Transcripts
Hey Megan, let's do a podcast. Great idea. What
Speaker:should we talk about?
Speaker:Today we are talking about the technology and experience
Speaker:infrastructure section of the AI Implementation Canvas.
Speaker:I didn't mess it up. It is a mouthful.
Speaker:It is so. So I know the implementation canvas really
Speaker:supports a cross functional team and working together and across
Speaker:boundaries. There are some sections of the canvas though that make
Speaker:sense for one or another function to lead, whether
Speaker:it's IT, the people, the business and so on. But
Speaker:I really like how the canvas provides questions and perspective
Speaker:for everyone to be involved in learning together. Yeah,
Speaker:you know, this is really one of the silo-busting
Speaker:features of the canvas is that everybody
Speaker:comes together to talk about these topics, even though different
Speaker:groups and teams within the organization might lead at different stages.
Speaker:We're at a point right now in the AI
Speaker:transformation space, not just within an organization, but
Speaker:societally in which old structures
Speaker:that used to exist are starting to become more fluid. It
Speaker:becomes easier to learn about other things. It becomes
Speaker:more imperative that we talk to each other across
Speaker:organizations in order to get this right. So the Canvas
Speaker:questions in each of these planning dimensions
Speaker:help non-subject matter experts, people who are not part of that
Speaker:function, engage with the subject matter experts in each area.
Speaker:In the book the AI Implementation Guide, there
Speaker:are series of questions that you can ask people you might want to invite
Speaker:to the table or to go find if you need help in any of
Speaker:these things. So that's a really key
Speaker:component here in the book. I call them planning dimensions,
Speaker:but really it's, it's 14 conversations to be having or
Speaker:topic conversations. And so you know, here we have these 14
Speaker:different sections or different planning dimensions or conversations
Speaker:and we've chunked them up into four different sections on
Speaker:the canvas. There's Strategic Foundations, Design and Implementation
Speaker:Enablers, Human Centered Adoption and Change, and as you
Speaker:noted, the
Speaker:rather-difficult-to-say-all-at-once technology
Speaker:and experience infrastructure. You said that fast. No
Speaker:problem. No problem for you. I say everything fast. It's a lot of caffeine here,
Speaker:Meg. Keep up. All right,
Speaker:so we hear a lot of talk about the technical aspects of AI
Speaker:implementation in organizations, but isn't that IT's role?
Speaker:Like they're the technical ones? How do we think of learning leaders
Speaker:and business leaders engaging in that conversation? You
Speaker:know, I feel like Meg, in order to be responsible partners, we can't
Speaker:just say, "oh, I don't know, it's technology!" And particularly
Speaker:when we have a technology that itself is
Speaker:kind of a black box. Right. It helps us
Speaker:become better partners with the IT team.
Speaker:It helps us become better stewards of our organization's
Speaker:resources and advocates for
Speaker:people who work in the organization if we participate
Speaker:and seek to learn. At
Speaker:the same time, I think that our ability to strategically
Speaker:ask subject matter experts questions about their work — that's something
Speaker:we've learned as learning design professionals, right? — helps
Speaker:them articulate their positions,
Speaker:what they're learning, and what they're doing to other people.
Speaker:And I will just say I
Speaker:have been in a number of situations in which the business
Speaker:leaders in the room didn't understand the technology or its capabilities
Speaker:and either didn't
Speaker:know the extent to which they weren't aware or were uncomfortable
Speaker:surfacing that. So we had this ability to come in as
Speaker:like, "hey, I'm here for the learning team and I'm here to help. How does
Speaker:this thing work?" And just know that there's probably other people around the
Speaker:table who had that same question and didn't have the ability or
Speaker:audacity in the moment to ask that question. Yeah. That makes me think about
Speaker:how as instructional designers, we come in
Speaker:and we're so used to being in that space of being
Speaker:comfortable not knowing. Cause we rely on those subject matter experts
Speaker:in those areas to come with the answers and tell us
Speaker:what they are. Okay. That's like. That's our superpower, right? Like one of our superpowers
Speaker:— we have lots of superpowers. We don't often... We live in them so often
Speaker:I don't think we necessarily appreciate them. But in the organization,
Speaker:we're the people who repeatedly learn new
Speaker:things well enough to teach it to other people. And that's
Speaker:a muscle that not everybody gets to flex. Yeah.
