Service as Software: When Your MediaTech Stack Starts Doing the Work for You

AI is moving media technology beyond managing work. The interesting question is what happens when the software becomes responsible for doing it.

By Philippe Brodeur, Founder & CEO, Overcast

This is the first in a three-part Overcast series exploring the shift from Software as a Service to Service as Software. Next, we’ll look at Why Enterprise Software Pricing Has to Move Beyond the Seat, followed by Stop Buying Video Software. Start Buying Video Outcomes. Links will be added as the series is published.

For 20 years, enterprise software has had a remarkably good deal.

We bought the software. Then we supplied the people required to make it useful.

A MAM could tell you where a video was. Someone still had to find the right version, check it, move it, clear it, prepare it, chase the approval and make sure it ended up where it was supposed to go.

AI changes that bargain.

If software can understand the content, understand the rules around it and understand what should happen next, simply presenting that information to a human starts to look rather old-fashioned.

The next generation of enterprise software won’t just help us do the work. It will be expected to do some of the work itself.

Welcome to the slightly awkwardly named world of Service as Software.

SaaS gave us tools. Now what?

SaaS transformed enterprise technology. It made software easier to buy, deploy, update and use across distributed organisations.

But the underlying contract barely changed.

Software provided capability. People converted that capability into an outcome.

The CRM helped the salesperson sell. The editing system helped the editor edit. The MAM helped the media team manage.

Service as Software moves that boundary.

My definition is deliberately simple:

Service as Software is software that delivers a defined piece of work or operational outcome, rather than simply giving people a tool with which to produce it themselves.

That distinction matters more than the latest AI feature list.

Click the sparkle

Almost every enterprise product now has an AI story.

Generate this. Summarise that. Ask a question. Find something similar. Click the little sparkle.

Some of it is genuinely useful.

But an AI feature and an AI-operated business process are very different things.

Recognising what’s inside a video isn’t enough to act on it. The system also needs to know which version is approved, who can use it, where it can run, what rights apply and what should happen next.

It needs content intelligence plus operational context.

A model might recognise the celebrity in an advert perfectly while having no idea that the rights expire on Friday.

That’s the difference between clever AI and useful infrastructure.

Spend humans wisely

Media businesses contain extraordinary amounts of undocumented intelligence.

Someone knows which master is actually current. Someone remembers which export breaks the downstream system. Someone knows the rights exception. And somewhere, Sarah is still receiving Teams messages because the official workflow doesn’t quite work.

Good people become the glue between imperfect systems.

That isn’t always expertise. Sometimes it’s expensive human middleware.

Service as Software gives us a useful way of separating the two:

• Judgement? Keep the human.
• Predictable exception? Automate the common path and escalate intelligently.
• Entirely routine? Ask why a person is still doing it.

The objective isn’t to remove people from media operations. It’s to stop wasting them on work that no longer requires them.

Count the work: the economics

This is where the economics gets interesting.

If software performs more of the operation, counting users becomes a surprisingly poor measure of value.

Imagine two media businesses.

One has 500 platform users but still needs people to search for assets, reconcile information, move files and coordinate routine workflows.

The other also has 500 users, but its content is understood, predictable work happens automatically and the same team can process considerably more video.

Which company is getting more value from its technology?

It’s undoubtedly the second.

That suggests a more useful set of enterprise technology metrics:

• Hours removed
• Turnaround reduced
• Assets reused
• External costs avoided
• Capacity created
• Work completed

Not everything needs a dashboard. But those measures get considerably closer to the reason the CFO signed the cheque.

Diageo didn’t need another repository

We’ve seen this at Overcast.

Diageo had more than 60 agencies creating and storing video across hard drives, desktops and cloud services. The problem wasn’t producing content. It was the operational work required to control, find and reuse it.

Overcast MAX became part of the Diageo Content Hub, creating a common environment for teams and agencies to ingest, manage, collaborate on and archive video.

Within six months, more than 500,000 videos had been ingested and AI-tagged, with 60+ agencies and 1,500+ marketers onboarded.

More importantly, Diageo reported £1 million saved in the first six months, while processes that had taken days, weeks and sometimes months could increasingly happen in minutes.

That’s the important bit.

Putting 500,000 videos into a system is an implementation metric.

Saving £1 million is an outcome.

And the difference between those two sentences is where I think enterprise software is heading.

The MAM gets a job

For years, we asked media platforms a fairly predictable set of questions.

Where is my content?

Who can access it?

Which version is current?

What rights are attached to it?

Where is it in the workflow?

Those questions aren’t disappearing. They’re becoming inputs. Once the platform knows the answers, a better question emerges:

What should the platform now be able to do about it?

It can surface the right content instead of waiting for somebody to search.

It can trigger the next predictable step instead of waiting for somebody to coordinate it.

It can apply known controls before content moves.

It can send the exceptions, with context, to the people whose judgement genuinely matters.

The MAM stops being somewhere work is recorded and starts becoming somewhere work gets done.

Find your human middleware

There is a simple exercise I’d recommend to any media CTO or Content Operations leader.

Pick one commercially important video workflow and follow it from beginning to end.

Every time somebody searches, waits, transfers information, repeats a check, chases an approval or manually moves something forward, mark it.

Then ask three questions:

• Does this require judgement? Keep a person accountable.
• Is it predictable with occasional exceptions? Automate the normal path.
• Is it completely routine? Ask why you’re still paying a human to do it.

You’ll probably learn more about your AI opportunity from that exercise than from another vendor presentation featuring 37 sparkles.

Enterprise software companies have spent two decades getting very good at counting users. I think the next decade gives us something considerably more interesting to count:

the work the software actually completed.

And once work becomes the unit of value, the way we price enterprise software starts to look rather different.

Next: Why Enterprise Software Pricing Has to Move Beyond the Seat.

Review Your Video Workflow with Overcast

Take one real video workflow. We’ll map where human effort, waiting and operational friction still sit, and identify which parts your technology should already be doing for you.