Tlao Cover Art 2023 Vector

Justin Watt – The Case for Automations, AI, Software – EP.237

My guest is Justin Watt, co-founder of Switchboard, a company that helps businesses become more efficient through automation and AI tools.
00.00
LinkedIn
Twitter
Facebook
Email

Episode Description

My guest today is Justin Watt, co-founder of Switchboard, which helps businesses with automation and AI tools to become more effective and efficient companies. He’s the smartest person I know when it comes to automation, and I’ve been really excited to record this episode with him laying out his philosophy for tools and how to approach streamlining key internal processes.

We talk about hype vs reality with AI tools, preparing your company for automation and laying the foundation through cleaner data, the three pillars of modernizing a company via people, process, and technology, and so much more!

One last note, I’ve launched my own search called Airframe Group and I’m looking for great companies and experts in industrial services and value-added distribution. Please reach out, always excited to meet owners and experts in the industry. Email me at [email protected].

Listen weekly and follow the show on Apple Podcasts, Spotify, Google Podcasts, Stitcher, Breaker, and TuneIn.

 

Clips From This Episode

The Three Pillars of Automation

Questions to Ask When Evaluating New Software Tools

Hood & Strong LLP – One of the nation’s premier full-service public accounting firms, Hood & Strong LLP provides buy- and sell-side quality of earnings, due diligence, assurance and tax services to search funds, private equity firms, and business owners and investors. The H&S Advisory team helps expedite a smooth, cost-effective transaction process that maximizes value and minimizes tax impacts for both buyers and sellers. To learn more about how Hood & Strong can support your M&A objectives, please contact Transaction Advisory Group Partner Jerry Zhou at [email protected].

Oberle Risk Strategies– Oberle is the leading specialty insurance brokerage catering to search funds and the broader ETA community, providing complimentary due diligence assessments of the target company’s commercial insurance and employee benefits programs. Over the past decade, August Felker and his team have engaged with hundreds of searchers to provide due diligence and ultimately place the most competitive insurance program at closing. Given August’s experience as a searcher himself, he and his team understand all that goes into buying a business and pride themselves on making the insurance portion of closing seamless and hassle-free.

If you are under LOI, please reach out to August to learn more about how Oberle can help with insurance due diligence at oberle-risk.com. Or reach out to August directly at [email protected].

Interested in sponsoring?

(00:00:00) – Intro

(00:03:03) – Justin’s career and background

(00:07:31) – The three pillars of automation

(00:11:56) – Examples of implementing new tools in current processes

(00:18:26) – How can you see through superfluous marketing?

(00:21:23) – What questions should people be asking when evaluating new software tools?

(00:23:06) – What do companies need to work on before they can consider implementing automation?

(00:26:20) – The AI hype & discerning noise from signal

(00:36:57) – Enablement infrastructures

(00:42:28) – What automations do you enjoy spending the most time on?

(00:49:14) – Reinventing HR Processes

(00:51:26) – The anatomy of automation

(00:53:10) – What are some quick wins you often see with companies you work with in automation?

(00:55:38) – What advice do you share with folks trying to implement automation?

Alex Bridgeman: Justin, thank you for coming on the podcast. I’m thrilled to get to chat about all things AI automation systems, definitely a hot topic now, but a kind of emerging area of new systems and tools that could be useful, and we’re all trying to kind of figure out where they fit, especially for companies that are like maybe not software but are more like hardware or distribution focused and whatnot. So, I’m excited to chat through what your perspective and your kind of earned and lived experience of these tools are. Kind of to prime that though, would you walk through some of the stuff you’ve worked on before Switchboard and some of these automation kind of research and consultancy work? 

Justin Watt: Sure, yeah. I started my career kind of like I think a lot of people get into their careers not knowing what they want to do, but I went to university for psych thinking I wanted to spend a career sitting one on one in a room with people hearing about their challenges, which in a way, I guess I do that now, but it’s very different. And I worked at a telco during that time and kind of got to see the, as a young, scrappy university student era, not knowing what I wanted to do, getting to see the inner workings of a kind of multibillion dollar company, to me, was really fascinating. It’s a lot of spreadsheets and whatnot, but it was interesting to see kind of how they were actually run day to day. And then from there, went to IBM, where I got to work on large scale software and IT implementation projects and see how that side of things works. But as I think a lot of people who come from that size of a consulting firm, you kind of get burnt out quickly because the the red tape has red tape. So moved over to working at a small product design and development firm called Metalab, which when I joined was 30 or 40 people. And they had just had the case study launched for helping design Slack. So as you can probably imagine, that’s a pretty good case study for a design firm to have. So while I was there over the course of seven, seven and a half years, we grew to nearly 200 folks. And I started kind of helping build out the project management function along with a lot of really smart people who did a lot more work on it than I did, and I moved over into operations. And during that time, Zapier was coming up as a cool new tool, and Airtable was coming out as a cool new tool, and we built our operations around these tools because it was so obvious that the things that everyone’s doing all day administratively don’t really need to be done to that degree. You can automate a lot of things. You can plug in a lot of tools and data. And so, over the course of that kind of five-ish years that I had moved from project management into operations and the company was growing, we got to build a lot. And as we were doing that, kind of building the airplane as it’s taking off, it became very obvious to me, especially when ChatGPT became public, that AI was no longer this kind of pipe dream of maybe one day if, or this could happen eventually, it was here and now, and you could plug that into a lot of these flows to help kind of supercharge them and add rocket fuel to these existing automations. So that led to founding what we’re doing now with Switchboard, which is helping small, medium, mid-market companies figure out their automation and AI within their company, and a lot more than that, but I’m sure we can get into that. 

Alex Bridgeman: Yeah, absolutely. What kind of automations or areas of improvement did you see early on and in that operations work at MetaLab when these tools started coming out? What kind of experiments and things did you try that led to some interesting places? 

