Join us for some insight into a key discussion we had internally, with our Head of Innovation, Matt Thomas.
Every agency is suddenly launching AI maturity tests, they are becoming “AI-powered”, “AI-native” or “AI-first” overnight. Buzzwords are everywhere, and for clients trying to understand what any of it actually means, it can become difficult to separate genuine innovation from noise.
During an internal discussion, Matt said quite accurately: “Everyone’s labelling AI onto absolutely everything.”
Most businesses now feel pressure to demonstrate they are using AI in some capacity, and this is exactly the problem of labelling themselves off a trending topic, without fully understanding it at a deeper level. Agencies are especially worried that if they don’t look like they’re adopting AI, clients will assume they’ve been left behind. However, at the core of it, using AI well is much harder than simply using AI.
Human expertise is still the most important part
One of, if not the most, important factors when using AI is the importance of maintaining a human in the loop. Again, another buzz term however, AI doesn’t know when it’s wrong. It can generate code, content, layouts, summaries and recommendations incredibly quickly. However, it can also confidently produce inaccurate information, generate flawed code, or create something that appears polished but is fundamentally incorrect.
AI responses can come across as confident, so much so that you believe it’s true. It can write code that looks right - but it can quite quietly break something catastrophic, and create content that is factually incorrect. This is why Matt stands behind the fact that human judgement remains essential.
The role of AI should not be to replace expertise. A person will catch the inaccuracies; the developer who has generated code with AI before it ships, a designer who checks an AI generated layout, a strategist that reads what AI has drafted and questions whether it is true. Without that person, you’re essentially gambling. Adding a human is not to be considered a weakness, it’s the whole point. Matt is passionate that AI does the heavy lifting, but the human does the judgement.
Ultimately it comes down to what that gamble is worth when compared to what the risk could be to reputation.
The difference between 'trust us' and proof
One of the biggest issues in the current AI landscape is visibility, and so if a company says they are using AI responsibly, how do you actually know? Matt’s view is that you are simply taking their word for it. But that lack of accountability is where risk begins to creep in because many companies are still implementing tools without proper governance, documentation or oversight. This really matters when AI is woven into client work, strategy, development as well as products.
Clients are trusting agencies with sensitive information, intellectual property, customer experiences and commercial decision-making. Without safeguards, there’s often no visibility into which AI tools are being used and how. There is no evidence of how client data is being handled and processed, whether humans are involved at all and therefore where accountability lies should something go wrong. That’s why governance matters.
As Matt explained:
"Trusting without proof is a massive oversight, because when working with an agency for example, the client is trusting that the agency has thought about how AI is used on their project, trusting the output, trusting that there is a record of what AI did in case something goes wrong. But what is the proof? Are data models being trained on the client’s sensitive data? Has any documented and proven auditable trail actually been taken? You’re trusting this is in place. This is a lot of trust for any brand to take on."
Faster doesn’t always mean better
For Rawnet, and other agencies alike, automation and AI-assisted workflows are helping speed up repetitive tasks, reduce friction and create more time for higher-value thinking. However, when clients expect delivery to be faster, the risk is that bad code, more bugs, and short term thinking start to creep in which has a higher risk of reputational damage. This is where accountability gets blurry - something that shifts fast is great for initial profit lines, but who is ultimately responsible… The agency? The AI vendor? Or the company themselves? The worry is that companies will start to lose sight of what is important, which is credibility. It’s the cost of what something will have in the future, not now. Companies, as well as agencies, that start to optimise purely for output volume or reduced cost, sacrifice maintainability, strategy and quality in the process.
The risks of excluding human expertise
We know there are well known examples of companies experiencing the consequences of poorly governed AI systems.
Poorly trained, maintained and governed chatbots are a prime example where companies have tried to replace humans but have ended up promising policies that don’t exist, have used explicit language to customers and under-selling products. All of these have hit companies reputationally as well as their bottom line on a significant scale. These examples are extreme, however they highlight a wider issue.
AI is not responsible for its own actions, the responsibility always falls back on the company owning it.
So when governance is missing, problems escalate quickly, whether it be financially, operationally and/or reputationally. However, where does accountability fall to?
Why governance matters to us
Our philosophy is to approach AI responsibly from the beginning and not just because it sounds impressive or because it’s trendy. It is because clients deserve confidence in how their work, data and intellectual property are being handled.
This is more than the middle ground, we have the ability to move fast with governance. Agencies can move fast with AI, of course they can. But if they don’t have governance in place, it’s just a trust exercise.
We are proud to be augmenting our workflows, with a human in the middle. The focus is not on shortcuts. We’re spending a lot of time making sure we’re doing it right to protect our clients’ IP and making sure what we deliver is sustainable and scalable for the future. Not taking shortcuts straight away just to deliver quicker and cheaper, but to actually make things a lot better. When used properly, AI gives teams more time to focus on strategic thinking, creative problem solving and innovation.
The boring parts get faster, which means the interesting parts get more attention.
Being able to move fast, with governance, takes team work. It is more than just working together, it’s how Rawnet is underpinned and an ethos that is threaded throughout everyone. It’s more than a nice company behaviour written on the wall and it is demonstrated by the quality as well as the speed of how we pivot and deliver, along with compliance underpinning everything we do.
