299: AI Won't Fix Your Knowledge Quality Problem - It'll Expose It; with Leslie O’Flahaven

AI can generate a clean looking knowledge article in seconds, but that speed comes with a trap: it makes bad knowledge easier to publish and easier to surface. Leslie O’Flahavan joins me to unpack a simple claim with big consequences for customer support leaders and knowledge management teams: AI won’t fix your knowledge quality problem, it will expose it.

We get practical about what AI can do well (drafting, simplifying language, reformatting for different audiences) and what it cannot do for you (understanding the real human reader who is stressed, new to the task, or missing context). Leslie explains why writing is more than editing and why relying only on revision can create “brain flab” as teams stop practising planning and organising. We also dig into a better way to judge content: stop asking users whether they like it and start asking whether they can complete the task with your self service help.

From there, we explore concrete tactics to make the reader real again: using daily ticket feedback as gold, building articles around customer questions, and designing knowledge base and FAQ structure that works for both humans and AI search. If you’re feeling AI fatigue, this conversation reframes the moment as an opportunity to do more meaningful knowledge gardening, not less.

Subscribe for more conversations on support leadership, AI in customer service, and knowledge base strategy, and if this helps, share it with a teammate and leave a quick review.

Transcript

Welcome And The Big Claim

Charlotte Ward 0:13

Hello and welcome to episode 299 of the Customer Support Leaders Podcast. I'm Charlotte Ward. Today, welcome Leslie O'Flahaven to talk about how AI won't fix your knowledge quality problem. It'll expose it. I'd like to welcome back today, um, after quite some time, uh Leslie O'Flahen. Leslie, lovely to have you back.

Leslie O’Flahavan 0:40

Um I'm so happy to be with you, Charlotte. It's across the miles and across the years.

Charlotte Ward 0:46

Again, miles and years is such a good way to put it put it. And uh I just loved the conversations we've had so far on the podcast, as you said, probably two years ago since we last recorded a session, maybe a little longer.

Leslie O’Flahavan 0:59

Maybe a little longer, yeah.

Charlotte Ward 1:01

Yeah, um, my fault. I know that. Um uh but uh thank you so much for agreeing to come on. You're just one of my favorite guests because you cover one of my favorite topics. In fact, two of my favorite topics, all bundled up. Uh good knowledge management and good communications, um, which uh the the bridge of which between is good writing, right? So um on that basis, I know who you are, but would you just mind doing a little introduction for our new listeners?

Leslie O’Flahavan 1:31

I'd be happy to, Charlotte. My name is Leslie, last name Oflahaven, and my company is called eRight. It's ephen W-R-I-T-E because you're right. I formed it in 1996, coming on 30 years next month, 30 years in business. My company's mission is to help people learn to write well at work. So many of my clients come from the support world, the contact center world, because we have some of the

Why AI Cannot Know Your Reader

Leslie O’Flahavan 2:00

most durable journeyman writers in these roles doing the best they possibly can to help unhappy customers find and use information. So, of course, we're living through uh an epic period in terms of the skills we use to write at work. And I'm here to support my clients and uh great thinkers like you to talk about how our world as workplace writers has changed because of AI.

Charlotte Ward 2:33

Ah, yeah. Well, thank you so much for the comp compliment that you she haunted in there. Uh unnecessary, but but thank you so much. Um uh what what a topic. Um, I know we're gonna open up an interesting conversation on this uh because this is uh good writing, good communications, good knowledge management is really been so um massively impacted by AI. And I would like to dive into that with you. Um I'm gonna look forward to it. So, Leslie, um I'm so looking forward to digging into this. Um I think I think, you know, as as I said before, the whole world of like knowledge management and writing well and drafting articles and customer replies and everything, all of that good writing that we've had so many expectations on our humans, particularly in support to deliver up to this point. That I get a pervading sense in the industry right now, certainly from non-support leaders in the industry that, and obviously vendors, sorry to say, um, but that like AI will fix all that. Like, you don't need to be good at writing anymore, you don't need to worry about the quality of knowledge, where it's stored, how it's written, who it's written for, all of those things, because AI is just going to solve it for us. So given that we're here talking about this, is it fair to say that you don't believe that's true?

