Everybody knows the enterprise needs to move towards a digital future. We speak of “data-driven organizations“. But in marketing in particular, but eventually in any other part of the modern enterprise, there is one question managers are asking these days: faced with the onslaught of artificial intelligence solutions and all the buzz about it, when are we really ready for AI?

There are lots of reasons to fear AI. Managers used to be completely unsure what AI is, what it is, and – most importantly – how much it costs. Do I have to hire new people for it? How much change is involved in general, from personnel to equipment to internal processes?

Let’s start with the most important question:

Where can this new tool fit into my business?

The first step is to draw up  a list of possible applications for AI in your business. This can be customer service functions, predictive functions (predicting churn rates or conversion rates for example), or functions linked to gathering and evaluating data for the C-suite. Where will the AI impact the customer, and where is it a purely internal consideration?

Once you have this list, choose the easiest use case, where deployment involves the smallest amount of change. 

The key here is to identify those cases which can be managed by the existing staff, perhaps even within the existing budget, and without big staff changes. More importantly, try to identify which use cases would need a data scientist — those can be notoriously hard to find and demand huge salaries.

Read also: How to Estimate an AI Project’s Data Needs

The challenge for the decision maker is to understand all the concepts, from AI to machine learning and deep learning, and everything under those headlines.

Read: Beyond the Hype: The difference between Artificial Intelligence, Machine Learning, and Deep Learning

Start With the Pain

Everybody knows that artificial intelligence is the key to competitiveness. To begin deployment in your enterprise, don’t try to understand big concepts or the ultimate potential of AI, but start with concrete use cases where you think AI might have an immediate impact on the bottom line. In other words, deploy AI where the pain is. 

Good salespeople of AI solutions understand this. They don’t try to sell you big words like deep learning but will help you solve concrete problems in your organization.

Talking about the technology rather than solutions is tantamount to a car salesman explaining the combustion engine rather than extolling the virtues of the car’s features.

It’s All About the Data

Once you have identified the pain point where AI will come in most handy, it’s all about the data.

Raw data is worth nothing if it cannot be manipulated to produce meaningful results and similarly, machine learning and AI cannot happen without sufficient and relevant data.

The data does not only have to be relevant, but it also has to be cleaned up and prepared for whatever use you are envisaging. Questions you are facing are: is there enough data to make the AI deployment feasible? Is the data readily available? Is the data biased? Do you have control over the data source? and many more questions.

I have also written a detailed description of general machine learning deployment which covers many of these issues in detail: Read Machine Learning in Organizations: First Steps

Managing Change at the Top

What it comes down to is the readiness of management to sit down and understand the questions involved.

Some managers may be unwilling or unable to grasp new concepts. Others may be willing to change, learn the new skills, and become change leaders who then spearhead the adoption of AI.

I often compare these to the advent of the PC in companies. Lots of employees were left behind because they did not want to or could not make the switch from the typewriter.

The HR impact of AI cannot be underestimated. AI will involve changes in business processes (streamlining or eliminating them), thus impacting anyone in the organization. But it may also involve output directly directed at top management (think of reporting). Only if top management is ready for AI should you consider a significant deployment.

Impacting the Customer

The final hurdle to adopting AI in the organization is the impact on the customer. Think about the relatively primitive chatbots used by airlines. You can only deploy these if you have a large enough customer base actually willing to use chatbots. If most of them prefer calling by phone, your lovely chatbot will gather spider webs.

This is not a joke: one of my customers thought of introducing an internet chatbot for insurance questions directed at senior citizens. I dissuaded them from the endeavor.

Incidentally, this last point of customer impact is often the most easily overlooked. An AI system making an automated recommendation for customers of a real estate agent, for example, caused utter confusion among customers as they found themselves confronted with choices they themselves would never make.

Read also: What Chatbots Can Currently Do for your Business

In Conclusion

In summary, deployment of AI solutions involves the following four touchpoints:

  • Which applications should come first?
  • Is there enough GOOD data to work with
  • Is management ready for deployment
  • How does the introduction of AI affect our customers?

You may also want to read:

How to Estimate an AI Project’s Data Needs

What Chatbots Can Currently Do for your Business

5 Ways Towards a Data-Driven Business Culture

Martin Hiesboeck is an international business consultant with a focus on companies’ operations in Asia.