How AirVu Media is using Productive AI to rethink agency operations

When it comes to AI, I’ve always believed in starting with one question: What problem are we actually trying to solve?

There’s a tendency to jump into AI because it’s exciting. You build something, get 70% of the way there in no time, and think it’s incredible. But getting that last 20–30% right? That’s where the real work begins.

At AirVu Media, we do a lot with AI, both within Productive and across the products and services we’re developing ourselves. But our approach has always been deliberate. Rather than building hundreds of agents or introducing AI into every possible workflow, we’re focused on a smaller number of solutions that genuinely change how we operate.

And some of the biggest wins we’ve seen have nothing to do with generating content or images. They’re about something much less glamorous, but far more valuable: running the agency better.

As an agency owner, I need to understand a handful of things at any given moment. Are we bringing in new business? Are we converting opportunities? Are we delivering the work? And are we actually getting paid?

I’ve implemented a framework called ACDC, which stands for Attract, Convert, Deliver, and Collect. Those four stages tell me pretty much everything I need to know about how we’re performing.

Before AI, getting that complete picture took quite a bit of work.

We’ve always loved Productive because all our agency data lives in one place. We use it for our CRM, deal pipeline, project management, finance, and pretty much everything else. But even with all that information available, I still needed to figure out which reports to build, which filters to apply, and how to connect everything to get the answers I needed.

Now, I’ve built a Productive AI Artifact that brings my entire ACDC framework together in one place.

What would’ve taken 10 or 15 different reports is now one Artifact that tells me everything I need to know about how we’re doing as an agency.

I can open it every morning, look at the week ahead, and immediately understand where we stand. If we’re falling behind on collecting payments, I know we need to chase invoices. If we haven’t been sending enough proposals, I know we need to focus on new business.

I don’t have to spend time finding the information anymore. I can spend that time doing something about it.

And the time savings have been massive.

I used to block off every Monday afternoon just to go through our numbers, understand what the data was telling us, and work out where we needed to focus.

That entire afternoon is now back on my calendar every single week because of one Artifact.

It took plenty of iterations to get it right, but that’s the thing about AI. You invest time upfront to build something genuinely useful, and the value compounds from there.

We organize our marketing department into five smaller teams, or pods. Understanding how each pod is performing is essential, particularly for revenue, client retention, and forecasting.

Previously, much of this work was done in Excel. Every month, I’d spend a couple of hours gathering data points, running calculations, updating spreadsheets, and sharing the results with the team. We’d even tried building the reports we needed in Productive before, but couldn’t quite get them into the format we wanted.

With AI Artifacts, we’ve been able to build exactly the view we need.
I can now see how each pod is performing, compare revenue and retention, and look at forecasts based on recurring client contracts.

For example, if a pod’s projected revenue is about to drop because several client contracts are up for renewal, I can see that in the forecast and understand what’s driving it. Previously, I’d have to piece that information together manually. Now it’s all there.

Beyond the hours saved, there’s another important benefit: I’m more confident in the data.

When you’re manually moving numbers between systems and spreadsheets, mistakes can happen. Having the information directly connected to our operational data removes much of that risk.

It’s been game-changing for understanding how our teams are performing and where we need to pay attention.

One of the biggest opportunities we’ve identified for AI is improving how work moves through the agency.

We handled around 12,500 tickets last year. That’s thousands of individual briefs, requests, and tasks moving between different people and departments. And when you’re working at that volume, even small inefficiencies add up.

Take task briefing, for example.

Previously, we had templates that people were supposed to follow when creating briefs for our design and content teams. But inevitably, information would be missing, fields would be skipped, or the person writing the brief wouldn’t know exactly what the receiving team needed.

That meant going back and forth, asking questions, clarifying requirements, and spending time on administration instead of actually doing the work.

So we built an AI agent called Leonard.

Leonard understands our standard operating procedures, which are stored in Notion, and uses them to ensure tasks are briefed correctly. The important thing is that Leonard doesn’t just follow a generic template. It understands the requirements of the team receiving the work. Someone creating a design brief shouldn’t necessarily need to know every detail of the design team’s SOPs. Leonard does, and it helps make sure those requirements are covered.

The result is more consistent, structured briefs and far less unnecessary back-and-forth. The biggest unlock with AI isn’t generating images or content. It’s streamlining operations, so your team can get on with doing really good work.