Leslie O’Flahavan 4:02

I yes, it is quite fair to say. In fact, if I were getting a tattoo, that would be the tattoo. AI is not going to fix all of our writing problems, but it is going to fix so, so, so many of them. So many of them. But the the reason AI cannot fix all of our writing problems is because the AI does not know the human reader whether the um individual piece of writing is incredibly consequential, like a proposal for funding for um uh, I don't know, a nonprofit organization that helps children who are starving get sufficient nutrition. We want that proposal to succeed. We want the organization feeding the hungry children to get the funding. So we we have human readers who decide whether the organization gets the funding. And the AI can generate a suitable draft of that proposal far more quickly than any human writer can, with uh so much less effort on the human writer's part. But when all is said and done, um, the human writer needs to decide whether that draft, that AI drafted proposal is ready for the human reader's eyes, for the decision maker's eyes. And on a much, much smaller scale, when we write a knowledge article about how to install a uh you know, a widget in your website or something like that, or when we write a knowledge article about how to troubleshoot a problem you're having with software, yes, the AI tool can write that knowledge article, yes, but it is the knowledge about where human users will struggle to understand the information in the article or how much detail they need that only humans can provide. And the problem is that without repeated practice in fixing draft content, human writers will lose the muscles, the mind muscles they need to do to

Brain Flab And The Writing Muscles

Leslie O’Flahavan 6:20

fix it. They they'll get brain flabby. It's brain flab, and I'm worried about it.

Charlotte Ward 6:26

And you should be. I I agree with you. Do you um you know I I wonder about this kind of in a very similar vein, I worry about the brain flab. Because I because I wonder if even um we we develop brain flab if we're not even writing content fresh anymore. I wonder if validation is enough exercise.

Leslie O’Flahavan 6:51

It is not enough exercise. You know, my entire professional expertise, the masters I earned 400 years ago, the um everything I know as a subject matter expert is about the cognition of composition. What does your brain do when you're creating something, you're writing something that is readable and effective, that achieves the purpose you had for it. And the writing process has distinct steps in it because each this the steps are distinct because they require different types of thinking. So we need to plan what we write, we need to organize it, we need to draft it, and we need to revise it. These are different types of thinking. If the tool, if the AI tool offers you something that is drafted, you will have a lot of practice in revising. That's the step left to do. Yeah. Because you your prompt may have helped been a type of light planning, right? Your prompt is like planning a document, but you didn't organize it and you didn't draft it. And it's just not reasonable to expect that you're going to be super good at revising it if you didn't do the other steps. Yeah, I agree. If you don't do them very often, and in some near future days, if you don't ever do them.

Charlotte Ward 8:20

I I completely agree with you, and I I think that um I think the inherent danger in completely focusing on one end of the pipeline, as it were, is to your point, you lose the skills. And and actually I I think there's um there's something to be said for, you know, um you have to practice the thing to be able to judge the thing as always.

Leslie O’Flahavan 8:43

Indeed, that's that's exactly what I'm saying. Yes.

Charlotte Ward 8:46

Yeah, and um I think that I think that the important note here is that you don't need to revert to writing everything, you just need to continue to hone or um you know develop those skills independently as well, so that you can do you can manage the other end of the pipeline, right? So I don't think this is a reversion to having to produce everything in order to write well.

Leslie O’Flahavan 9:14

No, no, don't worry, I'm not saying we need a sharp pencil and a tablet full of lined paper, and that's what we need. No, I'm not saying that at all. And you know, um I'm no snob about writing, not at all. And I know that good enough is good enough most of the time. It's absolutely good enough most of the time. It's just not reasonable to expect busy people at work to continue to grow as editors if they're never producers. It just it's contrary to what we know about how people develop.

Charlotte Ward 9:53

So is there I I think there's a an element for me of like knowing when it's appropriate to throw something into an AI and just have it do, let's call it the grunt work. I'm sorry to use that in your presence when it when we're talking about writing, but you know, sometimes sometimes you just need stuff in a document, you know? Indeed, yes. As as a as the as entry level, like it's a lot of lift sometimes when you've unpacked a whole thought process to then go off and write it. Um, or you've written some code and then to go off and write the docs from scratch. But sometimes it's appropriate for AI to take the grunt work. I will use that again, um, and and for you to finesse and validate and you know, um QA. And then there are times when it's appropriate to slow down a little bit and and use um and practice those skills. And I think there's there's all sorts of constraints in play. Some of the constraints are time, some are expertise, and some are audience, maybe as well, right? But yes. How how do you how do you make that judgment call about if if you are buying into the idea that you have to write occasionally to be able to judge appropriately at the other end of the pipeline, how do you make the judgment call about when to apply AI?

