Build vs. Buy GenAI: “The Most Overlooked Aspect is Time to Value.”

By now, the promise of GenAI is undeniable – offering transformative potential to streamline workflows, boost efficiencies, and deliver competitive advantage. Yet, for many organizations, the journey to implement AI is far from straightforward. What begins as an exciting proof of concept often faces roadblocks, including technical complexity, integration hurdles, and operational risks, leaving businesses grappling with wasted resources and missed opportunities. 

In a discussion with Dorian Selz, CEO of Squirro, we explore a question enterprises invariably face when setting off on their AI journey: Should they build AI solutions in-house or, instead, buy tried and tested technology. Of course, ultimately, the decision is not just about the tech, it’s about choosing the path that aligns with an organization’s strategic objectives and makes optimal use of their internal resources while steering clear of the common pitfalls that derail AI initiatives.

JO: How well would you say that business leaders understand the value of GenAI?

Dorian Selz: Frankly, most don’t see the transformative impact that it will have. Many still view generative AI as a glorified chatbot or a tool for simple queries. What they're missing is that this technology has the potential to completely transform their business models and processes. Similar to how digital music distribution disrupted the entire music industry, Generative AI can atomize and disintermediate processes, leading to a fundamental shift in how companies operate.

What are some of the biggest misconceptions that companies have when
starting their Generative AI journey?

They think it’s easy – that’s the biggest misconception. Many think it's as simple as applying ChatGPT to their business. What they fail to see are the complexities involved in data preparation, infrastructure, security, and maintenance.

Many companies stumble when starting their Gen AI journey. What are the recurring obstacles they face?

Companies often hit the same roadblocks, and these typically fall into three categories: strategic, technological, and operational. Even though it’s a question of corporate strategy first, many jump on the Gen AI bandwagon without carefully thinking through the technology’s business value. Because of that, they are unable to gauge its usefulness and its impact on their top or bottom line.

What about the technological and operational challenges?

First, many underestimate the importance of data preparation – AI models need clean, structured data to deliver valuable insights. As always, garbage in equates garbage out, and simply attaching AI to your system won’t make your data chaos disappear.

Second, scaling AI solutions requires significant infrastructure and resources, which can be costly and complex. Most organizations that run small pilots on a couple of thousand documents haven’t thought through what it takes to bring that up to scale: from the infrastructure to the types of embedding models and their cost-precision ratios. Because, contrary to what you might think, it isn’t cheap. The massive compute required comes with a high price tag, and the seemingly cheap $20 ChatGPT subscription doesn’t tell the whole story.

Then, I haven’t seen any companies run on a single data access scheme, granting everyone from the apprentices to the director the same permissions to access data. Building permission-enabled, secure GenAI at scale with the required accuracy is really hard, and 95 percent of companies that try to build it themselves will fail. Why? Because it takes expertise, and addressing that challenge isn’t their USP.

Finally, operational challenges often arise from underestimating the ongoing effort required to manage GenAI systems, which need constant monitoring, updating, and fine-tuning. Security and access control also pose significant operational challenges, especially with sensitive data.

What proportion of prospects first attempt to build the technology before reaching out to Squiro for support.

Most people start to build themselves. But lately, we’ve been hearing back from companies that we'd been in touch with months ago that ran into issues – insufficient quality, accuracy, or security to pass integration tests, for example. But, here’s the thing: For us it’s essential that companies gather some experience on their own before reaching out to us.

Why?

Because you need a knowledgeable counterpart on the buyer's side to have productive conversations about GenAI's strategic impact. Without understanding the technology's potential, they may have unrealistic expectations or suggest unfeasible applications. So, some initial experimentation is beneficial.

Beyond financial savings, how do customers benefit from partnering with Squiro?

The financial side is trivial. It costs a given amount of money for anyone to develop a GenAI platform. Distribute that over dozens of customers, and each one only ends up paying a fraction of that cost. But the most overlooked aspect is time to value. Instead of focusing on business strategy, business value, and organizational transformation, many organizations invest time and resources in developing a technology that they could acquire at a fraction of the cost, while freeing them up to focus on what really matters.

How do you expect 2025 to differ from 2024 in terms of GenAI adoption?

Reading the tea leaves, it looks like 2025 will be the year that most organizations understand they can’t simply build their platforms by themselves but, instead, need to acquire. It will also be the year where we separate the wheat from the chaff when it comes to the performance of enterprise level setups.

Finally, what recommendations would you give prospects or people looking for a GenAI platform provider? What should they be on the look out for?

I always recommend working with a provider who can prove that they have delivered permission-enabled deployments at scale with accurate results. Because these are three core components you’ll need for every GenAI use case, from a simple chatbot all the way up to a complex deployment with agentic workflows.

Ready to transform your operations with Squirro?

Get in touch or book a demo!