Justin Watt: Well, I think experiments is a good word because my theory that I think I’ve seen play out time and again is 99% of business challenges are actually people challenges. So, a lot of the challenges that we were looking to solve operationally were because you’ve got sales teams that they’re good at their job because they’re high level and jumping around 30 email threads with potential clients each day. And you’ve got project teams jumping around all these different clients that are active and whatnot. So a lot of it was trying to solve for how do we not have to rely on what is inside of people’s heads so much or in hopes that they communicate with others as to what’s going on or passing along information or updating some field in the CRM somewhere. So there’s a lot of kind of experimentation and learning through failure on connecting tools. So, some of the first things that we did was connecting tools so that when a CRM is updated, it can trigger in a project management tool over there certain templates being spun up or certain things being filled out so that, not relying on individuals in hopes that they go and fill something out. We actually had it filled out automatically, just because they updated one place, it could update in four or five kind of thing. 

Alex Bridgeman: You talked about these three pillars of automation that kind of prime the discussion on automation, like having the people, process, technology pieces figured out and organized is necessary before doing any of these system discussions and looking at different tools. Can you talk about those three pillars and maybe how they interact and rely on each other? 

Justin Watt: Yeah, I think a lot of- I’ve always seen the McKinsey stat of something like 70% of digital transformation projects fail. And I think partly that’s because digital transformation is sold as a silver bullet when really it’s a series of different things that have to happen in a company. But I think a lot of them fail because the people part is not really thought through and what the company actually needs isn’t really thought through. It’s just kind of here’s a cool, shiny new tool with this feature, and we think that’ll be, again, a silver bullet. Or someone builds out a process without any kind of thought as to how do we actually streamline that process? Or when it gets to the people part, how do we actually train people and get them bought in to use this new tool or process or way of doing things that ultimately will save them time, but like anything in human nature, change is uncomfortable. So that’s why I’ve learned over the years seeing it done with the product world of cutting through the jargon and acronyms of whatever new flashy trend is out there and just looking at the fundamentals. I think a lot of people probably overuse the term first principles, but I think it rings very true when it comes to actually improving operations within a company too. So that’s, to circle back to your question, the people, process, and technology side really has to work in tandem with each other. And that’s how I tried, at least the execs and kind of operators that we work with, I try to get them to think about it, of you can’t have any just one of those magically solve for whatever operational or business challenge you’re having. And so we try to get them to think first about what is the actual process itself. Often, I think a process is let’s say 30 steps and even just looking at it, again, those first principles of, does it have to be 30 steps? Like what is actually happening here? So more often than not, you sit down with a few different department heads and say, walk me through, as maybe a quick example, a sales lead from when they become a lead to when you sign them as a client or user or customer, what happens? And getting them just to play that back and then jotting it down in some sort of whiteboarding tool or even just documenting it in a Word doc, they often see, oh, why are we doing it this way? So, solving the process part usually is one of the easier parts because it’s just pure logic of once you jot it down, oh, we could cut this out or combine these things into one whatever it might be. And so, you often just get a quick win of 30 steps becomes 20. And then the fun part becomes, okay, of those 20 steps that are left, how do we automate as much of that as possible? And that’s where the technology pillar comes in of, okay, do we have the right tools currently to do this? Are we missing a tool? Is a tool that we have not actually going to solve this as well as something else could? So, solving the actual software, and I think software is tricky because the marketing teams at SaaS companies have become so good where they say, this is an all-in-one and you can do whatever you want with it, it’s going to be this magical AI power whatever. And I think people are getting burnt out on that quickly because these tools are not all-in-one and they’re not flexible enough to be able to fit each company’s unique needs and way of doing things. And so, kind of looking at the fundamentals of what do we need out of our software and then implementing the right software and then automating as much of that as possible is kind of that technology pillar. And then a big part of that is kind of architecting your data in the right way, which we can dig into. And then the people pillar is really about change management and documenting things and training it. I think a lot of people will sign up for a tool and say, here’s your login and good luck and just make this fit with our process. And a lot of success that I see at least where companies actually level up their operations is taking that time to say, what is the process? Let’s document that for someone in the context of the tool that we’re using and the process that we want to see and then actually training people on that. And it doesn’t need to be this sexy training program. It can be, depending on what the tool is, the process and the number of people using it, it can be kind of one on one training. It could be group training. Could be Looms I see be really successful, where someone only has to record the workflow, themselves doing the workflow once. And then they share that Loom with everyone and say, watch this and you’ll very quickly understand what to do. And so those three things combined of kind of the people, process, technology, once you can align those, and we can dig into each one of those if you want, but once you can align those, that tends to unlock actual success with rolling out kind of operational change. 

Alex Bridgeman: It’d be fun to go through an example. Is there an example that comes to mind where the team you worked with or maybe it was something in MetaLab even where you started with that foundational like, okay, what is the process? Who is involved in this? And then walked through into how can we use current tools or new tools better? Is there an example we can walk through to give kind of a framing for what’s going on? 