Leslie O’Flahavan 11:12

I think I I I want to be realistic. I think people are going to apply AI, they're use AI tools to draft their writing, to go from nothing to something. I think they're going to do it all the time. I I don't I don't want to be some dreamer. I think they're going to do it all the time. Writing, writing is hard. Writing well is really hard, and we have a tool that does a a really adequate job in a lot of cases. What I think we need to do is to make the reader more uh real

When To Use AI Drafting

Leslie O’Flahavan 11:50

to writers because that will support them to edit well. Even if they're not getting practice in producing writing very much because the AI tools are, we need to make the writer, and pardon me, the reader, a real known presence in people's lives. And and this is a practice I advocate for all of my clients who have knowledge bases and who have customer support organizations is gather a team of of readers or customers you can turn to regularly. And and you know, you can buy this kind of willingness to share for very for a Starbucks gift card. We don't we don't need to um get a whole budget line going for this. We we need seven real human people who represent different types of customers, or as I'm using readers, those terms are interchangeable in our conversation today. And we need to show them what the AI drafted and ask them very simple questions like is this too long? Do we need more screenshots, or do I have too many here? Um, you're new to this task. Can you follow the steps? And I think when we make the reader live real and accessible to people who are using AI tools to write, they'll be able to edit to refine. Otherwise, they really won't. It's such an abstraction and it's too much work.

Charlotte Ward 13:24

I'm gonna propose something a little bit controversial to you on that one. I I think I think you're to be clear, you're 100% correct. I totally agree with you. You need the reader to be a known entity, to have you need an understanding of how they will consume, how they will perceive, what their priorities are at the moment they're gonna be reading, uh the their command of the language, all sorts of things, right? You need to understand that about your readers and customers. I'm gonna suggest that you can get that for free, not even for the price price of a five bucks gift card. Um, we get that in, I mean, in support, we have such the luxury of getting that feedback every single day from our customers in every ticket. You get a ticket that says, I didn't understand this document. You get a ticket where the customer says, even not so overtly, but like, you know, I I have this problem. You can you can have a little conversation with them where you discover they followed something that's a KB article or a doc or a pop-up, you know, in the product or whatever that that has led them down a merry path or that they got bored reading or whatever, right? Right, right. Um, or they didn't understand because it was too obtuse. Um, I think we have a real opportunity and a the luxury of a ton of feedback every day in support if you just have put 5% effort in and going to find it. Never mind five bucks in a gift card.

Leslie O’Flahavan 14:52

Right, right. But you're so right. But um, you know, in a cynical way, I'll come back and ask you do we respect that feedback now? And do we act on the feedback now? Usually we don't. Sometimes we do, some organizations do, but we often don't. We we attribute the the customer says, I tried to follow this knowledge article, but it's way too long, and I don't understand what you're talking about in section three. And we say they're confused. Not we confuse them, they're confused.

Charlotte Ward 15:26

I think that's a I I I agree. I thought I do think there's a little bit of a culture. I mean, that's frankly like that's the non-KCS approach, isn't it? You should be, you know, if you're please, please, dear listeners, go and look at the KCS website and learn the basics. But but you know, if KB articles, if we just focus on those for a moment, um if they are not working, we should be improving them live and in the moment. And yes, um that improvement can be human and it can be AI, right? And I think I think the other thing that I I the other sort of controversial point I wanted to put to you was um was around using AI to almost simulate some of that where you you can actually apply AI and give it a persona for what you know, for want of a better word, and say review this with this in mind, you know. Yes, how does this read for you know a uh a B2 English as a foreign language um reader? Um, or you know, how does this read for non-technical? How does this read for you know another AI agent reading it, for instance?

Leslie O’Flahavan 16:37

Yes, and I think the tools are quite good at that, surprisingly good at that. You know, if you prompt the tool to um adjust the complexity of what you've written for a reader you or customer you can describe well, you know. For example, this person has been an administrator in our software and now they need to become a developer or something like that. They're changing role. I think the tools are very good at that. They're they're um very good at making language more simple. I find I I do trust the tools for that. But you know, um they that's that's an individual AI user's prompt. I'm I'm saying that when we have this hu this small group of seven willing humans, and all of us in the company hear from them, then together we have to, you know, it it happened on a Tuesday afternoon. We had a Zoom with our customer focus group, and on Tuesday they told us these articles are too long. Right. And now we all receive that feedback, not just the one individual who's using the AI tool. We can all get their feedback.

