The Power of AI in Transforming Insight with Fluency

In light of recent developments in artificial intelligence (AI), has of course entered the intricate world of data. As AI technologies continue to evolve at an unprecedented pace, their potential to revolutionise data processing, analysis, and utilisation has never been greater. 

As a data consultancy, Fluency is proud to have some of the world's best known brands as clients, whilst also able to offer support for SMEs and niche brands who are just starting out. Could we have built our business as fast without AI? Could we help brands build such impressive data fluency without AI? 

Matt Steel, Fluency’s marketing and business development director, sat down with David Lawton, head of data science at Fluency to discuss this. David shares his insights on the transformative role of AI in data-driven decision-making and what this means for brands looking to make use of AI in their data journey. 

Matt Steel: We utilise our AI ecosystem in some capacity for all of our clients. Can you explain your thoughts on AI in our data and insight world of AI, how it helps Fluency and what that means to the brands we work with? 

David Lawton: For me the most exciting thing right now is the idea of generalist AIs. Five years ago, when AI was starting to come on the scene, it was very specialist. It could do very specific tasks very well. So, we would engage certain providers who had really strong NLP offerings or image models, that was very much the cutting edge.

However, in recent years, we've seen AIs trained on ever bigger datasets leading to the birth of what are known as generalist AIs, more colloquially, that'll be your Chat GPTs, basically an AI that can turn its hand to anything. Whilst they've been around for a while, it’s only recently with these very large models that they have reached a level of quality that they are ready to be used in the wild. This is largely thanks to the big tech players, Facebook, OpenAI, Meta, who can sink huge resources into training with generalist AIs, now have started to overtake specialist AIs. 

So as a team, this is something that we've been looking to try and tap into because, and some of that always surprises me, a lot of these AIs are open source. So, Google, Facebook have released AIs so that anyone can download and use them. 

This has almost resulted in a David & Goliath situation, where the small tech players are going to have to make very smart innovations to be able to compete with the hardware advantage of the big players.

In my mind and from my position as Fluency it’s not desirable to compete with those brands; requiring huge investment in hardware or PhD-level talent for uncertain returns.

Where Fluency is thriving is by constantly monitoring the market, making sure we are always tapping into and leveraging the latest tech. That's how Fluency has been so successful, and can continue to stay relevant and competitive.

Our skill is to have a mixture of everything that’s going into the open-source domain, and creative thinking about how we can utilise this for the commercial benefit of our clients.

This is also looking at paid services. So, for example, Chat GPT 4, we use that to gain insights from the dashcam footage (which we analysed in bulk for an auto client) , where it is a clever old-school processing to clean up all the videos and then plug into GPT 4. That's not technically open source, but it is cutting-edge. Likewise the open source models often have very real hardware demands of their own to run, so we use AWS to enable us to access that hardware in a cost effective manner.

Fluency does not get reliant on any particular tech, because there's always the risk that tomorrow Google's going to produce a new model that supersedes it. It’s worth underlining that I’m not saying there is no longer a place for specialist AIs, we use them often, but I think we have reached a tipping point where generalist AIs more & more often will be the right tool to use, the role of specialist AIs in Fluency’s toolkit becomes more less relevant.

Matt Steel: One of Fluency’s differentiators is our stack of the world's best global data providers. Because the immense amount of data that we have in the stack can't be handled just by a small team of people. So, in order to handle that, we've built the best inbreed AI ecosystem. 

Our ecosystem is actually a relatively small number of products or brands, but within the process of building it, we've met with hundreds of products and brands to find the best ones. 

So, a client benefit is that they don't need to go out and meet with multiple brands and research all of the different products available to their business because we've done that for them. We will utilise our knowledge, our skill, and our ecosystem of AI to benefit their business.

Is it right then that our ecosystem is forever evolving? So as a new player comes on board, an existing product drops out of the ecosystem, and it just evolves, and updates itself?

David Lawton: Yes, to a degree. There are some of the partners who at the time were best in class, and were very much building their own AI solutions. We tend to find that the pace of innovation means it's difficult to keep up with the big players. Meanwhile, most of our newest partners have close relationships with the likes of OpenAI and Bard. so it’s important to find the balance between not being overly reliant on the market leaders and constantly evolving whilst mitigating risk and utilising the depth provided by the AI power houses. We work with AI partners that have the same mentality and are looking to grow as AI grows. 

Just to go back to the original point here, we're constantly meeting with AI companies and understanding exactly what they do. That is where we think “Okay, that's interesting. That's something that we could tap into”.  You then go into further meetings with them and check their credentials. We check that it really does work as there is a lot of smoke in this industry, where a lot of people say, “Oh, we can do these amazing things”, and you go and see their product and it can be quite often underwhelming. So the true gems are pretty rare. So yeah, we're constantly meeting companies to understand what everyone's doing.

Matt Steel: The other benefit, then, for brands utilising Fluency is that they don't need to employ new staff to go and learn how to physically use the products?