Justin Watt: Yeah, I’m smiling or laughing because one comes to mind. About a year ago, I posted a workflow on Twitter, and an exec at a growing SaaS reached out to say, hey, we want to do this for our company. And so the very first call we had was he wanted dashboards. He didn’t really care about the process and the technology. He just said, I know X is happening at step one, and I want to know by step 20, I want Y to happen, and I want a dashboard along the way of what is the, this is their sales process, so what is the number of leads we’re getting, how many of those are qualified, how many of them are converting to a call, of the calls, how many are converting to a proposal, of the proposal, how many are converting to etc., etc. And so I said, great, how are you doing that right now? And he said, well, we’re not doing any of that. And I said, okay, totally understand. That’s why we’re chatting. What data do you have in place to be able to do that? Well, what do you mean, what data? I said, well, if you wanted that dashboard tomorrow, regardless of us working together, how would you make that happen? He said, I would have to ping four or five different people to update their Excel spreadsheets, and then they would all send that to one person, and then that person would be able to report for me. I said, great, so you have no data infrastructure, none of your tools are connected, and your team doesn’t even have a process for kind of updating the stuff on a regular basis. And so that was to me kind of a prime example where I think a lot of folks want the end result of automated real-time dashboards telling them all these things, but they don’t realize it’s kind of working backwards from there to say, what is your actual data infrastructure? If a lead in a CRM is entered, then we can add that to 10 different tools. And I think that’s kind of step one is to say, because I don’t know about you, but I think a lot of people grew up saying, I want to be an astronaut, lawyer, engineer, whatever it might be. No one grows up saying, I’m really excited to just update spreadsheets all day. And so there’s this monotonous belief around updating spreadsheets. And so no one does it, even though they’re expected to or asked to, no one does it. And so part of it was saying, what is your source of truth? And how do we build everything else around that? And so we started with a CRM that they did already have, thankfully, but that CRM was not connected to any other tool. So for them, everything from invoicing to their implementation managers with this SaaS, they had to work with a lot of different teams to implement this SaaS. It wasn’t just kind of an off-the-shelf Asana or HubSpot or something. And so we had to work with them to say, okay, let’s start with a CRM as a source of truth if you want to be able to track all the sales stuff and then connect it to all the other tools so that from when someone becomes a lead to when they’re paying their first invoices, you know exactly what’s going on and all that data is talking to each other. And so to get there, you have to sit down with their team and say, what is your process? And that’s where the people side comes in often because on the people side, it’s, not in this case, I’ve seen it with others, though, where you’ve got two VPs from two different departments, and they don’t really get along great and they have very different incentives and very different needs of their teams and expectations from their execs. And so you sit down with them to say, here’s our proposed process. And they’re right away saying, I don’t want to do that or my team shouldn’t be responsible for that. And so that’s where the people side comes in of, okay, great, well, your exec wants all these magical dashboards. So how are we going to get there together? And framing it to them of it’s us against the challenge of getting these dashboards going, not you two against each other. So, kind of breaking down those barriers and getting people aligned of, oh, great, okay so, if we’ve got the data in one place, we actually- ultimately right now, it’s work to get this in place, but ultimately it will save us all a ton of time because our data is talking to each other. We’ve got things automatically updating so that when a salesperson updates a CRM, it will go and update their project management system. It will go and update their invoicing. It will go and update all these different things that ladder up to having the data to update a dashboard. So that was a bit of a ramble, but hopefully that kind of answers what you’re looking for in terms of kind of a case study of that. 

Alex Bridgeman: Yeah, absolutely. How often is part of the issue that they have software already, but they’re just not utilizing very much of it, or there’s just extra features they’re already paying for that they could be using, and that alone will take them a long way, even before looking at other tools?

Justin Watt: Yeah, I think a lot of teams have kind of shiny tool syndrome where they were seduced by a marketing page or a headline on a sales call or sales deck somewhere and said, great, that’s why we need this. And then they implement it in a week and don’t ever think about it again of how is this actually serving us. And so, it’s kind of the classic people use 10% of their brains, so the saying goes, I think people are using like 10% of their software. So I would say that a large part of it is there’s no one kind of responsible usually in an org to say, what are the features of this and how can it best serve us? And then kind of working backwards from there to actually use those features. And a lot of it is superfluous features or things that, again, marketing speak, they claim to do all these things and it won’t actually do that. But I would say that teams use a lot less than they think they do of the power of their software. Inversely, though, I think that there’s also, because of the ZURP phenomenon fading away and just all of these different reasons for SaaS companies needing to gain new sources of revenue, we’re seeing also the inverse problem where there’s a lot of kind of basic features that are now being hidden behind, excuse me, a lot of new features that are hidden behind enterprise tiers. And so, these companies who are using the software are saying, well, we want this basic feature. And now the SaaS company is saying, sorry, you’re going to have to 5x your monthly user costs to just be able to access these basic things like permissions. Permission structure is a basic thing that I’m seeing kind of starting to be hid behind more enterprise payment walls, which is kind of brutal to see. So it’s a twofold problem is what I’m getting at. There’s teams who have software who aren’t looking at all the opportunity in front of them of how they could use it better. And then inversely, there’s these SaaS companies kind of overcharging and adding a tax on, so to speak, just to use basic functionality that would serve them.

Alex Bridgeman: Is there any- you talk about the marketing like being really good. What is helpful in seeing through any superfluous marketing but also just trying to figure out what does the software at its core do and how would it actually fit? Are there any best practices you’ve come to for evaluating a new software tool and determining what’s real, what’s not, what fits with what I do, what doesn’t? 

Justin Watt: Yeah, I started using this analogy with a couple of clients. Have you ever seen the movie The Prestige, the Christopher Nolan movie? 

Alex Bridgeman: Oh, I don’t think so. 

Justin Watt: Highly suggest, first off. Highly, highly suggest. It’s a great, great movie. But the fascinating thing about that movie is he took a page from Tarantino, I think, where it starts at the end. And so the first 90 seconds of that movie tells you everything you need to know. But then you spend the next hour and a half figuring it all out, piecing it together, going, oh, cool, now I see why they showed us that at the beginning. And it doesn’t make sense until the end why they showed you that. And so I think software is a lot the same where teams don’t really start with the end in mind. They just see the marketing page and say, oh, shiny thing, great, let’s sign up and try it out and see if it works for us. And then they get three months or six months in and say, I don’t think this is actually serving us. And then they switch tools. And that kind of comes back to that shiny tool syndrome where people are seduced by marketing speak because they’re not really spending the time to actually think about what they want out of it. So why I use that analogy of the Prestige movie is starting at the end is where I see teams be really successful with choosing the right software and implementing it the right way by saying what is our actual process and how we’re working right now and what do we want it to be. So what would kind of our dream state or ideal state, kind of that starting at the end and then involving their team with that too. Because, again, a lot of times it’s an ops person or an IT person or exec saying, we were sold on this solution, but we haven’t really talked to the 20, 30, 80, 500 people that’ll be using it day to day. And so even just spending that bit of time with people to say, this is what we have in mind as a dream state, how would this tool serve you? Or what features would you like to see out of a tool that would serve us in that goal. And so starting at the end tends to be a really good way to go about it to figure out what do we actually want out of this thing and what would success look like for us. And then that gives you your feature set of, well, therefore we need these features or this functionality in the tool. And that gives you a much clearer kind of rubric or parameters of what you’re looking for. And it also empowers you during the kind of sales process with a lot of these SaaS companies because if you’re a big enough company, they’re going to say, oh, great, yeah, let’s hop on a call. And then you get on that call and they end up steering the ship for you because you’re just saying, uh-huh, that sounds cool, oh great. And you’re not owning the process. They start to own the process. And then you just say, yes, and next thing you know, you’re rolling out something that you don’t really know if you even really need or suits you. 