Charlotte Ward 17:50

So do you think, um, do you think that that feedback I'm sorry to make life so difficult on this podcast for you, Leslie, but I've got another difficult question, another difficult proposal. Bring it, I'm ready, I'm ready. Do you think feedback gathered in a focus group on a Tuesday afternoon is um let's say accurate? I I I sort of, you know, if you ask enough customers what they think of your UI, for instance, you'll you'll get enough people saying, I want that red button orange, or I think you should use I tag like sort of okay. They just feel like they have to have an opinion, and there are plenty of those opinions that you know, yes, are not particularly productive, let's say, or relevant or helpful or in line with the direction of whatever, whatever we're taking the product in, or whatever, right? But I I still feel like the same could be true, a little bit of a documentation or a KB focus group.

Leslie O’Flahavan 18:48

Do you indeed it

Make The Reader Real With Feedback

Leslie O’Flahavan 18:50

could? Yes, I do, but I wouldn't ask them, do you like it? I would ask them, could you do it?

Charlotte Ward 18:56

Right, right.

Leslie O’Flahavan 18:57

And that's a little bit different. Could you do the thing you were trying to do with the help I offered you? Yeah. Self-service. You're right. Absolutely, folks. So you're gonna get, you know, exactly the same number of people saying that reverse text, a white text on a black background is impossible to read, as you're gonna get saying that black text on a white black background is impossible to read, and then you'll you'll want all your Starbucks gift cards back.

Charlotte Ward 19:25

Yes. I mean, um one of my one of my one of my favorite sayings is uh um the greatest human emotion is neither love nor hate, but the desire to edit somebody else's copy, right?

Leslie O’Flahavan 19:37

Right. Right.

Charlotte Ward 19:38

Which is true. So if you say, can you make this better, they're gonna say, well, if you could just put you know, change that word. But I think that's I I love what you said there, and thank you for that clarification that we're not seeking feedback, we're effectively seeking efficacy. And indeed.

Leslie O’Flahavan 19:55

It's this kind of UX, CX, you know, and all the other Xs, you know, got together in the room and and uh can you do what you wanted to do with the help I've written for you? Can you do that?

Charlotte Ward 20:09

Yeah. So why seven? Is seven the magic number?

Leslie O’Flahavan 20:13

No, it's the it's not a scary number. Yeah. Because, you know, actually the the UX mentioned that I just didn't, you know, then uh people start to wonder does this group of customers whom I'm drawing on for opinions, for feedback, does it need to be statistically, demographically relevant? And you know, we have we have 15,000 paying customers. How could I take the opinions of only seven? But I I just really want organizations to have live humans, they can ask about whether the knowledge content was helpful. So I I was being wicked and arbitrary. It's a beautiful mix.

Charlotte Ward 20:58

I love wicked and arbitrary, though. I I I feel like that's the name of the fourth volume of my autobiography.

Leslie O’Flahavan 21:06

Right.

Charlotte Ward 21:07

J just to be clear, uh, volume two is currently titled. I mean, I haven't written any of them, so but I I I have just to be clear, uh, volume two is my favorite title, working title, ranting and persiflage.

Leslie O’Flahavan 21:20

There you go. I read all four volumes, written or unwritten.

Charlotte Ward 21:26

I'll certainly be coming, be coming to you for feedback. Um no, that's great. Um uh no, it it's really useful to think of it like that. Um so can you do it is the question to ask this group, this group of the magic seven. Um beyond so how how do we what do we do with that, with the output of that question? Do we just take all I mean, I would love to say we take all of the verbatims, give them to everyone in the company, everyone reads them. Absorbs them, uh, takes it on board, and we go and make the world a better place on the output of it. Is that is that realistic?

Leslie O’Flahavan 22:07

Um, I think when the number of inquiries coming in that that reveals that we have a problem in the knowledge base, and and now, you know, just like the cartoon characters blink and the dollar signs show up in their eyes, you know, when we know we have a knowledge problem and customers are contacting us because they can't fix something themselves, I think we do act on it right away. But um, you know, just I don't want to uh get trapped in this, you know, focus group idea that I brought up. It really I was really thinking of it, I'm really advocating it to enable people who are using AI tools to write from scratch who will not be able to edit what the AI produces because they just won't have enough practice creating the information from scratch. And the live access to a live human reader is one kind of a safeguard for that.

