The most critical decision in building out enterprise AI: Build in-house or bring in a vendor
From smart homes to robust intelligent edge ecosystems, AI is one of the hottest topics in tech and society as we embrace the new decade. While consumers are debating the benefits of one smart light bulb versus another, financial service companies are planning, adapting, and debating just how they can use artificial intelligence to empower their employees, delight their customers, optimize their operations, and transform their services.
Today, we’re going to explore exactly why a financial service company should consider building their own AI solutions, who’s best positioned to benefit from building their own tools, and how two similar organizations built their own solutions with two radically different outcomes.
Case studies in AI innovation
Let’s start in the world of the practical. In 2018, we were approached by two companies. Each wanted to build their own AI models from scratch and saw two very different outcomes.
The first was specifically interested in using AI to build the inaugural chatbots for their organization. The task was simple enough, company AI champions brought the concept to their leadership and those executives tasked them to build out a minimum viable product (MVP) to earn funding. Despite little technical experience and limited data science backgrounds, this team decided to keep their efforts in-house and build through online best practice guides and tools. It didn’t work out so well for them. The project fell apart, the budget fell through, and the AI efforts were abandoned.
And that leads to our first critical takeaway—if your future AI budget hinges on a no-budget AI trial, test an existing solution. The odds of in-house custom modeling success with no budget and no experience are slim at best.
The second organization we spoke with wanted to capture AI workloads, but like their colleagues, had no background and no clue where to start. Fortunately, their leadership was excited to pursue AI and had given them a small budget to build their AI MVP. This group immediately brought in an AI partner to help guide their process and build out three potential ideas. Of the three final deliverables, two failed, but the third, a sales analytics build, was well received by executive leadership and given a significant budget to create a data science department.
Notice something interesting from these two customers – success hinged on leadership buy-in, patience, and funding. Yet, it is almost always the case that leadership buy-in and funding requires a track record of AI success and a complete vision on what the organization will get out of AI.
Thus, we can see our AI build-it-in-house catch-22 — funding follows results, results follow funding. We’ve have seen this vicious cycle appear over and over again, and it highlights why Group #2 was able to succeed:
- They knew when to ask for help
- They didn’t hang the project’s success on one project.
Partnership is absolutely critical to the success and failure of these organizations. Whether you’re a shoe retailer, a salon, or a chemical engineering plant, organizations of all shapes and sizes recognize the business value of AI.
Finding an AI partner is essential to generate executive buy-in early in the building process. These are technical experts that live and breathe artificial intelligence but can also speak to the business value these projects represent. They understand exactly how AI can shape a businesses’ future success and know how to bring those visions to life. This combination of expertise and engagement can help not only deliver a more compelling final product but can help secure budget from skeptical leaders earlier in the building process.
Why build in the first place?
Both of our example companies came to the table with specific desires and little direction. In both cases, their decision to build came down to one key variable: control. In our examples, Group #1 wanted to own the project from start to finish so desperately that they refused to seek outside guidance and it cost them. Meanwhile, Group #2 was able to retain that control by admitting their own limitations and securing a partner that could amplify their vision. When building your own AI solution, you have direct input into how that product is shaped and applied. No one understands your needs, data, and opportunities better than you do. You control the quality of the deliverable, for better or worse.
Likewise, custom-built AI solutions are easier to:
- Reproduce
Few solutions will ever be as expensive as your first. As your data science team continues to iterate through, those previous experiences help expedite the process and help scale your efforts to other divisions.
- Explain
Often, the most valuable piece of a model is the explanation of why it works instead of the output itself. By building the solution yourself, you have a clearer understanding of how and why your model operates the way it does. This not only opens up the opportunity for new strategic business decisions to ripple from the understanding of your model, but also creates a potential for you to grow your thought leadership in the application of AI in your field.
- Secure
Every organization that touches your data adds another opportunity for outside threats. Despite their best efforts, outside vendors will never be 100% secure.
So, who should build?
There’s a place in AI for everyone, but that doesn’t mean that finance firms shouldn’t be strategic in their approach to how and where they spend their AI dollars. Group #2 found a specific niche within sales workloads where their AI solution was able to stand out but recognized that they couldn’t make that jump alone. Building your own AI solution demands a level of honest introspection about your capabilities, expertise, needs, and resources.
In the coming weeks, we will explore the other side of this conversation—why you ought to buy an AI solution, and who’s best suited for an off the shelf offering. Ultimately, there’s no right or wrong answer, only action, and indecision. If you want to chat about your AI options, don’t hesitate to contact me at [email protected].
