There is an incredible amount of optimism around Artificial Intelligence (AI) but there is also a lot of confusion about how you can benefit from it. This article explains the different categories of AI and how they add value to businesses and how you can begin to use them.
In 2017, Microsoft said that all their products and services would be infused with AI. The products and services enhanced this way will be the widespread way businesses will experience and use AI. This infusion of AI within products and services is providing new capabilities for users to exploit by improving working practices and results. Simply put, you can expect to be able to benefit from AI in many of the products and services you may be already using.
One example of this is the intelligence that is infused in automation tools. This has given rise to Intelligent Automation which incorporates AI to help perform a set of discrete, narrowly defined tasks that are incorporated into updated business processes.
Understanding AI for business
AI has been successfully applied to single tasks and has brought us many technological breakthroughs. This focus on single tasks is referred to as Narrow AI. Being able to unlock your phone with your face, auto tagging people in your social media photos and routing deliveries to your home are examples of where advances in Narrow AI algorithms, data science and computing power are delivering value.
Artificial General Intelligence (AGI) is the ability to exhibit human intelligence. This type of AI is exciting and somewhat scary. Many people hold the misconception that all AI is AGI. I believe AGI will likely emerge as a collection of advancements in Narrow AI. AGI is not here yet and is unlikely to be here soon.
Narrow AI surrounds us. It powers better information searches, identifies things like hate speech and recognises your voice in a group conversation. It will be the AI that your business will be able to exploit but there are challenges.
Challenges exploiting AI
Much of Narrow AI is built around the concept of building and training Machine Learning (ML) models to make predictions. These models are the output of combining an algorithm with data. For example, a combination of retina scan data and an algorithm to predict glaucoma could be used by software products to aid early detection of eye disease. In commoditised AI solutions and services, the hard work of building and training models has been done for you. Unfortunately, the perceived ease of plug and play solutions gives a false impression that AI is easy and can generate immediate returns on investment. There are challenges incorporating AI into your business. The main ones are:
- Code integration: How to integrate the AI services via their APIs into your existing systems and processes e.g., Azure Cognitive Services or AWS AI services.
- Product integration: How to integrate several task specific AI infused products into your infrastructure without creating a maintenance burden and creating technical debt.
- Data governance: How to share, process and control data across multiple services, products and teams.
- Inherent bias: If your machine learning is based on skewed or incomplete data then predicted outcomes could be questionable.
To maximise the business value with AI beyond using commodity services requires hands-on experience with Machine Learning (ML). Experience with technologies such as Python or R is considered essential. Getting to know different frameworks and APIs such as scikit-learn or Tensor Flow is necessary and can be overwhelming. Unlike the turn-key AI products, the data wrangling required is technical in nature and requires news skills to be learned, developed and retained.
Applying Machine Learning algorithms rarely finds perfect relationships within data. This means that people need to understand what the ML algorithms do, the results of applying them, the adjustments to make to improve them and importantly knowing when to stop trying to perfect them. Many machine learning experiments involve an extensive “confidence building” phase to confirm, refine or adjust algorithms. The value delivered is to remove the uncertainty from data analysis and decision making.
Establishing the right foundations
Triad frequently helps clients articulate and validate their IT strategy and transformation plans. For AI we use the Machine Learning Canvas developed by Louis Dorard as the basis for capturing ML tasks. Like the Business Model Canvas, it helps ask the right questions and validate the value of that AI/ML will bring.
Our Pathfinder review process fits well with helping clients with their IT planning. This is an extended form of discovery that gives clients a structured way of testing ideas while aiming to create value. The results can validate what to do and importantly what not to do.
Building an AI team
To avoid falling short of your goals with your adoption of AI you need to build an AI team and adapt ways of working. While technical experience is needed you also must consider what the cultural impact will be on implementing AI will be and how ways of working must adapt.
Building an AI team has parallels with the software development methods. Cross-functional teams with a mix of skills and perspectives will deliver the best results. Business and IT working alongside data experts will lead to innovative solutions and increase the chance of adoption throughout an organisation. Building and retaining data expertise is required to exploit ML. This requires a good understanding of your data and a good understanding of the type of ML algorithms you will be applying. It requires an environment that encourages experimentation. It also requires people, time and budget.
The goal of building an AI team is to support the business. The outcomes can be transformative. Retail can benefit from personalising shopping experiences and loyalty programs for each customer rather than a broad demographic. Stock management can optimise inventory management. Dynamic pricing can be applied to products and services to grow sales. Law firms can enhance how legal research is performed. Renewable energy companies can better predict energy output. Everyone can benefit from enhanced accuracy and efficient processes freeing people from repetitive tasks.
How Triad can help you with your AI aspirations
Triad can help you plan how best to exploit AI for your business. Our 30 years of experience in systems integration and development is directly relevant for talking the next steps and solving the code integration and product integration challenges. We can also help you build an AI capability and team to share, process and control your data.