TRANSCRIPT: What is your chat strategy? Interview with Michael Maurer, Data Analytics & AI Specialist at Microsoft

Matthew Cheung: Hi, I'm Matthew Chung, CEO of ipushpull, and the following is a conversation with Michael Maurer from Microsoft, who is a data analytics and AI solutions specialist. On the topic, what is your chat strategy? So Michael shares his thoughts on how Microsoft Teams can become an entry point into using AI, how data is paramount in the use of large language models, and then he goes on to share his ideas on the future of AI in business.

Matthew Cheung: Enjoy. So Michael, yeah, thanks. Thanks for joining me today. Um, so Michael, firstly, what do you do at Microsoft?

Michael Maurer: So I'm a data AI specialist at Microsoft, focusing on understanding the client scenario and trying to map it into the Microsoft product portfolio. So getting the right specialists on board for supporting my client to adopting Azure technologies like OpenAI, ChatGPT.

Matthew Cheung: So you must be very busy at the moment.

Michael Maurer: Yes. So in, in, in sales terms, we're in a luxury moment that the clients call us. Uh, I don't need to prospect, uh, identify, reach out. Uh, use cases are spreading. Uh, faster than we can follow up with right now.

Matthew Cheung: Interesting. And how many of those use cases are focused around chat and with teams?

Matthew Cheung: Because there's two interfaces or tools that you can be used with, um, with how people are going to market with these ideas. With chat GPT when it first came out end of November, all of a sudden that you know, GPT is a large language models have been around for a few years, but this new kind of interface on top of it suddenly changed the game.

Matthew Cheung: Now that you can have, you know, you can plug directly into the GPT via Azure and you've got Microsoft Teams. How many of the use cases you're seeing are plugging those two things together and how many is actually using GPT in other areas as well or outside of just chat?

Michael Maurer: I would like to take one step back and look at the chat GPT interface as such.

Michael Maurer: So now thinking of the way a human is interacting with a computer is currently changing. So in today's world, we have this kind of taskbar, you have the icons, you know, Per task you want to execute, there's an application for, and this changes now tremendously as you get a different entry point in interacting with your computer.

Michael Maurer: So instead of opening Outlook to write the email, instead of opening PowerPoint to do the slides, you will get that as an entry point so that you can interact with all kind of settings with the capabilities of these applications directly in natural language. So based on that, we see. It's just another channel you're starting that interaction with and interestingly, those companies who started adopting teams broadly.

Michael Maurer: So to have by topic, by account or whatsoever, um, uh, teams channels then created, uh, they are really seeing that also for cross Teams collaboration. So within Teams itself, and for example, with integration into CRM system, dynamics, Salesforce, others. So, um, it's getting a new entry point in maintaining certain data, accessing certain data.

Michael Maurer: So I don't need to go to Power BI to understand the latest sales report, rather than chatting just in my account channel to get the numbers for this account. As a report dynamically pushed, and this is something where I see that the entry point, the user experience, um, which, which started normally by the user is going to find the right application, open it and then executing the task.

Michael Maurer: It's getting more into mode that you. Have your preferred working style and there you find your chat enabled, um, um, kind of co pilot in the words of Microsoft, whatever you name it. But, um, here it's definitely kind of merging the. Access entry points. So thinking of once you have built, um, a chat enabled system, um, you can start having these entry points in your intranet in teams in like you did now in symphony, the architecture behind more or less stays the same.

Michael Maurer: So also the clients currently are reviewing their investments in open AI because sometimes it starts with a new eye in mind. But when you, when you sit next to someone really working with the tech, you find other UIs are the kind of sweet spot, the main entry points and integrating there does make more sense than searching for something else.

Michael Maurer: So with teams, definitely, um, from the, the adoption rate we see, it's a very fast adoption rate with teams since the COVID crisis. So it is well established in the FSI industry, or at least at my FSI clients. So having there an entry point is a preferred starting point for many of my clients. How,

Matthew Cheung: how do you see, um, teams being used going forward where in financial markets, historically, you've had Bloomberg Chat?

Matthew Cheung: Yeah, more recently, things like symphony, you have, you know, Reuters messenger, which is now workspace messenger and various different chat platforms, which have existed in different areas of financial markets or commodities, and each kind of area seems to have gravitated towards slightly different platforms.

Matthew Cheung: How do you see teams, um, Launching from the I suppose that the footprint that it has IE on everyone's desktop, but for a lot of people, it's used for chatting internally or doing video calls less so for financial market workflows when I'm talking specifically about, you know, kind of pre trade workflows, post trade workflows, things where there's interaction with clients and counterparts and so on today.

Matthew Cheung: Teams is not being, you know, the platform of choice for that. But as we go forward, how do you see that evolving?

Michael Maurer: So the interesting part here is that you're living in that kind of world of APIs. Now, so I see that, um, uh, service consumption is not limited to a user interface as, or if there is the possibility to integrate.