Alex Bridgeman: Are there any important questions or maybe lines of questions that folks should use more when evaluating a new software tool? 

Justin Watt: I would say a lot of really good software companies will have a clear roadmap. And so they will be able to tell you what they currently have and what they plan on releasing in the next… Usually they won’t go too far in the future, understandably. But if you’re able to share kind of the list of features that you want and then ask them the line of question of, show us these features and do you have this? But then also say, what’s coming next? What are the things of what I’ve told you about, what we need as a company with the software, what things are coming down the pipe? And I have found the best software companies are very open internally in that product and design and engineering and whatnot is sharing with the sales team, here’s our roadmap for the next quarter, the next six months, 12 months, 18 months, or we know that this is a highly requested thing, so we know that eventually we’re going to get to it. And so those open companies that are confident enough in both what they currently have and where they’re going and can answer those kinds of questions for you openly I find will tell you a lot about a company. If a software company both can’t really tell you if your needs will be served after you tell them what you need, and flip side, they can’t really tell you where they’re going, I find that a hard company to rally around because, well, if you can’t even do what we need and you don’t know what’s next, is this really going to serve us long-term? Because software tends to be a long-term investment. It’s not meant to be a one-off project that you’re working with for a month or two. It’s something that dozens and dozens of people are using for years and years. 

Alex Bridgeman: Yeah, certainly. And one thing you also talked about was many companies aren’t prepared for their workflows having automation in that the original data is either messy or not being recorded digitally. It’s in a spreadsheet or something else. Can you talk a little bit more about what you mean by that and what that implies, that companies need to work on before automation can work for them? 

Justin Watt: Yeah, a lot of- during kind of the scoping process of working with different teams on automation, a lot of them say, can you work with our tools though? Like a lot of teams feel that their software and their process is so unique that it takes an expert in their field with their company to understand it. And I always say to them, it doesn’t really matter. Like at the end of the day, they as users of those tools or the automations might see the front end, the Salesforce or the HubSpot or a pipe drive or whatever their CRM might be as an example. But at the end of the day, underneath all of those tools is the same set of data. It’s a lead has a name, they have a title, they have a location, they have an email address, all that kind of stuff. And so as long as they are using a tool that has an API functionality to plug it into other tools, and they’re actually using digital tools, then they’re usually five steps ahead of a lot of folks. The challenge with a lot of folks around automation is they say, we want to automate our sales process or our marketing content machine, whatever it might be. And then you find out they operate off of whiteboards and siloed Excel sheets saved on someone’s C drive. And you can’t really connect data that, well, you certainly can’t connect data that’s on a whiteboard or notebooks. And it’s really hard to connect data that lives in these kind of one-off ad hoc individual files on people’s computers and hard drives. And so that’s usually kind of step one is to say, how do you actually work now currently? And are you ready for automation? Because a lot of folks aren’t ready for automation if they’re still in kind of call it legacy pen and paper process, or if they are on local server, not cloud-based software, it’s also harder to do automation. You can do automation, but it’s not really worth it because it’s going to cost you an arm and a leg to set up the infrastructure to connect these local files and offline servers in a backroom somewhere to all these other tools that they use. So data really, really is important to be able to unlock automation. And then I think a lot of people don’t realize that AI works off of structured data that is interconnected. You can’t magically say what tasks are due today for me in an AI tool when all your tasks are in a notebook somewhere. And granted, that might seem very obvious, but I think a lot of people hear the hype about AI and they think, great, so we just roll out AI and it will automate and do all this stuff for us. And they don’t realize that, first, your data has to be in an accessible place for AI to look at it, whether that’s getting it out of a notebook or getting it off of the Apple Notes app on their computer or whatever it might be. 

Alex Bridgeman: Yeah, the AI hype right now is certainly very, very high. And we discussed, too, that some of that we’re overhyping, but pretty soon there’s going to be a lot of capability to rise up and match that hype. Can you give like a, what’s the balanced view on like AI as hype versus AI as a useful set of tools for us to start using more? 

Justin Watt: Yeah, it’s so early. I liken it to there was the era of AOL internet, and then suddenly everyone had high-speed internet through fiber, and then now we’ve all got 5G on our phones and we can access the internet anywhere. I think we’re in that AOL era of AI where it’s like it is a new cool thing, and it’s obvious where it’s going, but it is not matching the hype out there right now. But on the flip side, it is advancing extremely quick. Maybe it would be useful to share an example with a client that we had a while ago where, long story short, they have a tool that they implement for others. And so in implementing those tools, they’ve got, I think it was like eight or nine people on these calls all day with different clients of theirs, just implementing this tool. And so they had the classic, this call might be recorded for quality monitoring. And that was true. They did actually have a few people just listening to these calls to say, this is a newer employee. Are they covering the things that should be hit on in these implementation calls? And so, as you can imagine, there was hours, well, I mean, multiple full time people spending hours a week just listening to calls and kind of grading them and whatnot. And so with that team, we had started working with them where they wanted AI analysis of these calls, which logically makes sense and was doable at that time. The challenge was in an hour-long call, the transcript for that, that we run AI prompts against to look at the call to say are they covering X, Y, Z in that call, the AI capabilities at that time would time out because an hour-long transcript was too much content for the AI to review. And so, we had to do, not to get too nerdy with it, but like vector stores, we had to do all these things to compartmentalize the call into different segments and review just those segments. And so that was an example where AI could do it, but it was early days and it was really clunky and it took us longer than I think any of us would have wanted to implement it. By the time we were done that project, the context window or the kind of ability for AI to understand a larger amount of text at one time, the context window expanded so much that that all became moot, where it could listen to a three-hour call or transcript of a three-hour call and still be able to kind of keep the plot and understand all of that. So, I think that’s a good example of how quickly AI is advancing, that three months before that, it was really clunky to be able to build this and more expensive because it takes longer to build those kind of workarounds to make it work. And three months later, it was a moot point and could handle it. And so, I think that’s indicative of overall where AI is, is kind of the hype cycle is very real and the snake oil is also very real because there’s a lot of folks who I don’t think understand anything about AI. They’re just kind of jumping on the bandwagon to say, whether it’s SaaS companies saying, we’ve magically got AI powered whatever, or it’s consultants or folks saying, oh yeah, you can do all this cool stuff. And I think there’s this hype that is probably going to be for the next few years at all times six to 12 months ahead of where we actually are. Having said that, that example hopefully showcases where we are advancing quickly enough where a lot of these things are becoming very real very fast. 

