Takeaways
- Focus on business value: The most effective data teams understand the business context and proactively provide insights, actions, and recommendations that drive measurable ROI, rather than just fulfilling data requests.
- Automation opportunities: Look for repetitive processes in various departments (e.g., sales, marketing) that can be automated using data. This can lead to significant efficiency gains and improved customer engagement.
- AI integration: Encourage teams to start initiatives with AI and provide access to AI tools. Regular âShow and Tellâ events can foster innovation and knowledge sharing across the organization.
- Data infrastructure importance: High-quality data infrastructure is crucial for leveraging AI effectively, especially for techniques like Retrieval Augmented Generation (RAG) that enhance AI models with company-specific information.
- Talent retention: Create an environment that allows for continuous learning, experimentation with new technologies, and growth opportunities to retain top data talent.
Video
Transcript
Oleg
Hi, everybody! Welcome to Devico Breakfast Bar! Here we speak with different people involved in the business landscape, share their expertise, delve into the latest technology trends, and explore the ins and outs of IT outsourcing. Iâm Oleg Sadikov, and today Iâm excited to have Blake Burch, the co-founder of Shipyard, a data orchestration platform that allows data teams to launch, monitor, and share their business solutions. Donât forget to subscribe and hit the notification bell so you donât miss on new episodes. Hi, Blake! Could you please start by telling us about your journey from your early career experience to co-founding Shipyard? What motivated you to start this venture?
Blake
Sure thing. So, at the beginning of my career, I was not in a technical field at all. I actually started out doing account management for an advertising agency, where I was responsible at the time for running Google Ads, for OpenTable in the United Kingdom. And I had a lot of manual work that came along with that, that I was having to pull data into Excel spreadsheets. I was having to manually filter things to figure out how I should change a bid, or how I should change a budget, or which keywords we should start showing up for. And then, whenever I finished things in Excel, I would have to go and format it into a bulk upload sheet format that I could then just add to whatever tool we were using to manage ads at the time. This whole process seemed kind of menial to me. And it was something where I was just trying to figure out, âOkay, how do I make the work that Iâm doing take less time on my side and also be more consistent?â And so, I gradually started teaching myself various technical skills: SQL to be able to get data out of a database; Python to be able to send data to an API. And what that resulted in was me trying to figure out how can I take my everyday processes and automate them using this data. And that resulted in most of my work getting automated, and people started to catch on, and theyâre like, âWait, we want to do that for these other clients along the way.â And so I was like, âOkay, sure. I can write something to do that for your client.â Then I had to learn about abstraction and making things easy to use but slightly changed for every single client.
And that led to me ultimately kind of leading out and growing a data team at that agency as we found more and more opportunities of things that we could do to automate account management for our clients, but also to go even further and do interesting things like natural language processing on influencers so that we could react in basically real time and bid up and down based on products that we had that they were talking about. So, it was kind of a cool journey there to be able to go from being totally non-technical to leading out a technical team and building out solutions. But what I saw was that time and time again people kept on running into issues, not being able to know how to use their data effectively, and this was at big Fortune 500 companies. And I just realized that there was a bigger market opportunity to say, âHey, the types of things that weâre creating and the tooling that we had internally at that advertising agency could be used for way more than marketing.â And so, I worked with the CEO there to figure out how we could spin that technology off into its own separate tool, which is now what we know as Shipyard today. But I always kind of had an inkling that I wanted to be an entrepreneur. I graduated as an entrepreneurial management degree and was always interested in the tech space, but just finally found the right opportunity in the market to really sink my time into and to go after.
Oleg
Thanks for the detailed answer. What key lessons have you learned about entrepreneurship while building a Shipyard from the ground up?
Blake
Things never go the way you expect them to, man. Itâs something where you kind of just have to teach yourself to go with the flow, right? Youâre going to have ups, youâre going to have downs. And oftentimes, youâre going to have imperfect information that you have to make a decision on, and you have to live with that decision. It doesnât mean that that is going to be 100% your path. You can deviate a little bit as you get more and more information over time, but youâre going to have to just trust yourself to make the right decision in the moment and be okay with that and knowing that it could go wrong, things could go bad. Almost every single time that I think things are going smoothly, something terrible happens. And every single time that things donât feel like theyâre going as great, something fantastic happens, like a new opportunity or something else like that. So, you just kind of have to be okay with the journey.