Michael Maurer: Think of teams becoming a kind of agglomeration point, or as well, the Microsoft copilot story, an agglomeration point where you start the conversation, but then it gets rooted. To the kind of most, uh, relevant chat experience, for example, then to Reuters or Bloomberg capabilities who are underpin these kind of Microsoft co pilot experiences.

Michael Maurer: So that is something, um, from the envisioning phase, it's easy to think about that. And, um, let's wait what our product groups are going to bring in that area. Now, um, from the team's adoption itself, um, we see, for example, a lot of work being executed still in Excel as the main tool of working with data, consuming data, analyzing and, and, uh, you know, publishing nice people charts.

Michael Maurer: Um, however, Yeah. That is also now merging into that kind of co pilot experience when thinking of team as the starting points now, then leveraging what axle has underneath or your analytics platform like synapse. Those are coming closer together, and I think from a data governance perspective, many of my clients want to profit from that in order to get to know.

Michael Maurer: Where is what data actually used in the environment? Um, because in these tons of Excel sheets, it's sometimes difficult to understand where does the source come from and the lineage behind data quality aspects as well. So there I see AI is a conversation starter, but there are then dependencies about where to expose what data, as well as where to make it accessible for these chat experiences.

Michael Maurer: So for me, the interface with teams is the entry point. And then we have this kind of ecosystem around to think of. What is about sensitivity of data now? Should my copilot be able to access ABC or only A and B because C has a different sensitivity label and should not be taken into account? No. So that is something where you can think of different scenarios.

Michael Maurer: Um, where I see, Thank you. Teams as such is that entry point bringing the Office 365 graph. So the information about usage of documents together with analytics platforms and the kind of infrastructure services. So that is, uh, from my data AI perspective, a quite interesting sweet spot because you get already the graph information, you don't have to additionally add them.

Michael Maurer: How

Matthew Cheung: do you see the, the users who, who can? So you've got something like Power Automate, right, which is very powerful and, you know, gluing things together and creating workflows and so on. You said earlier your comment that Teams or any kind of platform, but Teams is, is the entry point where you can start a particular workflow, but then through APIs that are connected everywhere, you can then connect into lots of different things.

Matthew Cheung: How does that sit with no code development? Versus the API integration layer that actually needs to be behind the scenes with when you're talking to large financial institutions can be a bit of a mess, and it isn't as easy just to connect with our API because there's very old systems and things don't talk to each other.

Matthew Cheung: How do you see the no code developer operating in the world where there's highly regulated, difficult to access disparate data? How do those two things plug together?

Michael Maurer: And this is. Very interesting area because of this thinking off this self service on your data. This requires definitely a different data governance strategy and all these kind of buzzwords like data mesh came into mind.

Michael Maurer: No, however, Those who really started adopting these architectural principles, they can now profit from them because if you have really started defining data products with ownership and being able to vote about the quality of the data, you can then expose only with a certain rating quality data into the experiences.

Michael Maurer: So without that pre investment, um, it's that kind of difficult statement about garbage in, garbage out. Yeah. So you, in order to get the most out of your chat experience based on your data, you need to ensure that, um, certain quality levels are there. So thinking of that experience, um, where you. receive recommendations based on data.

Michael Maurer: Um, if relevance suffers, the users might stop adopting the service for them. So if you have 10 times a bad experience, why should you start with a chat the 11th time? So I think that is the, the exercise my clients are currently undergoing, not only developing that chat service, but ensuring the quality of outcome to have a broad adoption to make most of that investment in developing it.

Michael Maurer: And, um, there I encourage my clients to start testing as fast as possible. Now, thinking of a self service notion, most of my clients are in that kind of regulated industry with certain IT service management processes in place. So for them, it's a little bit, for the corporate ITs, the nightmare of having a self service platform and no one knows what happens.

Michael Maurer: But on the other hand, if you implement, uh, self service data platforms the right way, it will really free up IT from kind of being the servant to business in building the next report. Now, um, and that is something where we see interest. However, yeah, it's, for example, in manufacturing, we have clients who broadly adopted a self service power platform.

Michael Maurer: So having the business building, um, applications, and it reminds me that these Lotus notes applications. And they started adopting that broadly, and then they rank the applications by usage and taking them over into their corporate umbrella as soon as a certain threshold was hit. And this approach is in discussion in FSI, but right now it's really that more strict separation between an IT building a service ensuring SLA or OLA within the company, rather than having the business, um, doing the actual development themselves.

Michael Maurer: And it's definitely about the value a business employee can bring to the company. So is it worth Having the most special guy working on an application, um, in terms of salary, in terms of knowledge, in terms of client contact, which might be a better investment, not having him building these applications, rather than giving feedback and doing testing intensively.