Alex Bridgeman: So help us see through which part is maybe more on the BS side of things versus that’s a legitimate use case and a tool that works today, so we can figure out, okay, we’re hearing all this talk and noise, what’s actually useful here? What’s it look like today? And then how do you suss that out in the future? 

Justin Watt: Yeah, I think with we can call it maybe applicable AI, of things that you can actually do here and now that are genuinely useful, reviewing things and summarizing them or making guesses about what something means. So, maybe a good example is reviewing an investment memo in a financial use case where it’s often these 10, 30, 50 page documents full of fluff but also important financial figures that analysts or different folks in financial firms spend hours a day just reviewing and kind of summarizing for others. When you give AI something, along with a well-thought-out, well-structured prompt, say, review this and tell me X, Y, or Z about this document, it is already extremely good at that and getting better and better at it. There’s hallucinations and there is things that can go wrong. So, I always say use it as a first pass, but there should still be a human in the loop. So those kinds of things, whenever it’s summarizing things or creating content that is very basic, I think it’s already quite good at. So a very quick example is in the use case within an automation flower, within a process, within a company, is a lot of creative teams that we’ve worked with, their clients or their stakeholders will fill out a brief for them to say, this is what I want this design or piece of marketing content to look like, to feel like, the content that should be in it. And they give dozens of lines of inputs in these forms to say, this is what I need you to do. And so those kinds of things, AI can kind of summarize those. And you can say, act as a designer who is taking a brief and needs output in this format and summarize these kinds of points, avoid talking about this, lean into talking about that. And those kinds of things are really applicable in a workflow because if you’re in a company that’s getting 10, 20, 40, 50 of these creative briefs a week sent into a creative team, that really helps streamline their process. So that’s the reality, I think. That’s applicable AI. The other end of the spectrum that is pure hype right now, and I’ll stand by this if anyone wants to discuss it further, is anything to do with reasoning. So, a lot of people talk about AI as being a prediction machine, and I’m sure you know this too. The math of AI is really analyzing text and comparing it against its knowledge base to say what is the next logical word that goes here, and that eventually turns into an AI output. What AI can’t do is reason for itself to say, I want to give you a URL of a website and I want you to look at that company and figure out all of its competitors, all of its kind of market fundamentals, its total addressable market, its pros, cons, and then go to LinkedIn and figure out all the employees that work there, what kind of employees do they hire, what has their turnover been, anything that’s kind of a multi-step process where it has to figure stuff out for itself, it is not nearly there. And a lot of people talk about AI agents and talk about kind of multi-step AI workflows. And that is realistically a hack at this point. It is taking multi-chain prompts and adding all of these things together to figure out something, but it will not be able to reason for itself for a while. And I mean, some people talk about GPT-5 introducing that, who knows. But at this time, anything that involves kind of logical thinking or multi-step things where it has to go figure stuff out for itself, it is not nearly there. 

Alex Bridgeman: Yeah, I think I was using a chat bot the other day in setting up my Sonos because I’m having a really hard time. The worst part of every move so far in my life has been setting up Sonos. Like setting up the internet is usually like a breeze, but for whatever reason, these speakers just take like at least a week before like they work and I’m like restarting everything. So, I go through this AI chatbot, and it’s clearly, it’s not connecting. I’m using the same Wi-Fi. Like it all should be pretty seamless here. And it was clearly going through this massive checklist of all like possible problems. And it’s like, is it plugged in? Like, no. You’re just like going one by one. And I’m like 20 back and forths in and I’ve gotten nowhere. And it was just, the whole thing was just more like frustrating than anything. I just like ended it and tried to go- I went on YouTube and figured it out. So, it was just a frustrating saga that I feel like I’ve seen that example already before

Justin Watt: Yeah, and I think that’s why so many people have a disdain for chatbots, because that was one of the first, even before it was called AI, that was one of the first kind of call it machine learning things that people saw out in the wild when they’d go to visit a website, and it’s like we’re going to use the chatbot to find this before we actually introduce you to a human customer support person. And so many people have built that mental model in their mind of, oh, when you tell me you’re going to use a chatbot, it’s going to fail, it’s going to frustrate me, and it’s going to work terribly. And it’s gotten better, but not that much better because, again, it’s working only on what it knows, but it can’t go reason for itself to say, oh, you actually mean this. I’m going to go look somewhere else for that piece of information that you’re looking for. It’s just following a set of prompts that are run against a knowledge base that it has sitting in front of it. But if it’s not connected to the right things, it doesn’t have the right prompts and has to, quote unquote, think for itself, it falls apart very quickly. 

Alex Bridgeman: Yeah, I feel like my expectation when a chatbot replies is that I’m not going to get anywhere until a real person gets on. We’re going to spin our wheels for a couple back and forths here to get information about my problem, and then I’m going to like repeat that problem again once the real representative, like the real person gets on the chat and discussion. Like at some points, even with just figuring out like a credit card problem, I’ll like write up my question, my problem, AI goes back and forth, and the person hops on to start figuring out the problem too. And I’ll just like copy and paste what I said earlier. And I might do it like two or three times, and it doesn’t save, doesn’t connect. There’s so much wasted energy and wasted time right now, at least with most chatbot interactions I’ve had. 