Oleg
Agree. What strategies have you found most effective in leading and managing data teams to ensure optimal performance and collaboration?
Blake
So, I talk with a lot of different data leaders in the space, and the thing that I think really kind of sets the highest quality teams apart from others is having a true focus on what business value youâre driving. Iâve even seen some recent like state of analytics surveys and things like that that have come out where a lot of data teams are determining if they are valuable by seeing if people are happy with the work theyâre doing. But all that means is that you are happily answering peopleâs requests and grabbing the data that they want, creating a dashboard that they want. And it kind of gets your team into a state where youâre just doing request stops effectively. But thatâs not what a valuable data team does. A valuable data team has business context, and understands whatâs going on, and is able to dive in deeper with the business users to figure out what are they looking to change, what sort of things do they care about, what levers do they have in the business to pull. And then you proactively come with insights, and actions, and recommendations for that team. And then youâre able to actually verify, âHey, this dataset that I made resulted in this decision, which drove this amount of revenue or had this amount like ROI for the business.â And so, data teams that really focus on driving that business value and understanding how the business works, and theyâre acting proactively, theyâre going to have the best impact, and ultimately, have the best collaboration, and be seen as less of a cost center and more of an actual value driver, which I believe data truly has value or the ability to be.
Oleg
Can you share some practical examples of how businesses can leverage data to automate various aspects of their operations?
Blake
I think one of the easiest things to think about from practical examples comes down to sales and marketing. So, letâs imagine youâre a software as a service business, and you can track all the different click events that people are doing in the application, when they last logged in. And what you might want to do is get people to use very specific features. So, you can target people that have used the platform for more than a certain amount of time, but they are actively not using some feature you want them to use. So, you send them an email with guidance and tutorials, and ask them to reach out, and everything else like that. Thereâs also examples where maybe you could find people that have abandoned their cart and make sure that they are getting automatic emails in the e-commerce space so that they ultimately, hopefully, purchase that product for your business, or even your sales team themselves sending them notifications to say, âHey, these people exhibited a behavior that means they might be likely to buy from us. You should reach out to these points of contact.â Marketing and sales, in my opinion, is one of the easiest ways to actively use the data that you have in hand. And a lot of the tools nowadays make it easy to be able to react to this data. Itâs just on you to try and figure out how you link it up and decide, like, when someone hits this specific metric, weâll go and send them this series of emails, or notifications, or anything else like that. But itâs not just limited to sales and marketing. Ultimately, any process in your business that youâre doing day in and day out in a repeated fashion can usually be automated in some form, and usually, thatâs reliant on just a little bit of data to make sure that it happens successfully.
Oleg
In what ways do you see artificial intelligence being applied to streamline the process of writing code? And could you provide insights into the specific applications and potential benefits of AI in this area?
Blake
I am super excited about what AI is going to be doing for code development. The main thing is just trying to throw requirements at it, problems that you want to solve, and the exact description of what you want the end result to look like and seeing what itâs able to generate. So just as a tangible example, I had a period of time â it was like a 30-day trial for me â I was like, âIâm never going to write any code. Iâm always going to throw the problem at ChatGPT and see what happens.â And ultimately, it did pretty much everything I needed. So, some tangible examples, I would have webinars that were online where I could see the attendees. I would take the HTML of the attendeeâs page and be able to throw that into ChatGPT and say, âHey, can you make an Excel sheet for me that has all of the attendee names, their place of employment, whatever their job title is?â And itâs able to parse through all the HTML and provide me with that finalized file. I had an example at work the other day, where we had a bunch of different scripts that each had a different ID in them, and I needed all of them to be placed in a SQL statement I was writing to classify each of these scripts. And I could have gone into each script individually, copied and pasted it out, but no, instead I just decided to zip up the entire folder, throw it to ChatGPT, and say like, âHey, the ID kind of looks like this. Can you find this in all of these different files and then provide me with the already formatted SQL?â And it was able to do it. So, it doesnât have to be something big like that. Iâve used it for small things like, âHey, can you make this background transparent? Can you convert this JPEG to a PNG? Can you take this PDF and turn it into a Word doc?â Those are not traditional examples, but so many things like that can be done through these AI tools, and all theyâre doing is writing code in the backend. If you want to do that in a repeated fashion, you can copy and paste the code and deploy it anywhere you want. So, I do find it good to kind of throw most coding tasks for one-off utility scripts and things like that that youâre trying to get done and just to see how powerful it can be.