Matthew Cheung: How, how do you see. Microsoft for a lot of people sits on their desktop with different tools they can use for their own use and a lot of the time for internal within the company use less so when it's kind of, you know, kind of internal to external because of the nature of data sharing can be clunky and SFTP and APIs take a long, long time to connect together and so on.

Matthew Cheung: How do you see the growth of. Um, what's happening on a user's desktop and how they're communicating with people externally. And then with that, how, how do you see minimizing the friction, um, for people to communicate with each other? For example, we're working with you at the moment around having a Teams app, which sits on the app store, which is, uh, is going to be a certain threshold.

Matthew Cheung: So. Hopefully some companies can click the button off. They go, but because of the nature of some FSI companies, there's still various info sec. The companies have to go through. How? How do you see that moving forward in the next few years? Kind of minimizing the friction and making it much easier to communicate externally with clients and customers and counterparts.

Michael Maurer: So B to B, let's say services. Um, this was a topic Microsoft started with Active Directory in year 2015, I think. And that time, the first B2C scenarios and B2B scenarios came into that service. And over time, it was expanded into different, let's say, buckets, different areas of interaction. For example, in Teams today, you can have multiple identities joining your Teams client and then switching.

Michael Maurer: So that you can have external collaboration. So Microsoft with I push pull, having a separate team set up and a B2B scenario working like a charm. The thing is as soon as you have, for example, applications built and published into a store, then the question comes up from the internal it is the publisher trustworthy is the code of a certain quality.

Michael Maurer: And can DLP or whatsoever? And in these scenarios, um, now taking the example of the Teams application store, we have a certification process in place and getting through with a Teams or custom Teams app is also then something where we publish the certified applications on our documentation so that the internal ITs can review.

Michael Maurer: from client X. That application was published and I want to consume it because I have a business relation between that other company and myself. And if it's an code reviewed by Microsoft, then the kind of trustworthiness of the source is definitely bigger than just handing over a zip file, which is can still deploy and install in your team's environment now.

Michael Maurer: So looking at that, The mechanisms are sometimes not aligned across the portfolio of Microsoft. So having at Teams a different experience than if you publish your solution in the Azure marketplace or publishing data products in purview in B2B. So with Azure Data Share, for example, underneath, um, I assume That we will see further investment of Microsoft in this area to align the experiences and make it easier, more intuitive to start sharing information.

Matthew Cheung: So to finish off, it'd be good to hear your thoughts on, it's obviously been a very busy year since, you know, OpenAI released chat GPT and like you said, there's this moment in time where. From a sales perspective, it's this kind of luxury moment where, you know, customers are knocking at your door. Um, you know, similarly, we've seen the same thing in, in our business.

Matthew Cheung: How do you see that, uh, rate of growth continuing and proliferating using these, these AI kind of out of the box tools across different, you know, integrations, platforms, and so on? Yeah.

Michael Maurer: All of us probably have heard in the past of the Gartner hype cycle. And thinking of chatbot experiences. I have the impression that we had that first hype in the year 2015 2016 with that first chatbot wave.

Michael Maurer: However, the capabilities were kind of limited right now. I see that the, the, let's say the speed of any innovation we have seen in the area is unlocking certain scenarios, which no one has been able to think before. So my hope is that we are in that cycle of the hype cycle where we see then a continued growth until saturation right now.

Michael Maurer: I have the impression also the market is currently fiddling out scenarios worth investing in. So most often I hear, ah, let's use it for an optimized intranet search. But optimized intranet search is nice, but no one will create revenue out of that. And my client started thinking of where are services? Um, either saving cost or generating revenue.

Michael Maurer: So understanding the business model behind and I have the impression that the capabilities these services bring to market support in both areas, really interesting scenarios. So that's why I think, yeah, there will be a continued growth and, um, not the kind of second type. When thinking of, um, where will the market lead with technology, I'm, I'm very impressed that my product group opened up the AI studio for, uh, third party services plugging in.

Michael Maurer: So thinking of you come in with a co pilot experience with the first ask, and then you have, uh, third parties or open source hosted on Azure. So AI models. Uh, which you can then train or customize on your own to have specialized models running. So this time it's not the, uh, Microsoft has the one and only AI.

Michael Maurer: It's really an ecosystem of AI services now being able to integrate. And that's what makes me excited because I think that's the right approach for the market because no one knows what next is with innovation, uh, at that next step, you know, so I'd rather integrate, uh, the newest fancy stuff, which serves to a certain quality than having to stick to a piece of technology.

Matthew Cheung: And it's that interoperability and, and building block components. So that, you know, when the next thing comes along, it's very easy to slot it in. Great. Michael, thank you very much. Thank you, Matthew. This is the iPushball podcast series about what is your chat strategy? Over the next few months, we'll talk to leaders in the field and their views on the chat ecosystem, chatbots, and how AI is evolving the space.

Matthew Cheung: We hope you learn something new.