A tale of Human Centered AI
Lessons learnt and Best practices for the design of AI services
We all will remember the year 2020 for how Covid-19 affected every aspects of our lives. From traveling, to working, the pandemic has changed almost everything, including the way we feed ourselves. Indeed, due to social distance restrictions and lockdowns affecting most of our countries, even cooking at home became one of the major trends recorded in the last months. If this may have endangered the families of less expert cooks, it certainly did not frighten me or my wife, because we enjoy cooking and even more, I decided to start brewing my beer at home to accompany our meals.
What are the key elements for starting up your home brewery?
- Training with the basic processes: malting, mashing, boiling and fermenting at a controlled temperature.
- Setting up the technology stack: getting good ingredients and the home-brewing machine play a key-role.
- Initiating the production: after some time spent on self-education and preliminary experiments with ingredients, the machine and glass bottles, it is possible to launch the home brewing activity.
As it often happens, I realised that crafting my own beer is not that easy, and while meditating with the glass of my first ale in the hands, I realised how the brewing process is similar to the design of AI-based services.
Like “beer crafting”, designing AI-driven services requires clear goals and many trails & errors before reaching a desired result and the right tools. These kinds of projects are complex despite the fact that in many cases AI-based projects are managed in the same way as developing a web app.
How to avoid Failures when Designing AI Projects
The world around us is changing faster than ever before. It is changing in many areas like technology (AI, quantum computing, blockchain and IoT), economic (COVID-19) as well as environment (climate change). This fast change is a significant challenge for everyone. Especially in R&D departments, we strive to bring innovation, entrepreneurial spirit adopting a lean approach and so on, but the results are not so exciting. Moreover, the lean approach suggests to test as soon as possible the hypothesis and if they are not confirmed, trash it — fail fast, fail often — . Things behave differently when we handle AI-based services.
AI for Everyone
According to the “Gartner’s Hype Cycle for AI, 2020” main findings are:
- 47% of artificial intelligence (AI) investments were unchanged since the start of the pandemic and 30% of organizations plan to increase their AI investments, according to a recent Gartner poll.
- 30% of CEOs have AI initiatives in their organizations and regularly redefine resources, reporting structures and systems to ensure success.
- AI projects continue to accelerate this year in healthcare, bioscience, manufacturing, financial services and supply chain sectors despite greater economic & social uncertainty.
This trends is confirmed also by the Futurescape 2021 of IDC, that foresees that by 2022, 65% of CIOs will digitally empower and enable front line workers with data, AI, and security to extend their productivity, adaptability, and decision making in the face of rapid changes.
According to Worldwide Spending on Artificial Intelligence of IDC forecasts for AI spending a total market value for 2020 about $50.1 billion. Huge amount of money!
Lessons Learnt and Best Practices
Multiple factors come into the game when implementing AI projects. Data, skills, domain understanding, customer relationship and many others affect the success of AI projects.
Based on my experience, these are some common situations.
Business misalignment. At the beginning AI-driven technologies are seen as business steroids. Everyone talks about AI, NLP techniques or ML algorithms without understanding the differences and complexities. Effective business decision making is only possible by having accurate data-driven insight.
Soon, the misalignment between expectations from the organization management and what AI can realistically deliver becomes evident. Wrong assumptions, limited sets of information conveyed to the “Dev” teams and vague or inconsistent “why” — we are doing something — are the most common situations.
Focus on «why» rather than «how» and be less task oriented
Technology driven approach. Missing the “why”, it is the turn of IT people to support/find a solution. The “why” is quickly substituted with “what”. This is the most common situation in IT companies and creates false expectations. Creating new products and services starting from the technology makes sense in a limited amount of cases, in the majority not. One reason: the selection of technology is based on technical breakthrough rather than business values.
Identify projects that can bring value to a customer rather than choosing the project based on the technological content
Vinton G. Cerf, Vice President and Chief Internet Evangelist, at Google said “sustained competitive advantage cannot be achieved with technology alone.”