Oleg
As someone with experience in data strategy and automation, how do you foresee technology evolving in the future to enhance and automate business processes in this area?
Blake
I think AI is going to be extremely powerful for businesses. Itâs going to be integrated so much into the products and how they work that itâs kind of going to be indistinguishable from other things, like actually just writing code itself. But Iâve had a lot of conversations on this, and I really think the thing thatâs going to set businesses apart is a high-quality data infrastructure because a lot of the AI models right now are trained on very generalized data. But some of the best tools being created are augmented by any sort of data you have available on hand. So, one of the common methods right now is called RAG. Itâs Retrieval Augmented Generation, where after someone asks the question, you use search to look across all your documents, all your data to surface potentially relevant information that you then provide to the AI so that itâs able to provide the best possible response. So, this could be things like customer support documentation. It could be things like searching internal docs so that people can figure out what your business processes are. But all of those rely on you having very clean data sets with high-quality information thatâs available. And I think if people can figure out how to get their data pipelines, their data infrastructure correct, theyâre going to be able to reap the benefits of high-quality AI for their business.
Oleg
What advice would you give to organizations looking to integrate AI-driven solutions into their coding practices and business processes effectively?
Blake
I think it ultimately comes down to having good top-down leadership and expectations of AI usage. So, as an example on our side, in 2023, at the beginning of the year, we told the whole team, âHey, you need to start every initiative that you do with AI. It doesnât matter if youâre on the engineering team. It doesnât matter if youâre on the marketing team.â For us, it was all about experimentation. We need to make sure that everyone is trying to throw everything at the wall so we can actually see what sticks and figure out how to use AI more effectively. But itâs not just like setting that top-down; itâs making sure that you follow up on things. So, we have monthly events on our side called Show and Tells, where we are just having people show us what theyâve been working on, new things that theyâve figured out how to accomplish, which shares those ideas and creates this spark of innovation across the organization. So, I think really setting those things in place and making sure that people know, like, âHey, we want to use AI more frequently.â That is something that will really help make sure that it gets incorporated more into the engineering and coding space. I think the other aspect is just enabling those tools, making sure that you give people in your organization access to tools like ChatGPT, or the OpenAI API, or doing things with Anthropic, or Mistral so they can test things out, giving them access to Copilot. All of those are important towards making sure that AI can really get incorporated into your coding practices.
Oleg
The events you mentioned youâre having, are those kind of hackathon events?
Blake
For us, theyâre not hackathon events. Itâs mostly just having dedicated time on a monthly basis to have people show others internally what they have built and what they are using AI for.
Oleg
Oh, okay.
Blake
Because oftentimes, that information gets lost if itâs not being asked of or not being surfaced.
Oleg
What are some common challenges or misconceptions businesses face when considering the adoption of AI for code writing or business process automation, and how can they overcome them?
Blake
I think a lot of people expect it to be perfect the first time around. They give all of this information, and it generates just this perfect script that works exactly how they wanted it to. My best recommendation is thinking of AI more like a junior developer. You have to provide it with the right level of information, and youâre going to have to coach it a little bit, right?
Oleg
At this stage, a junior developer.
Blake
At this stage.
Oleg
Maybe in a year, we can upgrade it to the middle.
Blake
So, I think that yes, it can get better there. And I think it definitely will, but I think thereâs always going to be an aspect that is tied to language, right? The output of the AI is only going to be as good as the description that you gave of the problem that you wanted solved in the way that you wanted it to be solved. I think a lot of people struggle to provide that right level of information. Some people might call it prompt engineering. But to me, itâs really just having mastery over the English language in order to describe exactly what I want in detail. Because most people arenât good at doing that, and so the result that they get back from the code is that. âOh, well, actually you didnât do this one thing that I really wanted.â Most of the time, thatâs because I didnât describe it beforehand. So, it kind of makes you reflect and understand that you have to make sure that things are highly specific. And some of that could be interpreted in the future with AI models getting better, but I think thatâs one thing to really think about. And I think one of the examples that a lot of people have done with coding rather than writing the actual scripts is using it in the backend to automatically generate documentation or automatically generate tests. Thatâs one way that itâs hard to really get wrong because youâve already written the code. And itâs one way that can help accelerate what your team is building with minimal effort.
