Architecting AI: Why Knowledge Management Platforms Matter

Join Gary and Scott on The Macro AI Podcast for Architecting AI: Why Knowledge Management Platforms Matter. They break down Knowledge Management Platforms (KMPs)—smart hubs that turn chaotic enterprise data into organized, actionable knowledge. For leaders, KMPs cut the 20% of time workers waste searching for info (McKinsey), boosting efficiency. Zappos slashes support costs 20% with a KMP-powered chatbot, while Delta gets planes flying faster with instant maintenance logs. Scott highlights Toyota’s supply chain wins—spotting bottlenecks early, saving millions.  

The duo debates build vs. buy: third-party KMPs (Bloomfire, Confluence) win with seamless integrations (Salesforce, SAP), scalability, and security (encryption, 99.99% SLAs), outpacing in-house efforts. Gary and Scott then unpack the tech: NLP with BERT models powers semantic search, knowledge graphs (Neo4j) connect data, and ML (LSTMs) keeps it fresh—all in a stack with vector databases (Pinecone) and microservices (Kubernetes) for sub-50ms latency. Capital One’s Eno chatbot (80% automation) and Siemens’ downtime cuts (15%) show it in action.  

KMPs architect AI success—structuring data for faster training, linking silos for insights, and scaling globally, like Shopify’s merchant support. Gary and Scott wrap with a playbook: audit your data, pilot a KMP, and scale. Whether you’re a strategist or a techie, this episode builds the case for KMPs as your AI foundation. Subscribe for more! 

About your AI Guides

Gary Sloper

https://www.linkedin.com/in/gsloper/

Scott Bryan

https://www.linkedin.com/in/scottjbryan/

Macro AI Website:

https://www.macroaipodcast.com/

Macro AI LinkedIn Page:

https://www.linkedin.com/company/macro-ai-podcast/

Gary's Free AI Readiness Assessment:

https://macronetservices.com/events/the-comprehensive-guide-to-ai-readiness

Scott's Content & Blog

https://www.macronomics.ai/blog

Transcript

00:00
Welcome to the Macro AI Podcast,  where your expert guides Gary Sloper and Scott Bryan navigate the ever-evolving world of artificial intelligence.  Step into the future with us  as we uncover how AI is revolutionizing the global business landscape  from nimble startups to Fortune 500 giants.  Whether you're a seasoned executive,  an ambitious entrepreneur,

00:27
or simply eager to harness AI's potential,  we've got you covered.  Expect actionable insights,  conversations with industry trailblazers  and service providers,  and proven strategies to keep you ahead in a world being shaped rapidly by innovation.  Gary and Scott are here to decode the complexities of AI  and to bring forward ideas that can transform cutting-edge technology  into real-world business success.

00:57
So join us, let's explore, learn  and lead together.  Welcome to the Macro AI Podcast. I'm Gary Sloper here with my co-host as always, Scott Bryan, delivering the latest AI solutions to transform your business and compete globally. Whether you're a C-suite leader shaping your AI strategy or a tech enthusiast craving the details, we've got actionable insights for you. Hey everyone, Scott here.  Today's episode is Architecting AI.

01:25
why knowledge management platforms matter.  We're diving into knowledge management platforms, or also known as KMPs, and talking about why they're essential for building AI that actually works.  So expect some  business wins. We'll deep dive into the tech a bit and go over some clear cases why KMPs are your foundation for enterprise AI. So Gary, let's jump right in. Okay, great. uh

01:53
Maybe what we do Scott first is just kick off the basics around KMP. ah You know, the background is, and  what I'm seeing is businesses are really drowning in knowledge, right? There's knowledge everywhere and all different facets of various  outlets internally. You have customer data, you have operations manuals, you team expertise, but what I see it's often messy. So the data is housed in various silos.

02:22
outdated files, ah many are no longer valid to the current business.  So in your opinion, maybe take a shot at defining  what knowledge management platforms are, KMPs, and kind of why they matter. Certainly. Yep.  So knowledge management platforms or KMPs are  centralized systems, and they pull all that scattered information into one smart usable hub.  And they're not just storage like SharePoint.

02:51
but some campies can plug into SharePoint. actually can, campies can actually organize, update and deliver knowledge so that your teams and your AI, your AI engine can actually put it to work. So it's like the organizer or the architect of your enterprise brain. That's a good way to put it. It's really architecting order out of disorder. So for leaders, this is about ultimately efficiency.

03:18
McKinsey has said, you know, knowledge workers lose 20 % of their week, just hunting for information. I mean, think about all those legacy SharePoint folders that existed for many years.  A KMP can really slash that and move your teams into higher productivity, letting your people focus on strategy over, you know, ultimately scavenger hunts, I think. Yeah. Yep. Yeah. Let's, take a look at a real world example. uh

03:47
So uh Zappos, I think probably a lot of people are familiar with an online retailer and they're really renowned for customer service. So their KMP continuously centralizes updated product details, FAQs,  and their customer service reps or a chat bot can grab those answers quickly. So the efficiency gained equates to cut and support costs of up to 20%. So it's...