Justin Watt: Yeah, and the unfortunate part is it’s a lot of people either saying that they are able to do it and setting false expectations on the marketing side, or it’s, I guess, execs who are being sold a false promise, but they’re trying it anyways and making their company look terrible for it because it doesn’t actually work great. But how does a consultant say, yeah, we did it, we executed on the thing we said we would, but I imagine their NPS and customer satisfaction scores go down in the toilet. 

Alex Bridgeman: Yeah, definitely not as great. But moving on from chatbots, there’s kind of like two broad categories that you’ve looked at, system improvements or automation improvements, that’s enablement, and then infrastructure. I think it was a good kind of like primer, like what falls into each bucket before going into like anatomy of an automation and more and more case studies around certain automation points. Can you talk about kind of those two buckets, the enablement and infrastructure, and what you see fitting into both buckets? 

Justin Watt: Yeah, I think a lot of automation that we build is people will never see it. And so that’s kind of the infrastructure part of it, where it’s making people’s lives easier, but they don’t know it, or it’s enabling them to do something better. So to chat through infrastructure first, I think a good example was that CRM one that we were talking about where, as a lot of people listening can probably imagine, if you’ve got a person who is a sales lead that you then want to use for marketing purposes or data around that person for marketing purposes or for invoicing purposes, a lot of the power of automation is taking that information and not having to duplicate it manually and have a person go hire someone administratively to take the name, company, legal name, billing point of contact, like all of the stuff related to them as a customer or client or user, and go replicate that in your invoicing tool or your payment portal or your legal contracts or your marketing systems. The infrastructure with automation can take that and once it moves to say a certain point in the pipeline stage, say closed one, it knows, okay, this is now officially a customer. Let’s go take this information from the CRM and add it to those different places so that the marketing tool is already filled in for the marketing team. The legal team who has to then spin up a contract already has all the information at their fingertips or the invoicing team or finance team already has everything there. A lot of the questions that we ask when we’re doing strategy phases with clients, or one of the first questions I should say we always ask is what information are you constantly chasing or being chased for? And usually their face lights up because they are just so excited to tell you about all the time that they waste saying, oh, I’m on Slack or Teams all day or in my email inbox all day just asking people for these things. What’s the status of this sales deal? Can you tell me who the billing point of contact is? Can you tell me where they’re located so we know which tax regulations to follow? Whatever it might be or whatever reason they need it, just the sheer amount of time that people spend each day chasing information or being chased for information is obscene. And so that’s where the infrastructure part of automation, this might sound lame, but that’s why it excites me because no one wants to spend time doing this. Like, none of us signed up for a career of just chasing other people for menial information that, yes, is important and does unlock us doing our jobs, but that’s not what people want to do in their career all day. And so that’s why the infrastructure part of automation excites me a lot, because they don’t have to. Like, once you have it in one place, you can have it in five other places if you have the right tools and right infrastructure set up. So that’s the infrastructure side. And then the enablement side is kind of along the same lines, but it’s where people really start to feel it in their day to day because it’s doing parts of their job for them. So a good example is when someone does become call it closed one in a pipeline, a sales pipeline. Often there’s a ton of work after that to go and take- A quick example actually that we are building for a client right now is when something moves to that pipeline stage for them in the sales process, they then have a project team that is assigned to that and that project team needs to be briefed in and told what kind of project it is and what they’re doing and what the parameters are, what’s the start date, end date, the scope, all that fun stuff. And so previously, that was all locked in people’s heads, and they had to book a meeting with the sales team, get it all out of their heads because the sales team was not taking any notes about all of this. And so they would have to wait for this meeting and then hope that they got everything out of their heads that they had discussed during the sales process. And then they’d have to go into their project management tool and pick which template and set up that project, fill it all in and translate all of this kind of non-documented data into real data. And so, with automation, with this client, and I’ve seen it time and again, and I think a lot of people can benefit from this kind of stuff, is now their sales team uses an AI note taker in their calls. So every call that they have during the sales process is documented and tagged to that lead and what becomes a client. And so when they move it to the close one sales process, it goes into kind of a shared source of truth, runs an AI prompt against all of their different sales meetings and summarizes everything that was discussed. So then the team isn’t waiting for this meeting where they’re hoping to get stuff out of people’s heads. They can see the actual transcripts of the calls that were had, and they’ve got summaries of all the things that matter to them. And then from there too, it will go and say because there were three or four different things that were tagged in the CRM, it knows which template to choose and which project team this needs to be assigned to. So it will go to their project management tool and create that actual project for them, assign it to those people, and then it notifies them in Slack to say, hey, this is ready for you, here’s a link to your project, it’s set up. They go in there and they’ve got everything that they need already. So that’s kind of the enablement side of these things that people are spending hours a week just kind of updating stuff and figuring out what they need to go and set up, and the system is now done for them and gives them more than they got before with the enablement side of automation. 

Alex Bridgeman: What automations do you enjoy spending the most time on? Like is there a particular process like CRMs or sales or data organization or what have you? Like what areas do you find yourself enjoying the most and looking forward to the most? Like if there’s a project or a proposal, someone wants to work with you and they’re like, I have this problem, like is there a problem that you’re kind of hoping that they have so that you can enjoy working on it?

