Powering Enterprises with AI

In this chapter, we’ll talk about practical steps to level up your organisation when it comes to AI applications. The strategy here is mostly adapted for large enterprises with enough capital and structure to carry out a plan. On the other hand, I also take into account that large enterprises are often slower to adapt, more decision-makers need to be convinced, and thus a plan to power your organisation with AI might be harder to implement. It’s also definitely different than if you’re running a smaller organisation or if you’re a freelancer. I will treat these cases separately in the next chapter, Boosting Startups with Artificial Intelligence.

Let’s now go back to enterprises. They are organized into departments by their core function, ranging from sales/marketing to customer service, product development, et cetera. Because each of those departments has a different goal and thus is optimized for a different function, it’s worth first to analyze what level we want to re-organise with AI. Is it the whole organisation, a department, a single project? Depending on what you choose, there will be different obstacles, organisational change being the hardest due to inertia and lengthy processes involved unless you have already a great innovative culture at your workplace (I’ll discuss that as well). Before applying AI, you need to gauge your organisation’s AI maturity level.

AI Maturity Levels

Let’s look at the potential steps you might take to apply Artificial Intelligence at your workplace and what challenges you might face along the way. Gartner, a research and advisory company, has defined maturity levels for AI adoption as follows:9

Level 1 - Awareness: Conversations about AI are happening, but not in a strategic way, and no pilot projects or experiments are taking place.

Level 2 - Active: AI is appearing in proofs of concept and possibly pilot projects. Meetings about AI focus on knowledge sharing and the beginnings of standardization.

Level 3 - Operational: At least one AI project has moved to production, and best practices, experts, and technology are accessible to the enterprise. AI has an executive sponsor and a dedicated budget.

Level 4 - Systemic: Every new digital project at least consider using AI, and new products and services have embedded AI. Employees in process and application design understand the technology. AI-powered applications interact productively within the organization and across the business ecosystem.

Level 5 - Transformational: AI is a part of business DNA, it goes into every business process, and it is a natural framework to work with. Every worker knows the strengths and weaknesses of AI.

To that, I can add Level 0 - No Awareness, where there’s simply no discussions of AI or AI-related solutions at work, and workers are not aware of what exactly AI means and how it works. Unfortunately, this is the most common stage right now, even among the largest organisations. Following a study by McKinsey10 on Artificial Intelligence adoption, 53% of respondents are at Level 0, even though the studied companies were already more aware than most of what AI can do for them.

Fortunately, these statistics grow each year. Another study by McKinsey shows a 25% growth11 in AI adoption across various industries. Some major criteria why this happens can be tracked to:

• the cost of implementing AI is falling with AutoML solutions (automatic machine learning) and easy to use platforms, which automatically analyse data without a need for a data science team.
• the general awareness of how AI can be used is growing with growing coverage in major magazines and newspapers.
• there are more and more data scientists and machine learning engineers available on the market, thus it is easier to find experts to work with on AI. Especially if a company is willing to hire an external team from a software house to jump into AI.
• the track record and business case studies are more widespread. It’s easier for managers to see the real benefits of using AI in cases similar to theirs.

You can also apply the same levels of AI maturity to society as a whole. If you think about it this way, then we are currently entering Level 3 - Operational:

• There are numerous AI pilots going on around the world in each domain of our life. Most efforts here are made by private and public companies with substantial help from governments as well. Examples: traffic optimization, cashless shops, e-government, big data analytics.
• Many countries have already adopted AI strategies and approved a budget for AI solutions. Publicly administered funds are directed towards startups and R&D. Governments put money into universities creating AI faculties.

To achieve a higher level of maturity as a society, we need to:

• broaden the scope of AI education, starting with primary schools, introducing Python as the main tool for computer science classes.
• include AI and machine learning framework into education at business schools.
• popularise machine learning and data science knowledge among the general public, showing strengths and weaknesses of using AI, overcoming fear, and showing potential benefits.
• incentivise businesses to invest in AI solutions and use AI in their processes.

In general, I’m optimistic it’s just a matter of time when we transition to the highest level of AI maturity as a society. Artificial Intelligence is such a transforming force, with a proven track record already that it’s a question of when it will happen rather than whether.

Nevertheless, you can go into level 5 right now with your organisation by carefully planning and mapping your activities related to AI. There are two main tasks you have to do to start:

• assess the maturity level of your AI adoption,
• follow a list of recommendations to go to a higher level.

The first step is to assess the maturity level of the company by asking what the company is doing when it comes to:

1. team/people
2. data
3. tools for optimization and automation
4. vision and values

For example, let’s say we are concerned with the Marketing Department responsible for Customer Analytics. Data-wise this means we will be working with the following data:

• customers’ feedback
• effectiveness of in-store promotions
• planning and forecasting
• customer segmentation

The first step toward assessment of the AI maturity level of your company is to gather data from the department by asking your coworkers and looking through past projects. You should start by asking the following questions to assess properly how AI is used in your organisation.

