AI is the new black. Regardless of their industry, business leaders seem to understand that AI will impact their future in a big way - much like electricity, the internet and mobile have, in the past. This mindset also falls in line with the shift to a Tech@Core approach that we’ve found to be characteristic of the ongoing Fourth Industrial Revolution.
If you intend to stay in the game, your business needs to adapt to new ways of thinking and working. You’ve got to leverage technology in strategic decision-making to be able to discover new business opportunities.
Based on the myriad conversations we’ve had with our clients and market research that’s available in abundance, we hypothesize that businesses are finding it hard to identify a suitable starting point for AI implementation.
As is the case with any other technology, the objective is to discover opportunities that can have an enterprise-wide impact. But most business leaders appear to be overwhelmed since they lack the mental model to understand how they can derive value.
They’re delegating the discovery process to data scientists, who are certainly not the best people to determine business priorities. If you’re wondering where you can start your AI journey, we propose the following approaches. Also, this is assuming you have no prior knowledge of AI.
Andrew Ng, the founding lead of Google Brain says, “If a typical person can do a mental task with less than one second of thought, we can automate it using AI.”
Some establishments have already started acting on the sentiment and are using AI-enabled bots to resolve customer queries. For example, you have bots deciding whether or not a purchase should be refunded, or a mortgage should be approved. More examples that fit the one-second heuristic include examining security videos to detect suspicious behavior or determining whether a car is about to hit a pedestrian and deleting abusive online posts.
Our recommendation is to dive into the systems and processes within your enterprise to identify one-second tasks that are performed repetitively. And, then ask your data/AI teams if these tasks can be automated.
AI’s ability to influence decision-making in real-time paints it as an extremely valuable tech investment, and there’s more to it than automation. It can also take on the role of providing assistance and enabling augmentation.
For example, in both brick-and-mortar and online stores, AI is seen taking on the role of shop assistants, helping buyers find items and make purchase decisions - Walmart’s Bossa Nova, Sephora’s Color IQ, Amazon’s Online Recommendations engine are a few examples that come to mind.
Interestingly though, AI can assume different roles in different contexts within the same industry. For instance, in healthcare, it augments diagnoses and surgery to help the doctor make decisions. On the other hand, AI-enabled chatbots and virtual platforms assist patients with information gathering.
Understanding whether a said task relies on cognitive decision-making or data-driven models can ease the discovery process. In other words, figuring out whether you want AI to support, enhance or replace human effort is a useful starting point when evaluating a possible use case.
At Thoughtworks, we’ve partnered with a manufacturing company to build an AI-enabled application that can predict repair and maintenance schedules for heavy equipment. The application does all that while also accommodating advancements or postponements in service calls. It also provides easy access to critical information such as customer and equipment details that could impact sales opportunities for dealers.
Another example of a high-value prediction problem can be seen in retail, where AI is helping brands predict an optimum pricing that isn’t too high or too low, while still ensuring overall revenue. In banking and finance, AI-based machine learning can now detect potentially fraudulent payments, and even prevent them.
Using AI to anticipate problems related to buyer preference, staffing or supply-chain management, and designing relevant solutions can help you extract greater value in the long-term, owing to the cumulative effect of making such predictions.
Recognition – to identify and recognize things within unstructured data
Conversation and Human Interaction – to enable machines to interact with humans more naturally
Predictive Analytics and Decision Support – to help make better decisions
Goal-driven Systems – to find the optimal solution to a problem
Autonomous Systems – to minimize labour
Patterns and Anomalies – to find similarities and dissimilarities
Hyper-personalization – to treat the user as an individual
These patterns are evident in products that are looking to integrate object, text, image, voice, face or gesture recognition. They can also be observed in the ever increasing number of open source, cloud-based AI/ML platforms and tools - much like those offered by Google, Amazon and Mozilla, which encourage developers to build intelligent software applications, allowing businesses to implement them without necessarily hiring AI experts.
The availability of these resources has already enabled some of the AI applications we’re now seeing in healthcare, finance, security, fraud, intellectual property, retail, and consumer electronics. The good news is that this kind of AI eases the load on your development team.
For instance, if you’re a food delivery service, AI can help improve user experience, by recommending meals based on purchase history - clearly a customer-facing use case. On the other hand, it could also help forecast the number of delivery personnel required during a specific period. Deriving Value Requires You to Define ‘Value’
The key to exacting AI’s full potential lies in identifying the most suitable role(s) within your use case. The five approaches discussed above may overlap depending on the nature of your business. Different approaches will work for different types of businesses, it’s imperative that you define what ‘value’ means for your business so you can leverage AI to get there.
The next step will be to work with your data team to define the problems more concretely. It’ll help to make an assessment of benefits, feasibility (e.g. data availability) of proposed solutions beforehand, and taking a cross-functional, collaborative approach to it will ensure a more accurate scoping of the problem statement, enabling meaningful design and effective implementation.
A version of this article appeared in PCQuest.
If you intend to stay in the game, your business needs to adapt to new ways of thinking and working. You’ve got to leverage technology in strategic decision-making to be able to discover new business opportunities.
Based on the myriad conversations we’ve had with our clients and market research that’s available in abundance, we hypothesize that businesses are finding it hard to identify a suitable starting point for AI implementation.