04:14
You turn all of that knowledge that's in the KMP and turn it into action. So it's not just sitting on a shelf or really filed in a folder somewhere. That's interesting. I was researching a little bit and one use case that came up as Delta Airlines. It's another good example. It's not specifically related to call center customer experience, but really in their example, mechanics needed detailed maintenance logs, which we all know is very important when you're getting on a plane.

04:44
And really need to understand when plane is out of service So in their instance camp, he's put a lot of that necessary and highly pertinent information Right there for the mechanics so they didn't need to dig they need to go search or or make phone calls and that ultimately allowed you know Delta's planes to fly sooner and In an industry where minutes mean millions. That's a game changer, but more importantly just from a customer experience standpoint We've all waited at the gate

05:13
because there was a maintenance issue or something that needed to be cleared by maintenance. Yeah, that's a great operational example. ah Here's one in which knowledge management platforms connect the dots across teams for predictive gains. Everybody is familiar with Toyota, huge global company. ah

05:33
a company that many people view as a global benchmark for supply chain excellence. And Toyota uses a knowledge management platform to tie supply chain data together, spotting bottlenecks early. So picture a dashboard showing a delay from a tier two supplier in Southeast Asia, say for example, a semiconductor storage shortage, and the KMP will flag it instantly, letting the planners adjust before it halts a factory line all the way back in Kentucky. So for business, it's

06:03
smarter decisions  for AI, if built correctly, it's the structured foundation that AI really craves and that's why KMPs matter. Right, yeah, exactly.  They're the base  layer of operations in AI. So maybe we should dive into how you build it or buy it. Yeah, build versus buy. So that's a common showdown. uh So once a use case is identified, uh

06:32
You know, build a KMP in-house or buy a third party solution is the question. And for a lot of enterprises, maybe not as big as Toyota in the previous example, but  Toyota as of today's date, maybe the best bet may be to buy a platform. So we can, we can break that down a bit.  Yeah, I'm on board with that. So in-house sounds tempting.  basically have, and it's no different than a lot of things that have happened over the course of tech ecosystem history, right? Even cloud.

07:01
So in-house,  it really seemed  like something that you would want to do. So you have custom design. You ultimately have full control over it, but it's a massive lift. So you're not just coding. You need the AI expertise, scalability, security, which obviously is a big one.  And some third-party KMPs like Bloomfire, I have notes on Bloomfire versus E-gain, I will post later,  deliver that out of the gate, saving.

07:29
huge amount of time and a lot of internal headaches that could come out and potentially  hamper your organization. Yep.  And certainly money. Money is a  key component there. So building it in-house  eats resources. You're hiring specialists, uh developing. It's a distraction from your core business. uh Just going to another example, uh Shopify. They went for a third party to launch a KMP to support their 1 million merchants in the Shopify network.

07:58
And then 1000 support reps who assist them.  And they were able to basically build it,  quickly scale it. And they were able to stay focused on growth of the business. So that's a strategic win in the  build versus buy debate. Yeah, I don't disagree. mean, the integration is huge.  Vendors plug into Salesforce, SAP, Slack, et cetera.  You your team's collaborating instantly.  But if you build it yourself, you're really stuck coding those APIs and troubleshooting and testing.

08:29
especially while those deadlines pile up. I've seen companies  stall on that and potentially  great use cases that they had put down on paper turn into wasted resources and ultimately a potential business valuation  evaporates.

08:48
Yeah, certainly. um scalability is another angle to look at and debate. So some third party campies  handle  massive  user loads without breaking a sweat. So often they're leveraging distributed indexing like Kafka or Elasticsearch under the hood and  allowing them to handle thousands of queries very quickly. When you build it in-house,  you might not be able to

09:15
build that type of platform and you risk slowdowns or breakdowns. trust me, no one's happy with a lagging system when business decisions are on the line. Yeah. Yeah, I don't disagree. I mean, you also have security. I think GDPR, HIPAA, some vendors bring encryption. have various uptime SLAs, four nines. In-house, you're juggling compliance by yourself.

09:43
those costs can really balloon. You have to staff around engineers. You have to make sure you have updates around your infrastructure. So it can creep up on you very quickly. Yeah.  And some campes are  pre-built for compliance too. Yeah. So I think high level total cost of ownership seals it. So good SaaS vendors are predictable  with the vendors handling upkeep. It's not in your hands, it's in their hands.

10:13
So unless you're a very large enterprise with a deep technical bench,  third parties  is  the smart architect's choice. It's  faster, safer, cheaper in the long term. Yeah, agreed.  Buy it, build on it. So  maybe we dive a little bit into  the tech behind these platforms  and kind of go into that a little bit.  Sure, yeah. Yeah. Yeah, go ahead. Yeah.

10:39
All right, yeah, let's dig into it. So how do KMPs  architect AI?

10:49
Let's, uh, let's just say, let's kick it off with, uh, NLP natural language processing. So transform models like a BERT or BERT variants or T5 they'll tokenize the queries  and use attention to understand context. So a good chat bot will run on this. So 80 % of queries, um, are answered  by the chat bot because it gets the intent of the query, not, not just the keywords. Yeah. Yeah. Totally agree. You're leveraging.