Team

1. Is there a dedicated AI team working on machine learning and NLP algorithms: data scientists or machine learning engineers?
2. Is there a team of data analysts who analyses, cleans, and processes data?
3. Do some of the jobs/tasks at your department are tedious and repetitive? What’s the biggest obstacle in automating them via AI?
4. Is there a person responsible for managing the AI team?

Data

1. How is data stored currently? Do you store it in-house or using external services/cloud?
2. What kind of data is currently stored?
3. Do you clean and process data in any way upon reception?

Tools

1. Do you use existing tools for data analysis? What kind of tools are you using? How much automation is in the solutions you’re using?
2. Do you use any Machine Learning platforms to extract insights from your data? Which data do you submit to it?
3. Do you use algorithms built in-house by your data science team to analyse your data?

Vision and values

1. Do you think about applying AI to each new project?
2. Do you continuously try to improve existing processes by using machine learning algorithms and available AI systems?
3. Are there communication channels between the AI team and other teams so that AI can be easily implemented whenever needed?

Based on those questions, you will be able to assess the AI maturity level of the company by measuring it against Gartner’s description. The answers to the above questions would certainly allow you to pinpoint the AI maturity level and moreover might indicate what you should do in order to level up.

We will now go through each level to discuss how you can level up, no matter what’s your current stage. This applies to each department or an organisation as a whole.

Solving maturity issues

Let’s now talk about how a company can transition to the next level, no matter where it starts. Building AI maturity is crucial in staying competitive in our digital world. Enterprises are challenged and beaten by technological startups because of a lack of innovation or a too-slow adaptation to changes. In this section, I discuss each level and potential challenges associated with it.

Awareness (1)

In order to ascend to the very first level, one should organize a workshop for the whole department about AI, machine learning, data science, and their potential use in real-world scenarios.

Potential challenges:

• educating staff of new ways of doing marketing,
• overcoming fear of AI and explaining it in simple terms to everyone.

From Awareness (1) to Active (2)

To progress past the first stage, one needs to connect the company’s data to the system able to analyse data. This way, the first generic algorithms can be put into action, and the user will get first visualisations and preliminary cleaning of data. This can be an external ready-to-use system built by another company, more experienced with AI. Depending on the context, an enterprise can start with:

• social media monitoring,
• competition monitoring,
• documents analysis,
• invoice analysis.

Self-service analytics becomes a significant factor in making decisions based on data, as they allow us to observe threats and opportunities for growth.

Moreover, one needs to form a team of data scientists or programmers who would operate daily with external systems and machine learning models. At this stage, it can be done by hiring external teams to complete the first pilots and have a sense of AI capabilities.

The whole department needs a workshop to get accustomed to the machine learning framework and thinking about data and processes with an AI in view.

Potential challenges:

• finding talented people within the organisation or hiring a new team to use a machine learning system; can be solved using an AI consultant agency.
• learning to use external platforms; getting accustomed to new software is often hard; people need time and workshops to overcome fears.

From Active (2) to Operational (3)

Data Science Lead is hired to manage the data science team and introduce new AI experts, build a practice, and form a vision for future products.

The company starts using a machine learning/analytics system to experiment with machine learning models and extract data.

The company goes beyond analytics into more advanced methods like:

• NLP for analysis of customers’ feedback,
• prediction of future customer behaviors,
• segmentation of customers into similarity groups using AI.

Potential challenges:

• attracting Data Science Lead is hard; one needs to spend usually a lot to attract top-notch talent and create an excellent environment for growth, which would be open and research-friendly.

From Operational (3) to Systemic (4)

At this stage, the AI team should work on their own algorithms, which then would be deployed within existing and new products.

Implementing a visual system for building models will allow for fast iteration, testing, and deployment.

There is an active communication channel between the AI team and the rest of the department.

Potential challenges:

• open communication channels between the AI team and the rest; the way to overcome it may be organizing regular 10 min stand-ups to ask questions
• expanding the AI team with machine learning engineers; hard to attract talent - if Data Science Lead is really good, he’ll be able to do it though.

From Systemic (4) to Transformational (5)

The whole department is educated in machine learning frameworks and AI pipeline through a series of workshops, which are run on a continuous basis. This will provide broad AI training to the whole team.

To each project/team, there is assigned an AI supervisor responsible for thinking about how AI can transform a given project or change it substantially.

The company leverages the in-house AI team with its own solutions as well as available platforms/services and open-source code (GitHub).