Data strategy. Everyone knows “data is the oil of the 21st century” but:
- data content, quality and governance are continually neglected or postponed.
- data governance by itself does not always directly translate findings to production without clear business goals.
- many companies have siloed departments and accessing data is limited. This avoids business understanding, proper domain modeling and poses a serious obstacle to any project success.
Sponsor business understanding involving tech people from the beginning
I am sure the above situations sound really familiar to many readers and I am sure everyone agrees that the main reason of AI failures is: separate departments that work alone!
The Road Ahead
I really believe it is possible to reverse the course. Unfortunately it requires organizational culture transformation (real digital transformation) and adoption of Design Thinking techniques.
Adopting Design Thinking techniques together with an organizational culture transformation it is possible to increase the success rate.
Framing the business problem is a key determinant to understanding what are the business objectives. This it is the first step in any modern project but it is critical in the data science area. Start with “clusterize” your customers based on “industries”. A list of known major industries are:
- Agriculture
- Manufacturing
- Services
Companies, regardless of their parent major industry, usually have these departments:
- HR
- Finance
- Marketing
- Sales
- Supply chain, inventory, procurement
- IT
For any target customers (not one, execute many interviews!) and for a given department, identify suitable use-cases making each AI solution customer-centric. Design Thinking is crucial to secure the potential of AI. Design Thinking helps to focus more on the “change” we may introduce in the customer behaviour with the new solution. Be aware of technological “feasibility” in this phase.
Design Thinking is an iterative process that puts people first.
I have found some common use-cases and they are listed at the end of this work in the section Additional Resources.
Focus on the business problem first and then think of data; tools and algorithms will follow naturally.
Based on the chosen “industry” and “department”, start gathering data, map and put in place data governance procedures.
Developers, Data Scientists, Architects, Salesmen, Marketing experts, Managers all together bring their expertise and, adopting Design Thinking techniques, everyone starts figuring out what the user needs and how to provide it, instead of looking at a problem and immediately working toward a solution.
Knowing the business goals, understanding the customer domain and having data correctly managed ensure the definition of a technology stack is a straightforward task. The solution is not technology-driven but customer-oriented and is hard to be replaced with generalized solutions.
And last but not least: measure! We are used to being measured since we were little children, at school, during college and now at work, so everyone should be comfortable with. Choosing the right KPIs (Key Performance Indicators) can make the difference between a successful product and an expensive exercise. Customer satisfaction score is one of the most crucial customer-centric KPI. This score tells us if what we have done so far to match exactly the expectations, not our expectations, but our customers! Also extremely important is the System Usability Score. A lot of the time, I have seen nice AI driven features, explained so badly that is impossible to understand what they do and why someone spent time doing it! In dealing with AI, this situation is common, but adopting Design Thinking forces one to think with the end-user perspective focusing on what really matters for him not for us!
Conclusions
Design Thinking is creative problem-solving that focuses on addressing the needs of users. It is a technique and as with every technique it can be practiced and improved. AI enables endless use-cases. We must be aware that AI is just a tool and a platform. AI alone cannot be the source of innovation and for this reason we have to avoid restricting ourselves to thinking about technology as a sole-domain of innovation. We must put the customer as the central element of discussion. We have to resist the temptation of starting from technology because the customer is hard to understand. Customers buy improved processes and rarely technology; processes are followed by people and that is why in every modern office, if we want to introduce advanced features that leverage on AI for improve efficiency, the user/customer must be in the loop and adopting Design Thinking is possible to increase the success chances.
Additional Resources
Common use-cases by department
Supply chain, inventory, procurement:
- Predictive maintenance
- Predictive warehouse fulfillment
- Vendor management
IT:
- Fraud detection/Data loss prevention
- Data quality tools
- Conversational interfaces
- Help Desk
Sales:
- Credit risk evaluation
- Knowledge graph
- Sales monitoring and suggestions
- Billing & invoicing
- Track and solve customer issues
