How to build large AI powered engineering teams

Image from pixabay

What are we talking about here?

The hype around AI could be very real but to be fair that must be true for any technology that has the ability to change the world. Sit back and think about the last few days and one may find several references in the social media about how AI will change the way we live our lives especially in the area of healthcare, finance, entertainment or travel. Anyone who taps into the tech news or is somehow related to the digital transformation must know already about intelligent machines. Some of them may inadvertently perceive it to be just another fancy word for robotics or how it will take away their jobs someday but there is so much more to it that must be talked about. AI optimists believe it’s just the beginning of a revolution that has an ability to change everything about the world as we know it.

This has led to several organisations adopting AI strategy into solution building and add more value to their business. There have been many conjectures drawn about how to become an AI powered organisation and quite often they fail to converge simply because there are multiple dimensions to understanding AI. As a business owner it is important to know what path to take to embrace AI and imbibe into the organisation’s culture while minimising the risks and tread carefully.

The thing about culture is that it requires constant nurturing and attention. It evolves over a period of time and requires constant efforts. How does one start to build the right culture amongst various teams within an organisation? Considering it is expensive, difficult to understand and even more difficult to correctly implement, how does one proceed? It is prudent to create a framework for AI strategy to build the right culture that aligns with the organisation’s vision providing tangible benefits. A culture of innovation using artificial intelligence…

Double-clicking on the AI-path

The top challenge in front of an organisation is to figure out exactly where to start. Identify what business use cases are important to be solved by AI. Below framework can be used to create an organisation wide AI strategy which can be adopted by various teams.

Image by Author
  1. Develop the right expectations and set objectives

Irrespective of the perception of an organisation about its awareness on the latest in AI, it is extremely important to still brainstorm and form an opinion on what AI actually is and how it can be relevant for their business. It takes time and effort to acquire a basic understanding on AI and how it works. It is extremely important for cross functional teams to sit down and converge on what it means to them. This does not mean to brush up the mathematical skills and start solving complex equations but to know how powerful AI can be and how it can be used. Can the organisation really benefit from it before making any huge investments? What kind of problems does it solve?

This may require extensive reading and discussions, gathering insights from conferences. Inspiration from work done by others who are past this stage always helps. It’s often tough to encourage people to go ahead and move out of their comfort zone and gather ideas on how it can be relevant for their business. But this stage is very important to induce the right culture in the teams. A culture with enough freedom that favours innovation and a positive atmosphere about the emerging technologies. Always helps to stay away from the pessimists who believe — “This too shall fade away over time just like the others, so why bother learning it at all”. AI is not a tool or a library or a technology but it’s a paradigm shift, a way to solve complex problems efficiently like never before. It is here to stay, so hop on..

2. Explore the unknown

Get your hands dirty. Some of the popular use cases can be picked up to firm-up the understanding of how AI actually works. Find out various tools, frameworks and libraries which are preferred by others and learn them to develop AI solutions. How a problem can be solved or is usually solved by others in the industry. Learn some basic machine learning techniques. Identify regression vs classification problems, data cleansing needs, infrastructure needs etc. Find a way to compare the results of using two different algorithms. Figure out how chatbots are developed, how computer vision problems are solved, how forecasting solutions are built. What to do if training data set is highly skewed? How to set up data and ML pipeline. Reach a stage when it is easier to provide an AI based solutions to business problems or at least to have an opinion on which algorithm to use when.

It is imperative to know about how the AI solutions are deployed in production. What if the real data in production is much different from the one used during training. Can this be handled by retraining the AI models at run time on production or would it require offline analysis and training? How much data is enough for training purposes? What kind of impact it will have on the accuracy levels?

Experiment – analyse – discuss – repeat.

3. Just do it — develop multiple use cases

This is the phase in which teams should aspire and come forward to develop small solutions to solve some real problems relevant to the business of the organisation. Conceive ideas, do multiple POCs, demonstrate, gather feedback, brainstorm and improve. This is the phase in which teams really need to work together to show what they can do in the field of AI and how it helps the business. This phase is the most important one as it gives the decision makers confidence on the capabilities developed by the teams. Many decisions are needed to be taken in this phase like funding for team members, infrastructure etc. Decide exactly how relevant AI is to the organisation. This phase should provide enough clarity to decide how much investment should be made in this process. Get enlightened on whether it makes sense to do “AI research” from scratch or just do “applied AI” with what is available to use or a mix of both.

4. Grow and improvise

AI based solutions need constant attention and monitoring. Always be on the lookout for something new that comes up and is available to use. AI is such a fast paced field that everyday new research gets published. New benchmarks are set, improvements are made over the past AI models in terms of accuracy, performance and efficiency. Keep on maturing the solutions and make them better. Eventually this leads to a stage where teams improvise and come up with new innovative solutions. One could figure out ways to tread paths never explored before and make major breakthroughs. This builds proprietary solutions or patents in the name of the organisation and gives an edge over competitors.

Build guidelines around the AI governance. What should be the accuracy of the solutions for them to be well accepted in the organisation and its customers? How to minimise the human biases? Who leads the teams responsible for AI engineering? Define roles and responsibilities within the organisation to establish clear accountability.

5. Train and spread awareness

It is imperative to help other teams within the organisation to understand AI and spread awareness about the solutions developed. The more aware an organisation is on the AI based initiatives, the better the chances are to excel. This may require to give training sessions, publishing white-papers, newsletters, presentations etc. Others don’t have to start from scratch. Teams should help each other to learn, motivate and improve. Sometimes an idea may get transpired through these sessions which translates into a useful business use case.

Collaboration is the key when it comes to something as complex as AI. Also any AI based system is meant to solve a real business problem identified by the product team, to be sold by the sales team and branded by the marketing team. They all need to know how it works and what exactly it does. Hence the AI learning path needs to be adopted together by all respective teams. As the saying goes in IT world, no good ever comes from the work done in seclusion anyways.

Something that needs to be said

It’s evident that small steps in the right direction can really push an entire organisation to embrace a culture of innovation. It’s a constant endeavour to adopt the process and quite essentially, that is the goal itself. Avoiding unnecessary dissonance and keep moving forward.

Many believe that AI will do great things in the future. A future in which AI will make us more efficient and more productive. A future in which AI holds the key to solving global problems at a massive scale. A future in which AI helps us in making better decisions. A future in which it will outperform humans.

But some believe, that future, is already here..

Building AI culture in organisations was originally published in Towards Data Science on Medium, where people are continuing the conversation by highlighting and responding to this story.