David Lawton: Yes, and these meetings are very resource-heavy. There tend to be at least four people on the calls, the initial ones last about an hour, and then there are follow-ups and so on. But with tens of meetings, there's a lot of man-hours that go into that process.

Matt Steel: So, using AI allows us to be lean, fast-paced, agile, and competitive to the benefit of our business, but equally to the benefit of our clients. 

David Lawton: That's half of it. Say one-half of our business is AI. It enables you to offer a scale that's just not economical with humans. For example, we recently built a tool using AI to analyse video content, this tool would have looked at hundreds of hours of video. 

To do that with a team of analysts in an objective way is basically impossible because you'd have to divide it between a team of analysts. Everyone's going to have their own human interpretations of what that video was, so it's going to become subjective instead of objective, whereas AI is 100% objective and replicable. So, you can run it to scale very easily and run hundreds of hours of videos in minutes. Then you can do a deep dive into exactly what attributes it was pulling out and you've got that very clear audit trail. This project for me highlights the new way of working for Fluency. We used AWS to pre-process around 300GB of video so it could be ingested by the AI. We then ran it through a specialist AI for further pre-processing to remove unwanted parts & then passed it through to GPT-4 for the final analysis layer. So yes, I think it's about the scalability and objectivity that AI gives you that humans can’t. 

Matt Steel: Okay, so the call-out for brands who are considering the use of AI in their businesses is don't be scared about AI and to embrace the opportunities that AI brings.

But also, speak to experts that are in the world of AI, rather than try to learn and manage it themselves, because it's so fast-paced. By the time you start researching, things have evolved. 

David Lawton: Yeah, I think that's part of it. 

First, if you don't do this, you will be left behind. The second thing is, and something that I always try and do in our work, there's a perception with AI that what is good and powerful is a bit of a black box, and people are always very keen to talk about the errors AI can make. For example, with Chat GPT, people talk about hallucinations or just plain examples of how AI is going rogue. 

The work we're trying to do with Fluency is always about trying to demystify how AI has got to where it's gone, so we always try and use AIs. There are plenty of AIs out there, which are transparent and accountable, so you can follow through the whole train of logic. Then you can do the kind of human plus machine logic of “Oh, okay, the AI is going a little bit wrong here. Humans can step in and correct it”.

So for me, it's not about giving the keys entirely to AI. I don't think we're at that point yet, but it's about enhancing human decision-makers with AI. Therefore, the AI does need to be transparent and accountable, not a black box. And that's something that we always try and deliver.

Matt Steel: Where does Fluency see AI taking us in the next three years? The revolution has happened with AI, and we are a part of the revolution, but will there be another one, or a gradual evolution? 

David Lawton: Yes, I think ChatGPT is incredible, in terms of how it brought everything to the public sphere and I don't foresee a step of that size happening in the immediate future. 

In terms of where Fluency is, we are probably going to be leveraging more AIs. I know we have built a product called Total Experience Intelligence (TXI) which is built by our experts, but powered by AI. It's a breakthrough in strategic brand planning and performance management, leveraging over 60 performance metrics to reveal how audiences truly experience a brand, it is a comprehensive, always-on, single source of truth for decision-makers.

It looks at all the attributes of a brand, unpacking a lot of AI, some of the traditional data, and bringing that into a model to understand what experience works best for a brand and trying to enable brands to make smarter decisions in a very holistic way. 

So that's one thing that we're looking to do. This is just looking into a lot more of the generative AI whilst we're using that as a data scientist/analyst helper currently. It's not something that we're building too much into our final services yet. But I would foresee that becoming a kind of more fully automated pipeline, as we have already done with a number of other large language & image models. 

So, for example, we already use Chat GPT to help write copy for our clients, expand thinking and look for the patterns otherwise hidden in plain sight. This is not lazy or derivative, just the same way using that using a computer for design work is not... The difference is we can optimise existing processes across research, insight and the creative process and create efficiencies, always coming back to a human lens such as A/B testing to see what's working. Fluency here is constantly having that fully automated revision process of refining messaging for brands.

These are use cases we will lean even more into, and are the main things that I see us doing more of in future.

Matt Steel: So to summarise, Fluency is not an AI agency, but we're a data consultancy using AI to the benefit of our business and their clients.

David Lawton: Yeah, 100%. And I would personally be very worried if we were an AI agency because we cannot compete with Meta, Google, OpenAI, and Microsoft. Unless you have a tiny team of Einsteins who could do something incredible, maybe get around that hardware gap, then you could be alright! It's about standing on the shoulders of giants. 

Here at Fluency, AI is integral to our ability to manage the immense amount of data we handle for our clients. If you are not making use of AI in your data analysis, your business is severely missing out. 

So no matter how data fuelled you are (or feel!), you operate in a world of lean, fast-paced and agile competition in a data-driven world. Fluency exist to help you win and can fill the gaps to data-fluency. Irrespective of whether you are big or small, a fresh start-up, scale-up or a household brand, you need to be ready to embrace change and be prepared to make decisions with the best data, innovative approaches and people available to you. We'd love that to be with us!