Finding the start line of your AI journey
As is the case with any other technology, the objective is to discover opportunities that can have an enterprise-wide impact. But most business leaders appear to be overwhelmed since they lack the mental model to understand how they can derive value.They’re delegating the discovery process to data scientists, who are certainly not the best people to determine business priorities. If you’re wondering where you can start your AI journey, we propose the following approaches. Also, this is assuming you have no prior knowledge of AI.
1. Automation and Andrew Ng’s one-second heuristic
Full-blown automation completely removes the need for human involvement in any process – self-driven cars are perhaps what come to mind first, but other use cases of automation are seen in manufacturing, banking and cybersecurity.Andrew Ng, the founding lead of Google Brain says, “If a typical person can do a mental task with less than one second of thought, we can automate it using AI.”
Some establishments have already started acting on the sentiment and are using AI-enabled bots to resolve customer queries. For example, you have bots deciding whether or not a purchase should be refunded, or a mortgage should be approved. More examples that fit the one-second heuristic include examining security videos to detect suspicious behavior or determining whether a car is about to hit a pedestrian and deleting abusive online posts.
Our recommendation is to dive into the systems and processes within your enterprise to identify one-second tasks that are performed repetitively. And, then ask your data/AI teams if these tasks can be automated.
2. Assistance and/or Augmentation with AI
AI’s ability to influence decision-making in real-time paints it as an extremely valuable tech investment, and there’s more to it than automation. It can also take on the role of providing assistance and enabling augmentation.For example, in both brick-and-mortar and online stores, AI is seen taking on the role of shop assistants, helping buyers find items and make purchase decisions - Walmart’s Bossa Nova, Sephora’s Color IQ, Amazon’s Online Recommendations engine are a few examples that come to mind.
Interestingly though, AI can assume different roles in different contexts within the same industry. For instance, in healthcare, it augments diagnoses and surgery to help the doctor make decisions. On the other hand, AI-enabled chatbots and virtual platforms assist patients with information gathering.
Understanding whether a said task relies on cognitive decision-making or data-driven models can ease the discovery process. In other words, figuring out whether you want AI to support, enhance or replace human effort is a useful starting point when evaluating a possible use case.
3. AI for high-value predictions
Machine Learning, Deep Learning, Natural Language Processing and Automation have already made a huge impact on marketing, sales, customer service, employee engagement, recruitment and more.At Thoughtworks, we’ve partnered with a manufacturing company to build an AI-enabled application that can predict repair and maintenance schedules for heavy equipment. The application does all that while also accommodating advancements or postponements in service calls. It also provides easy access to critical information such as customer and equipment details that could impact sales opportunities for dealers.
Another example of a high-value prediction problem can be seen in retail, where AI is helping brands predict an optimum pricing that isn’t too high or too low, while still ensuring overall revenue. In banking and finance, AI-based machine learning can now detect potentially fraudulent payments, and even prevent them.
Using AI to anticipate problems related to buyer preference, staffing or supply-chain management, and designing relevant solutions can help you extract greater value in the long-term, owing to the cumulative effect of making such predictions.
4. Out-of-the-box AI
In the short time that AI has emerged to the fore of tech discussions, we’ve seen that every AI application fits these seven patterns that repeat themselves in different combinations.Recognition – to identify and recognize things within unstructured data
Conversation and Human Interaction – to enable machines to interact with humans more naturally
Predictive Analytics and Decision Support – to help make better decisions
Goal-driven Systems – to find the optimal solution to a problem
Autonomous Systems – to minimize labour
Patterns and Anomalies – to find similarities and dissimilarities
Hyper-personalization – to treat the user as an individual
These patterns are evident in products that are looking to integrate object, text, image, voice, face or gesture recognition. They can also be observed in the ever increasing number of open source, cloud-based AI/ML platforms and tools - much like those offered by Google, Amazon and Mozilla, which encourage developers to build intelligent software applications, allowing businesses to implement them without necessarily hiring AI experts.
The availability of these resources has already enabled some of the AI applications we’re now seeing in healthcare, finance, security, fraud, intellectual property, retail, and consumer electronics. The good news is that this kind of AI eases the load on your development team.
5. Customer-facing applications vs. Internal Processes
Another way to discover AI use cases is to simply look at all your internal and customer-facing processes, and determine which is in more urgent need of a tech facelift.For instance, if you’re a food delivery service, AI can help improve user experience, by recommending meals based on purchase history - clearly a customer-facing use case. On the other hand, it could also help forecast the number of delivery personnel required during a specific period. Deriving Value Requires You to Define ‘Value’
The key to exacting AI’s full potential lies in identifying the most suitable role(s) within your use case. The five approaches discussed above may overlap depending on the nature of your business. Different approaches will work for different types of businesses, it’s imperative that you define what ‘value’ means for your business so you can leverage AI to get there.
The next step will be to work with your data team to define the problems more concretely. It’ll help to make an assessment of benefits, feasibility (e.g. data availability) of proposed solutions beforehand, and taking a cross-functional, collaborative approach to it will ensure a more accurate scoping of the problem statement, enabling meaningful design and effective implementation.
A version of this article appeared in PCQuest.
Disclaimer: The statements and opinions expressed in this article are those of the author(s) and do not necessarily reflect the positions of Thoughtworks.