Speaker:Yeah, that's cool.
Speaker:So even though we're talking about infrastructure, there's like a people side here
Speaker:to consider, right? I know in an earlier
Speaker:episode, you talked about how "people" is, like, all over the Implementation
Speaker:Canvas. So people need to be able to use these AI
Speaker:systems and solutions in the workflows and to support the
Speaker:tasks that they're doing on a daily basis. Tell
Speaker:me more about how we're planning for that people side — that user experience —
Speaker:here. You know, this one, Meg,
Speaker:was something that emerged from the research as
Speaker:I was asking people about
Speaker:their stories with AI implementation in their organization.
Speaker:And we experienced it ourselves, right? So
Speaker:some of the research and stories that I heard from people, like "really cool
Speaker:technology, but didn't account for the fact that
Speaker:people may work in a certain way." So here's. Here's a story, and I actually
Speaker:heard this story. This is a synthesized story. I heard
Speaker:from two places. It was almost eerie. It was the same week's worth of
Speaker:interviews. And I was like, "seriously, do you two people
Speaker:work together?" And they don't. It was really eerie, but almost the exact same
Speaker:story. You have customer service personnel and they rely
Speaker:on a knowledge base, right? So in the moment, they've got a call, they
Speaker:don't know what to do. They search their knowledge base. And if they can't find
Speaker:an answer, then they escalate the call to a supervisor
Speaker:or a more experienced person on their team. And the business's
Speaker:goal is to have as few escalations — right,
Speaker:escalating up to a supervisor or manager or a more experienced person — as
Speaker:possible because, well, there's a couple things, right? It helps relieve
Speaker:the manager, it gives agency to the call center rep, and
Speaker:for the customer, they get a first call
Speaker:resolution. They get somebody who knows what they're doing.
Speaker:And so each of these organizations was
Speaker:trying to give better tools to the service
Speaker:reps so that they could solve those
Speaker:problems on a first call and not have to escalate.
Speaker:In both cases, this is crazy, right? In both cases,
Speaker:what they did was they were moving from a system that
Speaker:you could search with keywords and you had to learn the keywords, and
Speaker:that was a barrier to learning and becoming fast at your job. But once you
Speaker:knew that, you could like cruise through that knowledge base because you
Speaker:knew the knowledge base like the back of your hand and you had the right
Speaker:keywords. And they replaced that with a chatbot
Speaker:assistant that required that you type like something
Speaker:much more like a whole sentence into the chatbot. So you're
Speaker:actually slowing down the service rep
Speaker:and forcing the service rep to think about how to query
Speaker:this properly in order to get a consistent answer over and over
Speaker:again. And consistency of answers, by the way, is something that large language
Speaker:models don't necessarily do well unless you've really throttled
Speaker:them. Probably a topic for another conversation.
Speaker:So what we were actually doing in these two — not we, but like these
Speaker:companies — these two stories, was slowing down service reps.
Speaker:And actually they didn't change
Speaker:their escalation metrics. They weren't able to
Speaker:improve the first call resolution because they didn't think about this
Speaker:user experience component. And there are
Speaker:other examples. Even on our own team, right, Your own tiny team. Early on in
Speaker:the days of AI, we had like making all these GPTs and doing all this
Speaker:really cool stuff. And by the way, we have a really, really fantastic set of
Speaker:really well-honed GPTs with very specific instructions and best
Speaker:practices, and people look at our best practices like, "wow, that's pretty cool!" Probably another
Speaker:episode. But
Speaker:I remember one of our
Speaker:senior learning designers, Lauren, said, "yeah, but it's a pain in
Speaker:the neck these. Because I constantly have to alt-tab over and copy and paste and
Speaker:do all this stuff, and it's not integrated into my work. It's
Speaker:yet another tool I have to go and use." Yeah.
