Demystifying AI

Starting AI on the right track

Now that cloud infrastructure can support the access and use of large volumes of data, businesses are at the point where that value can be tapped. Fueled by new technologies like ChatGPT, optimism is high around the potential of AI with regard to automation: AI is going to do all the things we don’t want to do and free up lots of employees who can work in more productive areas of the business.

That optimism is not proving to be particularly well-founded, not because AI can’t accomplish “all the things” but because the path forward has not been as effective as it needs to be.

I was recently joined by Cameron Turner, Vice President of Data Science, and a group of senior technology leaders from a range of industry sectors to consider how to move forward with scalable AI initiatives that make a positive impact on business.

These are the topics and insights we covered in a very useful discussion about how and where to take the organization’s first steps into AI.

The pendulum swing of AI

People are euphoric about the potential of AI to do automation, that AI is going to wash the dishes, drive us places and do all of our laundry (literally or figuratively). However, some of the early promises have yet to deliver.

What we are finding is that equilibrium is really on the ability of AI to superpower and enable the people and the talent that is already within our organization. It’s not just artificial intelligence or natural intelligence, it is just intelligence—human plus machine, doing better than either alone.

The great data tension

The potential energy of all of your data has been sitting behind a dam where people can not access its value. But just because you are unable to access this data and its potential value now, doesn’t mean you can’t in the future.

When speaking to clients, there is always a great idea in mind of what the potential could be. The data is there but organizations could do better at predicting, personalizing, or understanding the upstream supply chain and downstream partnerships through the data that we have access to. But they all have the same sentiment of “we don't really have the way of integrating or encapsulating that data or presenting it in the way that fits into our current processes and our current organizational structure”.

You want to create a bridge between all of the data assets that your organization has stored and the value that can be captured in the market, but how can you do this right away without building a giant data platform?

Although data maturity journeys will look different, every organization can benefit from this approach to data preparation.

1. Walk before you run

   Make your leap into AI more of a hop. In today’s digital environment, it is possible to start very small and scale from there. Identify an area of opportunity that can be pursued with small financial, team size, and technology requirements.

   Going small limits risk and can get the initiative into the business (producing results) quickly. It also allows efficient iterative improvements before the tool is scaled out further. Select a short time to value, high ROI opportunity, and then encapsulate that value in the form of products that can scale without heavy human systems.

2. Align AI with overall business strategy

   Like other parts of the digital infrastructure, AI is a tool that needs to support and enhance the direction the company wants to go. What is your superpower? What makes you great? Finely focus your AI initiative on the answer to those questions rather than attempting to design a large, all-encompassing program.

3. The right place to start will make sense

   In any company and on any working day, there’s an amount of rote and repetitive activity. As soon as you do something three times in a row in the same way, you can code that thing and let it happen automatically.

   Whether it's form processing, looking at quality diagnostics, food spoilage rates, or one of many other items, there are rote activities that can be optimized. Think in terms of this type of repetitive process when considering your AI initiative.

4. Avoid the “Field of Dreams Model”

   Be sure that you can answer the question “how is this actually changing the business” before you start. You can almost get drunk on the ability to answer every question with your data without actually ever achieving the outcome. So be sure to stay grounded in why you are doing this in the first place.

   Like any major technology initiative, moving forward on AI with a “when we build it, they will come” belief is not the road to success. Tangibly connect your AI initiative to the business by way of a user or stakeholder who sees value in having that capability in their area. They will not only give you the answer to “So what?” but will readily see the value of the outcome in the work environment.

Demystify the AI development process

There's not much difference between AI technology development and other forms of technology development. This means that the plan and design steps will be familiar to IT staff and, presumably, business users.

That said, approach this as a transparent and collaborative undertaking with the business. Don’t just capture requirements, go into the development cave for three months, and come out with version 1. Instead, apply an iterative process that includes cross-functional stakeholders who are looking at all aspects.

Part of the planning can be quite quick and straightforward: Get everyone into a conference room for a few hours, ensure you have the right data access, and start capturing questions on a whiteboard. Run some SQL together and see what’s there. Out of that process, you can codify applications and scale to stakeholders who weren’t in the room.

Deemphasize the need for data scientists

It is a common belief that data science must be a core skill in an AI initiative. This has created something of a “sky is falling” reaction in business media, as the shortage of data scientists is cited as a big challenge for businesses.

In terms of in-house staff, including a data engineer with deep knowledge of the company’s data is more important than having a team of data scientists. Data storage and access are the “fuel” for your AI activities, so having a detailed understanding of your data will produce a high ROI. Acquire the needed data science skills through a partnership with a trusted third party that can support your development process as you build your internal competencies.