11:16
embeddings, vectorizing text into a searchable space and matching queries to answer  with cosine similarity.  Delta's mechanics used it. Search a KMP, uh get those logs in seconds.  Really made more sense. uh So Scott, when you start trying to tackle the structuring of the data, what are your thoughts there?

11:43
Yeah, sure. So knowledge graphs, uh they organize data in a  super powerful way. uh Concepts like, you know, say product  X that you might have in your company becomes a node. And then that node is connected by edges to things like sales or support. So you have these concepts inside of the business that are all connected via knowledge graphs. uh

12:09
Just take, for example, Siemens in a case that I read, in their platform, they take this a little bit further  by using graph neural networks or GNNs, and they're linking sensor data to maintenance records to reduce downtime substantially.  So just  to summarize that, it's a clear  connected blueprint  for smarter operations when you're using a KMP in this manner. Yeah, graph neural networks are cutting edge, in my opinion. ah

12:38
selling at uncovering insights from those data relationships. Staying up to date is just as crucial. Why don't we cover how KMPs ensure their knowledge essentially remains current? Yeah, good point. So machine learning is really what keeps it current. So tools like long short-term memory networks or LSTMs or gradient boosted trees, that's how you're going to monitor how the

13:08
content is used and flag anything that's outdated. So based on, you know,  monitoring how often it's used. Some third  party platforms will combine uh term frequency and verse document frequency. So that's just a method that will weigh the word importance  with a custom ML machine learning layer that will do retraining every week. So it's a self updating system, which is really essential for keeping that knowledge

13:37
fresh and alive, which is what's going to  power your business  in the most accurate and current state. Yeah. mean,  the tech stack architect in some of these KMPs is really impressive. mean, it's blending several layers for top performance.  You have vector databases like Pinecone or FICE, uh spelled F-A-I-S-S. I'll have that in the show notes. They deliver blazing speed while their knowledge graph sits at the

14:05
it's organizing all that data. On top of that, natural language processing handles the queries with ease. So that's really impressive. Then you have microservices, know, powered by FastAPI and orchestrated with Kube and Kubernetes, connecting everything seamlessly. And Redis caching ensures the response is under 50 milliseconds. So think about that. Speed is obviously important for real-time predictions, especially if, you know, you're

14:35
out on a runway. m Yeah, certainly. Good stuff. So that's the tech architecture. uh With these types of solutions, the business gets instant access  and the AI is ready with clean structured data to fuel it. Yeah,  I agree. So Scott, let's chat about some KMP's use cases for enterprise wins.

15:02
Sure, yeah, it's about data readiness. AI needs  structured inputs to perform,  and KMPs deliver just that, and they help cut the training time.  So just back to that Toyota global supply chain example, the uh machine learning system predicts disruptions because the KMP is highly organized, which  saves millions and helps them perform better,  better output across the board. Right, so the value of

15:32
you know, for KMPs and the customer experience is big. If you go back to your Zappos example, their chat bots tap a KMP. It's answering queries fast. It's a better customer experience, lower costs. You may even have that customer hopefully become a repeat customer, or maybe even add more things into their cart. Another area around collaboration is a huge angle we covered a little bit. KMPs can break that silo, right?

16:04
Yeah, exactly. So marketing operations, R &D, they can all be linked with relevant data. So for example, know, TechPerm One covers a product gap by graphing support tickets alongside  sales data, and then turn that raw information into actionable insight. It's perfect example of a smart system at work. Right. And agility too. New regulations always come up. We've talked about that on prior shows and K &P

16:33
re-indexes those  policies quickly.  The AI adapts, it's flagging  the environment that needs to be compliant, which hopefully keeps them out of the news.  Yeah, exactly.  Yeah, so that's the launchpad. uh Efficiency, decisions, scale. So without a solid knowledge management platform, the AI can oftentimes be lost or it's just harder to get to a system that's really working for your business use cases.

17:01
And then every one of them would be a missed opportunity. So Gary, what's the next step? So I would say the next step is really to audit your data, make sure it's ready and ask that question internally. uh Definitely test a third party KMP with an NLP attached. You know, look for graphing  and integrations. We talked about, you know, API capabilities. So really

17:27
you before you actually even go down the pilot path, make sure you've documented this and all of the individuals  and you know, business leaders are all on the same page. I've seen too many times where folks want to get into a pilot  and they're excited about the tech and the problems they believe it will solve, but they forgot about other parts of the business. So really making sure you have a well architected plan that will allow you to measure the wins during the

17:55
during this  testing phase. And then as you implement, it allows you to scale in a much more efficient and smart manner. Yeah, perfect. All right, well, I hope you found this episode useful about knowledge management platforms  and  all the use cases and technical information that we wrapped in there. um So  if you can, please subscribe to the Macro AI Podcast. And  please feel free to send us any feedback by visiting us at macroaipodcast.com.

18:25
uh Any more interesting episodes ahead?  Awesome.  Scott, always a blast. See you next time.  Same. Thanks, Kerry.