Speaker:Interesting. User experience: always
Speaker:and forever important. Projects are going to live or die based on user experience.
Speaker:I can imagine that there are whole new governance practices
Speaker:that are newly important in the organization when it comes
Speaker:to AI too, right? Yeah... And doesn't that sound so boring? I mean,
Speaker:governance. Yeah, I know, like rules.
Speaker:You know, I think governance, all kidding aside, governance is
Speaker:becoming really, really important. It is important. I think it's a word that like five
Speaker:years ago I never heard of. Yeah, I didn't think much about it.
Speaker:At least not organizational governance, right? But this is.... So
Speaker:let's define it. Yeah. It's how we evaluate, approve, and
Speaker:monitor the various systems that an organization uses. The data
Speaker:they have access to, the people who have access to them, and the impacts and
Speaker:the actions that we allow those tools to do.
Speaker:What this does is, in a
Speaker:well-governed environment, there is a team,
Speaker:a function, and a set of processes, as well as
Speaker:organizational culture, that helps ensure safe,
Speaker:compliant, transparent use of AI tools in a
Speaker:way that the organization.... What it means is the organization has
Speaker:your back when something goes wrong because they've had your front
Speaker:and by upfront looking at those tools and making sure
Speaker:that they are going to be doing what you expect them to do. So
Speaker:sometimes people say like, "oh, I can't use that tool. It's not been approved at
Speaker:work." And it's really annoying because you can't use the coolest
Speaker:new tool that you find on the web because it's not been approved.
Speaker:I really like knowing that before I start using a tool
Speaker:or opening a browser on my, you know, on my computer that
Speaker:it's going to be safe to use. I don't know about you
Speaker:— we have a lot of our clients' sensitive data. It
Speaker:really does behoove us to make sure we're taking really good care of that.
Speaker:Yeah. So I think my takeaway here is
Speaker:that when we're talking about technology, it's more than just the technology
Speaker:itself. It's more than just the code or
Speaker:the hardware. Totally, totally. Right. It's technology,
Speaker:it's people, it's the user experience, and
Speaker:the rules — the guardrails keeping us
Speaker:safe. Yep. I like being safe. What's one more thing
Speaker:you got for us, Megan? There's always one more thing, isn't there? There is.
Speaker:Hence the well-caffeinated life. There's always one more thing.
Speaker:You know, I think one of the things here is that
Speaker:there are multiple roles for the learning team
Speaker:in this particular section. Because not only can we
Speaker:be advocating for the user
Speaker:in user experience — we don't have to be a UX designer to be advocating
Speaker:for the user. We build the training for this, right? And
Speaker:we see these problems, I think, earlier than perhaps
Speaker:other functions do because we look at things differently. Again, part of our
Speaker:learning designer superpowers, right? But
Speaker:also, right, so we're participating on the project, but then also, there may be
Speaker:learning and communication efforts associated with governance, associated
Speaker:with the user experience and how the workflows are changing.
Speaker:So there's multiple roles for us here.
Speaker:How'd that go? You know, I wonder if we
Speaker:overplayed the difficulty or the perceived difficulty of saying
Speaker:technology and experience infrastructure. There's a lot of big words, but
Speaker:they're also words that we deal with all the time.
Speaker:Yeah, well, and they don't end or start with the same sounds,
Speaker:which is usually what trips people up. I love how you see
Speaker:that. You see things I don't see. Thank you. You're welcome.
Speaker:This is Meg Fairchild and Megan Torrance, and this
Speaker:has been a podcast from TorranceLearning. Tangents is the
Speaker:official podcast of TorranceLearning, as though we have an unofficial
Speaker:one. Tangents is hosted by Meg Fairchild and Megan
Speaker:Torrance. It's produced by Dean Castile and Meg Fairchild,
Speaker:engineered and edited by Dean Castile, with original
Speaker:music also by Dean Castile. This episode was
Speaker:fact checked by Meg Fairchild.