Oleg
Thanks for the detailed answer. How do you navigate talent scarcity in the highly competitive fields of data analytics and engineering, and what strategies have you employed to attract and retain top talents, in particular your company?
Blake
I donât think I have all the answers for talent scarcity. I get still difficult being a smaller organization on our side. The best thing that we do on our end is providing high-quality content, whether itâs on our blog or on our YouTube channel, thatâs enticing and engaging towards technical people that they would come to respect and come to enjoy. And thatâs something that can help us set us apart. But when it comes to retaining, thatâs something that I feel much more confident on than the actual procurement of talent, where itâs really all about making sure that youâre constantly aligning with the growth opportunity that people want. So, oftentimes, people donât want to just do the same job continuously. And as a business, if you want your business to grow, youâre constantly going to need new things that youâre digging into. I think as long as you create an environment where people are able to grow, learn new things, experiment with new technologies, and are able to see that constant change, thatâs one great way to make sure that people are going to stick around. And I find that the people that have that sort of desire and want to learn and the curiosity, like those two things go hand in hand, as long as youâre able to help facilitate that, youâll be able to retain top talent for a while.
Oleg
I know that in the early stages of your company, you outsourced your development needs. In your opinion, what are the main advantages and disadvantages of IT outsourcing, and how to navigate these factors to maximize the benefit while mitigating potential risk?
Blake
One of the biggest advantages in the early stage was just being able to get access to talent that we didnât have yet. We didnât have full-time hires for various skills. And so, one thing that we did initially was hiring out a front-end engineer on a project basis to be able to help us take the application from the UI and UX that we had built on our side and being able to put all of that sort of scaffolding and stuff in place. So, the expansion of skills that you donât currently have and the ability to get that work done relatively quickly because youâre hiring someone thatâs more of an expert at it â thatâs super good there. One downside, though, is that thereâs a lot more project management that goes into things than I think a lot of people realize. And if youâre not very clear about your expectations, if you donât have a very good idea of how something needs to be built out, then youâre going to have a lot more back and forth. Itâs not something where itâs like an easy set and forget it. Like, âHey, I outsourced this project. So, now itâs just going to get done.â You still have to check the code. You still have to talk through things. You still have to go back and forth on revisions. And so, I think a lot of people just donât think of that aspect. So, in general, I recommend just being more specific about exactly the work you want done, how you want it done, and processes, and all that good stuff. The same thing youâd have if you were hiring an employee to do the work.
Oleg
Reflecting on your business operations, are there specific tasks or projects that you believe could potentially benefit from outsourcing, and what criteria would you use to determine their sustainability for external collaboration?
Blake
So, kind of relating to the answer I just had, anything that itâs project-based, I think, is the best opportunity for us, and anything that doesnât rely on other dependencies as much. So, weâre in a very unique situation because our product has a lot of open-source templates that we built. They are all built in Python. And the ultimate goal is that they can integrate with some service to upload and download data. It needs to run on its own in Python outside of our tool. So, thereâs no dependencies. We know the exact tasks that need to be done, and theyâre very concrete things that we can hand off to someone. And thatâs what we spent a lot more time outsourcing in the early days of our product, just because it was much easier to manage there. I wouldnât mind outsourcing things for various data projects that we have either. If we had all this event data, and we wanted to try and figure out anomaly detection that we could add into the platform, that doesnât require them to have any knowledge of the platform. It just requires them to build out some machine learning model to ultimately kind of score things and return back the right results that then we can react to in the product itself. So, really my kind of recommendation for the outsourcing stuff is just finding things that are very concrete projects, with finite deadlines and endpoints that donât require too much knowledge of the internals of your organization or the internals of your product.
Oleg
Blake, thanks for the answers. I really enjoyed having you on the video podcast. The information youâve shared is definitely valuable for our auditory, and especially your amazing journey from not being a tech guy to becoming a guy that can do coding, Python knows, and understands how all this stuff works. Itâs respectful. Not many people are capable of doing that.
Blake
Well, thank you. I appreciate it.
Oleg
Thank you for joining my podcast, and wishing you all the best in your entrepreneurship journey. If you enjoy our discussion and want to stay updated on future episodes, donât forget to subscribe and hit the notification bell. That way, you will not miss on the latest insights and conversations from Devico Breakfast Bar. See you in a week!