When it comes to embracing Artificial Intelligence (AI), there’s a pretty broad spectrum of business approaches. Some remain sceptical or have yet to see the relevance it holds for their sector or organisation. Others are exploring the wealth of opportunities it presents, eager to gain an advantage and prepare their teams for the future.
Within this gamut, the percentage of businesses that have taken a concept from idea to implementation is still relatively low (30% according to Gartner, Inc.). This means that examples of purposeful AI integration that offer real value are limited, and potential applications remain unrealised. As a result, visualising practical uses for the technology can be difficult. Delivering a competent, cost-effective product is even harder. However, with the right people and processes, it’s certainly achievable. What’s more, the results can be transformative, as our latest AI launch for Tier 1 construction company Willmott Dixon goes to show.
Through AI, we’ve helped Willmott Dixon optimise its information repository The Knowledge Hub and gain a competitive advantage in bid writing. Delivering time savings of 28+ hours per average bid, estimated cost savings of £147,924 per year, and creating space for innovation, the tool has enhanced its new business outcomes and the working lives of those behind them.
Taking this proven experience in the execution of AI products, at Tribus we aim to make the technology more accessible and applicable for our partners. In this article, our Creative Director, Nathan Parks, shares his experience and advice on taking an AI idea from concept to creation.
So, Nathan, how do you deliver an AI product?
As you can imagine, this is a hard subject to summarise as there are so many different outcomes, intricacies, and opportunities involved. Clearly, we’ve only just scratched the surface of what AI can achieve, and the technology is evolving rapidly. So, there’s no one-size-fits-all approach. There is, however, an adaptive framework for discovery that we have honed over time to ensure we progress in the right direction.
Start with a strong foundation
So many organisations want to use AI but have yet to define why they need to use it. This positions AI as a problem to solve, rather than recognising it as a potential solution. But if you want to build a successful AI application, you need to start by defining the business problem or opportunity you want to address first, and then you can find a use case for it.
Data quality: Do you have good-quality project data to support an AI product? For example, a digital database that is free of duplicate data or blank fields. Could AI be used to improve your understanding or access to this data? You may have come across the acronym GIGO before (‘garbage in, garbage out’), and this sentiment still stands. The more robust and relevant your data set is, the better your outcomes will be.
Technology: Does the approach require AI specifically? Or, are you looking to achieve something that software or a website with well-designed navigation could fulfil?
Cost-benefit: Will the efficiencies or opportunities that your product delivers outweigh the processing power it requires and the ongoing costs associated with this? A responsible supplier can help you estimate the ongoing costs you may face, given the visits/usage you expect to see each month.
Human involvement: AI often works best when combined with a human’s expertise, perspective, or oversight. Can your business resource this need? Can you strike the right balance?
If your idea still stands after questioning, it’s time to take it forward.
Follow a flexible framework
Every AI specialist will have their own take on how to deliver an artificial intelligence product, and these approaches will evolve at speed alongside the technology. At Tribus, we use a flexible framework that includes four core areas of delivery. Refined through our work on projects such as The Knowledge Hub AI tool, they combine the structure needed to work efficiently with the flexibility required for true exploration.
Discovery: This crucial step involves challenging the brief and delving into the details behind it so that we can uncover any hidden barriers or opportunities for improvement. From workshops and site visits to data audits, we’ll conduct activities that allow us to gain an in-depth understanding of your business and how the product will provide the most value.
Design: Once we have clear deliverables and our direction in place, we begin to craft the experiences users will have when using the AI product. This includes the look and feel of the user interface as well as the journey they take to achieve their desired outcome. We can also create proof-of-concept prototypes at this point to test our approach and ensure the data quality is capable of producing the required outcomes.
Development: During the development and engineering stages, we transform the designs into a fully functional product. This means building any models, algorithms or systems that allow the AI tool to perform. It can also include managing data pipelines, training models and integrating these models into other applications.
Deployment: Throughout the process, we can provide the consultancy and content necessary to launch your product successfully and secure ongoing support. This may mean training a technical team to champion AI internally, or producing explainer animations for external use. It’s all about supporting the transition to new ways of working.
Our approach to AI is based on our existing digital development process, but factors in the new opportunities and outcomes we’ve experienced when utilising AI. It’s also delivered with more flexibility to accommodate the wealth of possibilities that AI offers.
Let’s take the design process as a practical example of how our approach changes. When designing a more traditional digital product, such as a website, we use logic. This means the user journey can be plotted out following a decision tree format to arrive at one of many expected fixed outcomes that we can design a user interface (UI) for.
AI applications, however, are less predictable. Depending on how it’s applied, the technology can independently identify cases, find patterns or generate content that we can’t possibly foresee. Of course, this is where it adds value, where its power and potential lie. But because we can’t always predict what the output will be, designing a rigid user interface can be more difficult. In this instance, we may need to enter into the development stages before any designs are finalised. This allows us to see the outputs of the AI tool and how it’s interpreting our requirements, before feeding this insight back into the design and the UI for it.
Prepare for the challenges of AI
As we’ve touched on above, the factors that make AI unique and powerful can also present challenges that you may not have experienced with traditional digital products. However, working with an experienced team, it is possible to prepare for some of these differences in advance. By having certain measures in place and communicating them with the wider business, you’re more likely to secure their confidence and support for the project. It also helps you to keep within your budgets and timescales.
Let’s take the testing stages as an example of how your approach may change. AI products are open books in terms of how they can be used (way more so than a normal product). You need to treat them more like black boxes to be guided, rather than a product that is coded in a very specific way. Because of this, testing is way more crucial with AI than with other digital products. It also becomes far more important to involve your end users in this stage of development. AIs use companies' datasets; that's what they are experts in, but you need the business’s input and knowledge to check that these have been interpreted accurately and that the information they provide you with is relevant and contains the correct, expected content.
Keep communications open
On many of our AI builds, including the Willmott Dixon Knowledge Hub AI tool, the most valuable insights and instructions we receive result from honest discussions with our project partners.
This approach to constructive communication begins in our discovery workshops, where we align everybody's thinking on requirements, test assumptions and glean expectations. At this stage, talking openly without intent or prejudice, we can iron out a lot of misunderstandings and cement a shared understanding of what’s possible. We may even uncover some hidden potential for the project, too.
Once the design and delivery are underway, we keep conversations open by holding regular catch-ups. Here, there is an acceptance from both sides that change is likely to happen, and we use our time together to identify where it needs to happen in the interest of time, budget, and bettering the product.
One additional step we introduced on the Knowledge Hub was sharing our initial, light-touch concepts with the client at an early stage. This allowed us to explore our ideas as proof of concepts, get valuable feedback to shape the build, and gain more traction and momentum for the product internally.
As we’ve shown in this article, Artificial Intelligence holds a wealth of undiscovered possibilities. As such, it’s impossible to outline one definitive route to producing an AI product. At this stage of the technology’s adoption, the majority of work is still investigative and exploratory. No one can claim to have conquered every application or outcome. However, with a combination of the above tactics in place and an experienced partner on your side, your journey is sure to be productive. If you can keep an open mind, focus on the bigger picture, and be willing to explore the multiple solutions on offer, you’ll be sure to reap the rewards.
If you’d like to learn more about AI tools or Tribus’s approach to development, please get in touch.
