The build-versus-buy debate has a new answer for media companies

The build-versus-buy debate is almost soothingly familiar to anyone who has even a modicum of experience with technology conversations. But recent advances in AI are giving new shine to an old question: what should you build yourself, and what should you buy?

At a recent Wall Street Journal Technology Council Summit, Work-Bench co-founder and general partner Jonathan Lehr offered a useful way to think about it: “buy to operate and build to differentiate.”

White the distinction is simple, it’s still an important one, forcing media companies to ask which parts of the stack are worth owning, and which simply aren't.

Why the standard dichotomy fails 

For the past few years, companies have tended to treat build-versus-buy as an all-or-nothing decision. Build a proprietary AI stack and own the technology. Or buy platforms and move faster.

Neither approach works particularly well on its own.

Only 30% of agencies, brands, and publishers have fully integrated AI across the media campaign lifecycle, according to the IAB's State of Data 2025 report. At the same time, roughly 90% are using general-purpose tools like ChatGPT or platform features like Smart Bidding. The gap between using AI and actually integrating it into the way teams work is still significant.

That gap isn't necessarily about access to technology, but rather about where media companies are putting their effort. 

Try to build everything and you quickly run into engineering and maintenance costs. Buy everything and you can end up with a collection of point solutions that don't share data, context, or workflows.

The answer sits somewhere in between.

Buy to operate

There is plenty of AI infrastructure that media companies simply don't need to reinvent.

Foundation models, vector storage, orchestration, evaluation tooling, observability, these are increasingly well-developed capabilities. Building them internally can consume significant engineering time without creating much differentiation.

Lehr's point was that off-the-shelf software can get companies “80%, 90% of the way there in terms of capabilities.”

Take the 90%.

A useful test for the buy column is: if a competitor could purchase the same thing tomorrow and get essentially the same capability, why build it yourself?

The goal isn't to win an engineering competition. It's to put your engineering resources where they actually create an advantage.

Build what you know 

The more interesting work starts after the foundation is in place.

Every agency has things that can't simply be purchased off the shelf: its audience taxonomies, its approach to measurement, its understanding of clients and categories, its planning methodology, and the decisions its strategists make every day.

That's where the investment starts to matter.

Build the connective tissue between your data and your workflow,  the logic that reflects how your teams actually work, the evaluation criteria that tell you whether an AI workflow produced something a strategist would trust and put in front of a client – a competitor can't simply license that acumen from the same vendor.

The common thread is expertise: build where your media company knows something the market doesn't.

The infrastructure distinction

There's a catch, though: you can't build differentiation on top of infrastructure you can't actually work with.

This is where some AI strategies fall apart. A point solution might produce a useful output, but if it sits outside your environment, relies on data exports, can't follow your governance model, or doesn't let you add your own logic, you've created another silo.

That makes it difficult to turn a purchased capability into something uniquely yours.

The infrastructure underneath your AI should make that possible. It should sit close to your data, work within your security and governance requirements, and give your teams room to build on top of it.

That's the role Akkio plays for media companies. Akkio runs inside your cloud environment and sits on top of its data, giving teams a foundation for productionizing agentic workflows across campaign strategy, audience building, analysis and modeling, media mix modeling, and performance measurement.

Akkio also provides observability and governance so teams can understand how workflows are operating and maintain the controls required in an enterprise environment.

Additionally, because the underlying model landscape keeps changing, Akkio is model-agnostic. Media companies don't have to rebuild their architecture every time a new model becomes the one everyone is talking about.

Two questions to ask before your next AI decision

The right answer will look different from company to company. Security might be the priority for one firm. Proprietary methodology might matter most to another.

But before investing in another AI capability, two questions can help:

1. Would this capability differentiate us? If a competitor could buy the same thing and get to the same place, it's probably something to buy rather than build.
2. Can we build on what we're buying? If the platform can't work with your data, governance, or media-specific logic, it may solve an immediate problem without helping you build toward something bigger.

The answers will result in a more focused AI roadmap.

Instead of trying to build an entire AI stack, media companies can spend their resources on the few layers where their expertise actually gives them an edge, and use existing technology for everything underneath.

Want to learn more? Watch our webinar, Beyond Build vs. Buy: The AI Integration Imperative for Media Companies, or request a demo to see how Akkio helps media companies productionize agentic workflows inside their environment.

‍