next time, do keep closing the gap between what gets sold and what your customers are actually achieving. Thank you. Thank you.

Thanks for listening to Breakthrough SaaS Growth with the Jasons. If you enjoyed the episode, follow our show and please share it with someone else in SaaS. You can connect with us on LinkedIn, YouTube, or at BreakthroughSaaSGrowth.com.

And if you know a brilliant guest or a sponsor who'd be a strong fit for the audience, send them our way. Until next time, keep closing the gap between what gets sold and what customers actually achieve.