Charlotte Ward 23:07

Yeah, it's that, isn't it? It's an it's it's a mechanism for um uh knowing your reader effectively. Indeed, yes, yeah, yeah.

Leslie O’Flahavan 23:17

That's what I was really thinking.

Charlotte Ward 23:19

Yeah, um, no, understood. And I think that um I I think what I'd like to sort of explore with you um is what other mechanisms we might have for knowing our reader, and and then beyond that, how do we apply that to, given everything we said at the start, which is AI can drop into this knowledge maturity model, this knowledge development process, the QA and everything at different points? It can draft, it can QA, it can rewrite and summarize and improve, and it can get things into a certain publishable state for different audiences. Um, first of all, what are uh other mechanics do we have for getting to know our reader? And then how do we decide where to apply AI on the journey?

Leslie O’Flahavan 24:08

Mm-hmm. I think we our readers give us uh uh gold and we we don't realize we're being given gold, but our readers give us their questions. And and um if you think about uh any kind of stored knowledge, whether a knowledge-based article, an FAQ, or even an email template as an answer to readers' questions, then when they give us their questions, we are really getting gold from them. So and and we could think about what a wonderful prompt we can we can create for an AI tool if we're asking it to draft an answer to an actual reader question, the way the reader would ask it. So as you and I were talking about earlier, you know, the the AI, the generative AI tools do much better if you give them a persona to write for. Well, what better thing than the user who has these questions? And I have uh long thought that we could replace the familiar structure of many knowledge articles, which is title, then some kind of summary, then uh procedure or instructions, then we sometimes have exceptions, and then we have linked resources. You know, we have about four or five sections. I love the idea and recommend it that we replace those sections with reader questions. That the that the knowledge article not have uh a heading, for example, that says uh resources, but it would be where can I find more information? And it doesn't have the heading procedure or instructions, it has the question how do I, and then we restate the title, how do I add a user who lacks the correct permissions or something like that? So this is this is a great way to um consider the user to be able to phrase their questions either because you just know them

Rewrite Knowledge Around User Questions

Leslie O’Flahavan 26:20

or because you're collecting the questions they ask. And and to write a consistent prompt for know uh stored knowledge that is is really good for AI search as well as human search.

Charlotte Ward 26:36

Yeah, I love that idea. I love that idea. Um and I feel like that um sort of instantly connects your reader to the written form as well, because it it feels like it's speaking to them rather than just providing information.

Leslie O’Flahavan 26:54

Yes, and it connects the writer who's whose needs we're discussing a lot this afternoon, who is, you know, more a person who's instead of uh moving from from thought to keyboard to product, they're more like operating a calculator. Yeah, yeah, yeah. And so if we we want to give we want the human who's prompting the AI tool to be able to represent the human reader as much as possible in the prompt and get a better product.

Charlotte Ward 27:26

Yeah, yeah, absolutely, absolutely. Um so we we've talked about AI drafting, we've talked about AI maybe helping us structure articles, um, and AI can do the validation as well, particularly if you if you structure the article in such a way that um it's already formatted to answer the questions and to go off and kind of almost uh you know, um, well, QA. QA, you know, uh I I guess I was trying to think of a word for like battle test, I suppose, is you know, not very nice, but yeah, go and try it out. Um so that there's there's multiple stages in this uh development of a piece of knowledge that we can apply AI to. My feeling is that we should somewhat randomise where we get humans involved and where we get AI involved. I feel like the right path to keep ourselves as human beings, as writers, in touch with every stage of the process is to get involved in every stage of the process at some time. That's still that still means you can apply AI broadly to every stage of the process for speed and for um scalability, all of those things. But but the as humans to to everything we said at the start, we shouldn't just concentrate on you know clicking the publish button.

Leslie O’Flahavan 28:53

Like you have to be involved in the yeah, yeah. You you do have to be involved. And and you know, I think about the team leads and other managers whose um efforts can help the human use uh the staff continue to understand what a good article is. What a we we talk about what a good self-service knowledge article is. We should talk about it a lot. Um, because we want we want, first of all, we want everyone who works for us to share the same sense of what good is. They may have worked at another company, and maybe that other company said these knowledge articles have to come in under 500 words or something really arbitrary like that. We want we want our team to agree on what good knowledge is, you know, so uh so that we can use these AI tools in the same way. And again, agreeing on what's a good example of communication is a human task. That is not an AI task.