Justin Watt: We’re talking a lot about the sales process. I actually love working on the HR kind of people ops side of things. Because I think the employee experience and onboarding experience for so many companies is, everyone wants to do the right things and make it enjoyable, but it’s often not. And so, an example of what we’ve built multiple times now is, and you’ve probably seen this in your career, and I know I have in mine, where you as a new employee sign your offer letter, and over the course of the next few days or weeks, you are getting bombarded by, can you fill out this tax form? Can you fill out this thing for payroll? Can you tell us what size your clothing is for swag? Can you give us your address so we can add it to payroll and do this for sending your laptop and we need this? And suddenly you’ve got IT people saying, I’m going to set up your accounts and what’s your personal email so I can send the initial password, all that fun stuff. And you’re answering all of these different questions to all these random people you’ve never met and you haven’t even started your job yet. And then when you start your job, people are going, oh, right, yeah, you started today. Okay, here’s some folder full of stuff that I just want you to read today, and your laptop’s not set up yet, and, and, and, and, could go on for a while. And there’s so much power and literal revenue to be gained by having, or margin, whatever the role might be, but to be gained by people being onboarded better and being up to speed and effective in the role that you’ve hired them for faster. Not to mention their employee experience of, oh, this company has their shit together, for lack of a better word, because my employee onboarding was obviously so well put together that they must have everything else, a well-oiled machine, so to speak. And so a lot of the automations that we get to build in HR make me happy because we get to hear them that, oh, yeah, our team is happier, or our new employees, I should say, are happier because their experience is better, and our team is getting all this time back because they’re not chasing all this stuff. So, a good example of that is to go back to the 10 different emails that someone will get before they start, for a couple of different companies now, we’ve automated where kind of like sales is pipeline, hiring is really a pipeline of first interview, second interview, maybe there’s a case study or test or whatever it might be and then references and offer letter and all that fun stuff. And so a lot of HR systems, modern ones, will have it set up as a pipeline and they’ll have APIs where we can plug it into different tools. So, we treat it as that pipeline of once a new employee hits a certain stage in that pipeline, it will go and trigger this stuff happening. So their IT tooling is set up. The HR kind of person is notified of things being provided to them. They, on the inverse, will get one form where they only have to fill out their information once of here’s my name and address and all that fun stuff. And because you can connect all these tools and kind of going back to that infrastructure side of automation, you can connect the payroll system to the HR system and the IT system to that, whatever it might be. When it hits that stage in the pipeline, you can trigger all these automations to have happen for their onboarding. And then when they fill out that form, the information they provide can disseminate back to those platforms. So then they’re not getting 10 emails from four randos at this company they’ve never met before that they’re giving all this personal information to. They only have to fill it in once. So they get one email, one form, they get to fill it out once. And on the flip side, within the company, there’s all these people who are then saying, here’s the information that you needed, your tool is already updated, and half your work is done for you for onboarding this person. So that’s why HR side excites me, because employee experiences can always be better, and they’ll never be perfect, but automation helps them get there. 

Alex Bridgeman: The survey sounds pretty awesome because I’ve definitely been in roles where you’re like you’re giving the same information multiple different times and it’s not even anyone’s fault. It’s just like different systems need the same information. Not everyone is always on the same page of what everyone needs. But yeah, Survey seems like a pretty good fix for that. It’s automatically disseminated to different tools and whatnot. That seems like a pretty clean fix for that. I’m surprised more folks don’t do that. It seems very easy to implement. 

Justin Watt: The cool, I think at least, the cool thing about automation is you can kind of treat it like you do a product. Like the version of Uber in the app store right now is not the same version that launched four or five years ago. It has all these new features and new additions to it for better or for worse, whatever people think about Uber. But point being, you can kind of treat your process like you do a product where everything I’ve just explained to you is kind of version one. And we’ve now seen a couple of companies take it even further. There’s one example where we built all that. And then the HR team was so excited that they were like, what’s next, what else can we do? And so, a good example for them was every employee and their hiring manager or their manager when they start is supposed to develop a 30, 60, 90 plan in their first two weeks together. What does that first 30 days of success look like? What do we need to do to get there? And so on and so forth for 60 and 90 days. And so they were treating this as reinventing the wheel every time of every new employee, they have to go figure out what’s your role, what are all the things that generally in that role lead to success in the first 30, 60, 90. And so we started with one department and then moved through three or four others with hiring managers in that department to say what are your 30, 60, 90 templates? If you were to templatize this, what do they tend to be? And so the call it version two, if we’re to productize workflows and process in a company, the version two that we built was everything that I explained before, and then we templatized their 30, 60, 90 plans, and then because we knew and had that data coming in of what is this person’s role, who did they report to, we could then go automatically spin up a template of that 30, 60, 90 and send it to, through Slack, we would send it to their hiring manager to say, here’s the first pass done for you. So then it’s not this first two weeks of them being like, oh, right, I got to do that. I got to go figure this out. Half of it’s already done for them. And it’s not meant to be- this is still a human process where that person might have specific needs or their role is a bit different than a counterpart where there needs to be some updates to that 30, 60, 90 plan. That’s the cool part of like you get to evolve these processes. Once you’ve kind of got a baseline, you get to go even further with it. And that, again, saved hours of time for every new employee that joins and that’s a better experience for everyone. 

Alex Bridgeman: Yeah, absolutely. What else beyond the- I feel like there’s more to dive into on the HR process side. What other aspects of HR do you find, besides the onboarding, do you find the most energizing or fun to reinvent? 

Justin Watt: This is more of a personal one because I’ve worked within IT teams for so long, but it kind of still blows my mind at how many people in IT roles still treat it as a manual job themselves. Like, these are people who are working with tools, but I think, and it’s not a knock on them because that’s kind of, in a lot of cases, all that they know. But a lot of folks in IT will spend all this time spinning up all these accounts where your average SMB employee, I think, is using like 25 or 30 different apps nowadays. And so, for every new hire, IT has to go and create all of those accounts, but when you sit down with them and say, okay, which ones have APIs that allow us to spin up accounts based on information from other tools or data points, they go, oh, I didn’t know we could do that. And next thing we know, the 25 or 30 apps that they had to go manually provision and then on top of that, all the email groups that they have to add that person to based on their role, all the different shared files and folders and servers that they have to add them to, you can kind of templatize that to say if someone is hired into X department, they’re generally going to get these tools, added to these email groups, access to these files or folder structures. And so a lot of times, IT teams are spending hours per new hire doing this, and you can automate a lot of that. You can’t automate all of it, but you can automate a lot of it. And to me, that is such a fascinating thing because I think everyone has that experience at some point in their career of starting a new job and going, oh, we don’t have your laptop yet, or oh, you don’t have that tool yet, we forgot to set that up before you started, or you’re two months in and you’ve got a deadline on something and you realize, oh, I’ve never been added to that email group, I have no idea what’s going on there, or I need access to all these files and I was never added to that, and then you’re chasing for the next one, two, three days the IT team to say, can you add me to this, or your manager, whoever it might be. And so, the IT side of it is always ripe for disruption, so to speak, with automation. 