Decision-making is based on data and is enhanced with AI.

Potential challenges:

• explaining how real AI works - without overhyping it and showing weaknesses as well as strengths; it often takes time to do that part (workshops, seminars, one on one meetings).
• building a framework for what is being done internally and what externally when it comes to data and algorithms; it takes an understanding of what the goal of the company is, where lies the competitive advantage, and what will be the most important in 5 years.

Fostering a culture of innovation

One of the key components of being successful in implementing AI at a transformational level is building your company’s culture around innovation and AI in particular.

A culture of innovation means creating a culture where every employee feels he or she has some level of autonomy to think independently and find new ways to solve problems. If you’re the one in charge, it means you should put forward your decisions, but leave an open window for new ideas coming from your team. Trust is important.

You should encourage taking a reasonable risk and allow the uncertainty of results if you want to foster innovation. Developing new ideas is an iterative discovery process and as such needs people and ideas to flow freely. Understanding that this does not always lead to a product is also a key. Failure is an important aspect of building an innovative culture. You can’t hide failures. Instead, talk them over with your team. A truly innovative culture rewards people for their involvement in generating ideas and executing them, it doesn’t just pay for ideas. You should think of innovation as a mindset and a framework for running a business process rather than an end goal.

Openness is another critical aspect of building an innovative culture. The innovative culture welcomes different points of view, different perspectives and seeks to associate disparate ideas and technologies into new products and services. You never should hear an argument “we’ve never done that before” as a way to shut down ideas – on the contrary, it should be considered a challenge worth pursuing in the innovative culture.

How to shift an existing culture

Changing a corporate culture is a difficult task, and it should be done slowly. Don’t force it; adaptation to changes at a massive scale is always painful. It should be done in steps:

• create a vision for the future,
• describe your vision in a story and pass it to your closest employees,
• create a plan of actions you want to reward and how,
• prepare materials for your employees and the necessary retraining programs.

There’s no change without tracking how employees answer to senior management strategy. You should hear feedback and act upon it to make this shift towards innovation as smooth as possible. In the end, there are three ingredients in shifting a corporate culture towards innovation:

• rewards,
• hiring,
• retraining.

By rewarding innovative behaviors, hiring new talented people, and expanding the skills of your employees, you’ll be able to create an excellent environment for innovation.

We should remember that the emergence of AI in the workplace requires a massive re-skilling of employees at all levels, especially when it comes to:

• adaptability,
• social intelligence,
• communication,
• problem-solving.

Re-training of employees is crucial to implement hybrid solutions of people collaborating with machines on a daily basis. In some positions this is already common (think traders). Still, adoption of machine learning, especially the most cutting edge research, will nevertheless require new skills to use available tools in the best way.

That’s why building a mindset of life-long learning (courses, seminars, books) plays a crucial role in creating an innovative culture. Moreover, it builds trust whenever an organisation invests in its employees and cares about their growth.

Hiring AI Talent

Often the most challenging part of executing an AI strategy is hiring talented people. There’s a shortage of talent on the market, and the best are usually picked by Google, Facebook, Amazon and other big tech companies. Even large enterprises which are not strictly technological in nature lack talent. So in order to overcome this challenge, it’s critical to understand how to attract talent to your organisation.

Computer scientists, in particular data scientists and machine learning engineers, are looking for an intellectually stimulating environment. More often than salary, what’s important is the possibility to learn, solve creative problems and be challenged intellectually. On the other hand, the fewer meetings and administrative burden,s the better for them.

Those conditions often are hard to meet in organisations with a corporate, rigid structure. That’s why changing your culture slowly to a culture of innovation is a necessary first step. And this has to be as practical as it gets. Focus on results.

With a culture of innovation installed, the next crucial thing is to have a clear cut strategy for what kind of problems you want to solve with AI. The more down to earth and practical it is, the better. You want to attract open-minded, hard-working, bright individuals, and for that, the best way is to present them with a meaningful challenge, something which is technically hard and something which solves a real problem at the same time.

Do research on various platforms about the potential payroll to see what you should offer. Then don’t limit yourself to just posting job interviews, outreach to people you want to hire, especially to senior data scientists who always tend to have a job and are seldom looking for one (talent shortage again!).

Building an innovative culture in enterprises

Summing up this chapter, we have covered AI maturity levels and discussed how to progress to the next level, no matter where you start. It’s crucial to become an AI-enabled company to stay competitive in the future.

A key component to progress in AI maturity is fostering an innovation culture. We’ve discussed how it’s possible to change corporate culture. It’s a slow and steady process, but it pays off in the end. An environment where innovative ideas are rewarded and not frowned upon is perfect for attracting top talent in machine learning and data science.

---