VIDEO TRANSCRIPT:

In 2026, every company faces the same choice. Do you build your own AI or buy something that already exists? Make the wrong call and it could cost millions. We asked our experts to weigh in.

90-95% of AI pilots fail simply because they're done in a silo, for a point need.

Let's see where we failed or why it failed.

LLM developers are hard to come by, right? Those don't just grow on trees.

Building means creating your own AI tools and infrastructure, tailored to your data and business. Buying means using off-the-shelf platforms or ready-made AI solutions that plug in fast.

We're dealing with a lot of our clients on this exact question right now, and I think what really is driving a lot of the buy need is the urgency to launch something quickly, the ability to actually prove out some of the value in the use cases.

I think a lot of it comes down to kind of their people process and technology that's already in play. Do they have people who have the skill and talent to actually develop these AI tools?

They've got tools. They have so many tools coming out of there. Their ears, they've got teams on the ground building their own tools using tools for their specific function. We're now having conversations about, okay, we're drowning in tools.

Speed is tempting, but rushing the decision can cause chaos to follow fast. Before you even think about building, you have to deal with your data.

To agentic AI and AI to work best, you desperately need clean data. You need data that makes sense. It's garbage in, garbage out is real with AI.

90-95% of AI pilots fail simply because they're done in a silo for a point need as a point solution where data inputs or certain inputs are curated to a level that is unrealistic yet within an enterprise lens. And so for that reason, you can't scale. And so then, of course, a pilot fails and lands flat.

Once you've dealt with your messy data, then there are some reasons you might want to build.

AI needs to be owned by the brand. It's not something that can be dedicated to a platform.

I think that the path is absolutely to build, and the reason being is that you want to build in a way that is going to scale with your organization and not be tied to a single solution vendor.

No one else has that same retail mix and that same context that you do. that you do that's where it really makes a lot of sense to build but you need to build in those foundations first to be able to really get the value out of that investment

On the flip side, sometimes it does make sense to buy an existing product. Not everything needs reinventing.

And if there's a platform out there that you're able to plug in that can help agent builder capabilities, context libraries, ontologies. That is a game changer in terms of speed. Things that may take you six months to a year to build may take you a month or two to configure.

For example, dynamic pricing. We know that there are some extremely mature vendors and solution offerings for dynamic pricing that already plug in some of the super complex inputs.

It's often a really good option to buy a tool that's integrated in tools that are already part of people's daily work. and help encourage usage and adoption of those tools.

But sometimes, does it make sense to do both? Build for the long-term differentiation and buy for the speed.

You don't want to be so bogged down in... Setting up the foundational capabilities and the infrastructure needed for you to scale.

You create the space for organizations to be able to pilot and test and move quickly when appropriate, but also having a top-down enterprise looking strategy that sets you up for scale, sets you up for longevity, adoption and value realization.

We're going to see a lot of change, a lot of instability. The out of that, we're really going to start to see the longer term, bigger impact of this wave of AI transformation.