We’ve built a couple of other agents around the same idea.

Terry helps format tasks so information is consistently structured and easy to read.

Barry acts as a brief decoder, helping us understand incoming requests and what’s actually being asked.

None of these agents is particularly flashy. But that’s exactly the point. We’re not trying to build AI for the sake of having AI. We’re addressing the small, repetitive problems that create friction across thousands of tasks.

And when you solve those problems at scale, the impact becomes substantial.

One of the challenges we’d identified was the gap between winning new business and actually getting started on the work.

Sales would close a project, but there could be a week or two before everything was properly resourced and ready to go.

We realized we needed to start thinking about delivery much earlier in the sales process.

Now, once a deal reaches about 80% certainty, we’ve set up triggers that initiate the resourcing process. Our delivery team can start by reviewing who’s available, who has capacity, and how we should allocate the work before the deal is officially closed. It means we can be much more proactive instead of waiting until something lands on our doorstep.

The other big change is how easily we can access information through the Productive AI Assistant. When we were first involved in the beta for Productive’s AI agents, I remember saying that my dream was to be able to talk to my agency. I wanted to be in a meeting, discussing a new project or brainstorming something, and simply ask a question about what’s happening in the business.

That’s something we can actually do now.

If we’re discussing a new project and wondering who we might assign to it, I can ask Productive. I don’t have to stop the conversation, open several reports, and start digging around for the information.

That ability to ask questions about the agency, in the moment, is incredibly powerful.

We’re not approaching AI with the expectation that one tool should do absolutely everything.

We use Claude extensively and have connected it to Productive through MCP. That allows us to bring operational data from Productive together with information from other systems, including our accounting software.

I get daily briefings that help me understand where I should focus based on what’s happening across the business. But if I’m simply asking a question about information that already lives in Productive, why would I go somewhere else to do that?

That’s why we use both.

Productive AI is incredibly valuable for understanding and acting on our agency operations. Claude helps when we need to combine that information with a much broader dataset.

We’ve also built our own AI workforce for SEO and GEO, comprising 26 specialized agents working across research, strategy, and content creation. We’re currently producing more than 100 pieces of content a month across our clients using this approach. We’ve even developed a way for those agents to work within Productive tasks, hand work to one another, and eventually pass it over to a human for review.

For me, what’s interesting isn’t just having agents doing work. It’s building a system where people and AI agents can collaborate within the workflows the team already uses.

That’s where I think things get really exciting.

We’ve got an entire roadmap for where we want to take AI, and we’re currently in what I’d call phase two.

Phase one was about learning. We wanted to understand what AI was genuinely good at, where it could add value, and which solutions were worth investing time into. The important part wasn’t just implementing tools. It was getting people to actually use them.

Adoption is everything. You can build the most sophisticated AI agent in the world, but if it doesn’t make someone’s day easier, they’re probably not going to use it. That’s why something like Leonard has been such a success for us. It solves a problem that affects almost everybody, and the benefit is immediately obvious.

Phase two is about operations. We’re looking at how we can continue to improve the day-to-day operations of the agency, reduce administrative work, and streamline our processes. One of the things we’re working on is connecting our content production process directly to publishing.

The idea is that once a piece of content has been reviewed and approved in Productive, changing the task status could trigger its publication to the client’s WordPress website. It may sound like a small improvement, but when you’re publishing more than 100 pieces of content across multiple client websites every month, the time savings can be enormous.

Looking further ahead, we’re continuing to build specialized AI workforces for our marketing services. 

The goal isn’t to replace our people. It’s about combining the strengths of AI agents and humans so our teams can spend less time managing work and more time delivering great results for clients.

Without a doubt, it’s worth it.

AI is already having a profound impact on how we run AirVu Media, and I think we’re still at the beginning of what’s possible. But my advice would be to start with the problem, not the technology.

Don’t build 100 agents just because you can. Find the processes slowing your team down, understand why they’re inefficient, and build a small set of solutions that genuinely improve how people work.

For agencies, especially as they grow, understanding operations is everything. Utilization, delivery metrics, resourcing, profitability — these are the things that determine how effectively you can run a business.

Productive gives us the operational foundation to understand those things. And with AI, we can do so much more with that information. For me, the real promise of AI isn’t simply doing more work faster. It’s giving people the time, information, and freedom to focus on the work that actually matters.

And that’s exactly what we’re building toward at AirVu Media.