Charlotte Ward 29:59

Yeah, yeah, I couldn't agree more. Um I and I think it's interesting now now that we're bringing this to the team side, um something that I've been sort of pondering certainly since rebooting the podcast, and and towards the tail end of last year as well, um, is and and everyone in the industry is, is what is the impact of AI on our jobs? Like actually, if AI is able to write the knowledge, surface the knowledge, answer the question, close the ticket, um, what is left for a support agent or a support engineer to do? And um I think the answer is increasingly but not entirely knowledge management. And and I think that it's um the way I've begun to phrase it is that support has to, the the the individuals, the teams have to s shift their centre of gravity away from being on the coalface with customers and almost taking a step back slightly to also be stewards of the knowledge. And so they be so they become even the support teams that think they've been doing KCS forever and are actually like constantly writing and improvement articles, improving articles, I think that there is a migration to be had uh to what I would what I've been calling sort of a much truer kind of knowledge work state rather than frontline with customers today. Of course, you still and should be working with customers, that's why you're there, and that's where you're um where the most value is to be had from your skills as a human being, right? But um, but a bigger portion of the work is going to be about this, isn't it? It's going to be about working alongside AI, uh, all of those stages. Um, I actually think that's quite interesting, personally.

Leslie O’Flahavan 31:52

I do too. I mean, you know, most knowledge collections I've encountered in my work are um, I don't know, they're kind of maybe 20% glittery and 65% shabby. And then I'm not describing that middle percentage. I don't have a good adjective for it. So the humans who may, there may be fewer humans because of AI, and they may have less fewer customer contacts per day because of AI. Yes, so true. But but they could turn their talents to deciding how to improve or jettison that shabby knowledge content. And this has been true for every knowledge base that I've encountered, you know, outside of a knowledge base with 20 pieces of knowledge in it that's you know, for a startup that just launched and and so and knowledge, knowledge uh gardening, knowledge maintenance and gardening does does not end even when the audit of the knowledge base has ended, because we need new knowledge products all the time, and uh uh we need to combine them. And so you're right. I I think that's a very a very productive way to think about how the likely loss of support jobs might also be a period of increase of knowledge management work. Yes. And you know, pre-AI, we wouldn't turn to a a frontline customer service agent who had been on the job a short time and maybe wasn't super well educated and was much better on the phone than on a keyboard. We wouldn't turn to that person and say, please help us clean up our knowledge base because we knew that that person did not have the wordcraft skills. But because of AI, they may not need those wordcraft skills. The tool may be able to fill in any gaps they have as long as their customer IQ is high and their willingness to do clean, accurate work is high. And

Support Work Shifts To Knowledge Gardening

Leslie O’Flahavan 34:10

you know, that's this this idea that um I well I'll just say I my career has been filled with well-intentioned, industrious people who struggled to write well. Filled. I have met thousands of people like that, smart, capable, industrious people who just struggled with writing. And to think that this tool, these tools might help them advance without their burden, I'm thrilled. I'm thrilled about that. That's beautiful.

Charlotte Ward 34:48

Yeah, I uh yeah, I've met so many of those people as well, you know, for and for so many reasons, you know, for for sort of individual challenges or just because it doesn't come naturally to their that you know to the the traditional remit of their day job. I mean, for me it's kind of the classic example is the engineer who can't write like customer-facing documentation, you know. Right. Um but it helps. Um and uh you know, I think that this is where being involved in the knowledge production should should be a a job that we make sure humans are applying themselves to so many parts of the life life cycle. Um because if you really are saying, yes, I manage knowledge, I'm the knowledge base manager, but really all you're doing is looking after six AI agents and you're hitting uh publish or approve or whatever at the end of at the end of a train of agents, well, that's not interesting work. And I think we lose all of the opportunities and all the value of humans there if that's what we consider knowledge management in the world of AI to be. Um that's not interesting work. And I think everything you're describing, getting involved at every part of the process alongside AI, is interesting. And uh humans should do interesting work.