Alex Bridgeman: Yeah, certainly. You had this concept of like an anatomy of an automation. Can you talk about what you mean by that? 

Justin Watt: Yeah. We try to think of especially the enablement side of automation as you would a product that you’re building. So there’s kind of the front end, the back end, and the logic layer. And to be able to build a good automation, you have to figure out those three parts. And so that often comes back to the kind of process and technology side of those three pillars of the people, process, and technology. So the general anatomy of an automation is what is the back end or what is the data that is needed to be able to do this automated thing? And then what is the front end, as in, is it a notification email or a Slack message or Teams message that sends someone to notify them something’s done or ready or approved or whatever it might be. And then what is the logic layer? So between that kind of front end and back end, if it’s a good example is the HR stuff, where if the back end is this person is a operations manager and this person is a salesperson, they’re probably going to get a different set of tools, a different set of access and provisioning of things. And so you have to figure out what is the data points that we need. In this case, it’s who is the person, what is their role. And then what is the logic part of it, which is if it’s this role, provision X, Y, and Z, and if it’s this role, provision one, two, and three. And then the front end of sending someone the notification to say, hey, so-and-so has been hired into this role, we’ve provisioned these tools, can you double check and then send them their login information? Whatever it might be, that’s just one example. But the general anatomy of an automation is what is the backend data, what is the logic layer, and then what is the frontend that you need to deliver to someone. 

Alex Bridgeman: What are some quick wins you often see companies see opportunity for in some of these companies that you work with? Like if I’m an executive leading a company, ideally I’d love to get my team on board and excited about working in these automations and some of these sound more complicated than others. Like, where are the quick wins you often see that can kind of build some internal momentum and get some people on your side? 

Justin Watt: I would say in the abstract, before going into specifics, in the abstract, it’s wherever they have the best process and tooling already. Because a lot of folks hire us and they, this is on us to better expectation set all times, but a lot of folks will hire us, and then they are disappointed that they don’t have the automations up and running two weeks later because they didn’t actually have any tooling, their team has no actual process around that workflow, whatever it might be. And they haven’t really thought through that part. And so that tends to be a bit of a disappointment for them of like, oh, right, to do all this sales enablement automation, for example, we need a sales pipeline, and we don’t currently really have one, we’ve been winging it for the last three years or whatever it might be. We don’t have a proper CRM, we just kind of keep track of it in our email inboxes and a spreadsheet somewhere. So, I would say, in the abstract, that’s why we look for teams that already have kind of decent process and tooling in place because they’ll see a quicker win that way. And then specifically, we try to focus on inter-department automations as kind of a quick win because then you start to build momentum where people start to see the power of automation where the HR one is a good example or that sales one we’re talking about where you have multiple teams that don’t have the same role, don’t tend to talk to each other outside of when they need something from each other to empower their own role or their own department. And so a lot of people have this fear around automation of like, oh, it’s going to eliminate all the jobs and not really. I mean, candidly, eventually one day we will get there, but that future is far away. And so right now, AI and automation is augmenting teams. And so, the quicker win that we see is interdepartmental fit processes because then teams see the value of it. And then they start talking with each other, like do you know that this is happening now for us? We don’t have to worry about this anymore, whatever it might be. And then it just becomes easier to do the hard things, which is the teams that have no process right now or no data infrastructure, they get to point to the team that says, oh, they got a win in two months. And so, once we figure out kind of the infrastructure part of it, everything else becomes quicker and we can do that too. 

Alex Bridgeman: What advice do you often share with folks who are trying to implement more tools and automation in their businesses? 

Justin Watt: I always start with source of truth. Like what is their source of truth for data? That’s often why we don’t have any sort of commission agreement with them or anything. I like to stay chaste that way. But we often recommend Airtable for teams because usually they’ve got a dozen plus tools and no source of truth. And so, the fastest way to enable automation in a business is having a single place where all the data is fed into because that gives you the foundation to build the rest of the house on. And that foundation of a single source of truth then is a place to plug into from all these other tools and all these different teams that need that same data. An employee is a good example. An employee is going to have a name, a location, a job title, a salary, a start date, eventually an end date, all that fun stuff. And so, you can imagine that that single source of truth, then you can plug your payroll platform into that, then you can plug your IT systems into that, your project management, your whatever it might be. Having a source of truth for all of your data is really key and that’s usually the first piece of advice. And one of the first questions I’ll ask is what is your source of truth? And 80% of the time, it’s we don’t have one. And so that’s a good place to start. 

Alex Bridgeman: Absolutely. So, if folks want to work with you, how do they reach out to you and pitch you a project or tell you about what they’re working on and see if you can help them? 

Justin Watt: For you, I will say publicly my email address is justinwithswitchboard.com, but generally, Justin Watt on LinkedIn. If people want to search that, they’ll find me pretty quickly, I imagine, and then just_watt on Twitter. I share a fair amount of things there in terms of what we’re building and kind of what we’re seeing in the market that even if folks aren’t ready to work with us, it can be useful for them. And then if they want to learn more, withswitchboard.com kind of breaks down the services that we offer a bit more if they’re interested. 

Alex Bridgeman: Awesome. Justin, thank you so much for coming on the podcast. I loved talking through automation and tools and things that go faster. So, I really appreciate you sharing your time. This has been a lot of fun. 

Justin Watt: Thank you, Alex. Take care.

Related Episodes

Subscribe to our newsletter

Join small company investors, search funds, private equity firms, business owners, and entrepreneurs in reading the Think Like An Owner Newsletter.

Search
Generic filters