Leslie O’Flahavan 36:12

Well, and then here's another kind of interesting work that uh we haven't talked about yet. And um, you know, we're not just writing with AI, we are writing for AI. So the tools that serve the AI tools that serve content to human users or within a knowledge system, they vastly prefer the knowledge base articles and the FAQs to other types of content because these products are so well structured. And if you think about it, the the search the search that we might uh the search terms we might use when we're using AI to search are almost identical to the question in an FAQ. So to the title in a knowledge article. So the the AI search tools are going like a waiter, they're going to retrieve the food and serve you the knowledge article quicker and more prominently than they're going to serve you other kinds of content, which might be uh more detailed, um equally accurate, but not structured as easily for the AI search to parse.

Charlotte Ward 37:27

That's uh such an important point on human behavior, actually, as well, because when I think about how I search with Google compared to how I search with Claude or with ChatGPT, they're quite different, even though I'm looking for the same content ultimately at the end of the day. You know, I might search in Google something like uh, you know, display settings iPhone. Right, right, right. Just some keywords. But when I'm talking to Claude or ChatGPT, I'll phrase the full question. You know, how do I change my font on my iPhone on this screen? And then it'll serve and it'll answer me in a fully formed, you know, response. But it it actually, to your point, if I'm asking the question a way that's already captured out there, it's much more likely to serve me the right answer first time.

Leslie O’Flahavan 38:20

Right. And it's it's going to serve you the KB article, whether it's correct or not. Because from the AI tools perspective, what what the human user asked and the information stored in the KB were very similar.

Charlotte Ward 38:37

Yeah.

Leslie O’Flahavan 38:38

They were so similar. So it's going to bring that KB article so that that kind of 65% detritus in your in your in your knowledge base, it's coming up. It's coming up, yeah. It's coming up, right? So there's there's all of the motivation we the strong motivation we feel to have worthwhile, clean, and readable knowledge base articles and FAQs because they're coming to the front ahead of other types of content that we have, you know, uh that we have on our on our websites and our in our vast collections of knowledge.

Charlotte Ward 39:18

Yeah, yeah. And I think this is where a KB structure and format has it has its strength. You know, it should, it's okay that it duplicates some more reference-based docs, right? It's formatted for different consumption, which brings us right back to the point at the start, which is being aware of who your audience is and what their context is, right? Right. Yeah.

Leslie O’Flahavan 39:40

And one of your most important audiences is AI search.

Charlotte Ward 39:44

Absolutely, 100%. Um thank you so much for your time, um, Leslie. This has been, as I expected, a wonderful kind of uh exploration through the whole ecosystem and how we can uh uh apply AI to all of these problems, but also how we need to apply

Writing For AI Search Plus Closing

Charlotte Ward 40:05

humans to all of these problems as well. It's so important. And there's a bunch of opportunities there, which um I hope has been inspiring to those people out there who are being feeling a little bit of AI fatigue at the moment because I know they're out there.

Leslie O’Flahavan 40:19

Indeed. And I think two of them are right here on the spot.

Charlotte Ward 40:25

Um all right, well, um AI fatigue and excitement at the same time. I think it's a double-edged sword, right? Yeah, I'm I I let I let my I let your comment just just play around my head for a little bit. And I was like, yeah, but no, I think it's it's both. It's like, okay, this reads like AI, but also I can't deny I'm using AI for everything as well.

Leslie O’Flahavan 40:48

So yeah. I I completely understand. And I'm not I'm not fatigued by worthwhile discussions like yours and mine today. I am fatigued by spin and by empty discussions. And I think just we are just talking about AI so much that there's going to be some empty discussions and some, you know, spin.

Charlotte Ward 41:12

Yeah, 100%, which which brings us right back to the start again, doesn't it? Which is, you know, the notion that AI will just solve all of your problems is is spin. Uh so yeah. Um, I I know that I've taken some uh some really strong ideas, not least about how to structure my knowledge base articles going forward, uh, and uh some hopefully some inspiration I can take to my team. And I know the audience listening will have as well. I expected nothing less from a conversation with you, Leslie. Um, what a delight as I expected. Thank you so much for coming on um after a bit of a break. Um, will you come back very soon and uh and talk about uh communications and writing and AI and all of the things again? Because I know you'll have a lot more to say. We come back soon.

Leslie O’Flahavan 41:56

I would be delighted to come back soon. I'll just between now and then I will just miss you.

Charlotte Ward 42:02

Oh, oh thank you so much. I'll miss you too. Um thank you so much for your time. Um looking forward to the next time. Thank you again. Me too. That's it for today. Go to customersupportleaders.com forward slash two nine nine for the show notes, and I'll see you next time.