Expert Advice for How to Incorporate AI Into Your Business

“Artificial intelligence is going to be transformative,” yada yada yada, but how do you really approach the problem of implementing AI in business? Do you need to build a whole new AI team? What about the pitfalls, or the practical steps you need to take to create organizational change? These are questions on every leader’s mind.

IDC1 puts it well: “As GenAI becomes front and center of conversations in the technology industry and beyond, the underlying question being asked by the market is: how do organizations accelerate their journey with GenAI to deliver business value?”.

We’ll begin to answer these questions with tips from AI experts we interviewed (you can find the rest of their insight in the Enterprise AI eBook). But before getting into their advice, we have to cover two important aspects that are foundational to a winning implementation of AI.

2 essential aspects of implementing AI in your business.

In order to realize the value behind the AI buzz, you need to become an AI enterprise: a business that is engineered to extract practical, tangible value from AI.

An AI enterprise effectively combines AI with more foundational technologies: data and process.

Why you need data.

Data is essential for AI, as you’ve probably gathered given the reams of data that LLMs and other forms of generative AI need to operate (GPT-4 has more than 175 billion parameters). The more data you feed an AI (and the better that data is), the better the AI’s output will be. AI’s usefulness corresponds to the data it’s given. Keep these three considerations in mind when thinking about the kind of data you want to provide AI. The data should be:

Fresh. AI is most helpful when it can access the latest data. The more recent the data, the most reliable and effective its support will be. (Of course, historical data can also be critical, too, particularly for industries like insurance that sit on decades of data.)
Yours. To make AI valuable in your organization, you need to feed it on your data, not a generic knowledge set from a public AI model. This way, it can provide tailored, actually relevant insights for your company. (A private AI strategy ensures you can do this safely.)
Available. How often is data locked away, thanks to data management blockers? To make data available and accessible by the AI, you need a strong semantic layer (one unified place to see and interact with your data) over your enterprise data. You’ll need to gather your data for training the algorithm, and/or querying it in real-time to supplement your questions. Without a semantic layer (like a data fabric) over your data, you won’t be able to do this and create valuable AI models.

So obviously, you need good data to incorporate AI into your business effectively. But good processes are just as important.

Why you need optimized processes.

Process is the mechanism that operationalizes AI. While AI is a powerful capability that adds value to your data and your employees, it’s not the only thing you need. You’ll need to be able to route a lot of work to and from AI, between it and automation technologies and employees. This is why process is so necessary.

AI can do a lot, but it can’t run your organization, and you’ll need sophisticated workflows to manage the handoffs and ensure AI and the other aspects of your process are working seamlessly together. Working together, process automation and AI can accomplish much more than they could separately.

Now that we know the fundamentals, let’s hear how experts at organizations like Amazon Web Services and KPMG LLP recommend incorporating AI into your business.

8 tips for implementing AI in business operations.

There’s great pressure from every direction to bring AI into your enterprise, not least because of the need to keep up with competition and customers. But the way forward is not exactly clear. That’s why we interviewed experts to provide advice on where to begin, along with other relevant AI topics like data privacy, trends, and risks.

1. Make AI implementation a top-down strategy.

This advice was echoed across experts: leadership needs to drive AI implementation. Piyush Bothra, Field CTO, Principal Solutions Architect at Amazon Web Services (AWS), points out that companies adopting machine learning and AI as part of their strategies now are setting themselves up for a trajectory to outpace their competitors in the coming years.

“How can your business emulate this? He advises, “Companies who are doing well pull AI in as part of their operational planning. . . mandating that business leaders, product managers, data scientists, and engineers get together and start thinking about how they can improve the customer experience using machine learning and artificial intelligence.”.

Hasit Trivedi, CTO Digital Technologies and Global Head – AI at Tech Mahindra concurs, “The success of a project hinges on obtaining management’s alignment with the strategic plan.”

Can an AI initiative be led from the bottom up? Sure, but Hasit adds, “Ideally, this alignment should stem from a top-down approach, unless there are remarkable instances of effective bottom-up leadership. Although a bottom-up approach may occasionally yield positive results, the consensus is that a top-down strategy tends to be more effective.”

For those struggling to convince leaders, the hype surrounding AI may help. Wipro’s Global Head of Strategy, Strategic Partnerships & Solutions Piyush Kumar says, “The best organizations take a top-down approach, where the c-level buys in to the critical business value and differentiation AI can drive. This may be easier to achieve with the news hype and promise of generative AI.”

2. Build a strong data foundation.

Again, we need to emphasize the importance of data to your AI initiative. AWS’s Piyush Bothra, points out the necessity of providing “the right tools and data for the teams to become successful in this journey. As we all know, machine learning and artificial intelligence cannot run without data. If there isn’t easy and well-governed access to data for engineers and data scientists, then those organizations will fall behind in terms of innovation.”

Wipro’s Piyush Kumar concurs, adding: “You need the right data available to get the output or results you want. . . . All that is critical for driving results and proving the value with confidence.”.

He encourages businesses to ask these questions to prepare their data for AI:

How are you cleansing your data?
How are you preparing your data?
How do you do feature engineering on top of that?

3. Map AI to business goals.

AI is worth pursuing, but only in the context of business goals.

When asked what sets top performers around AI apart, Todd Lohr, Principal in KPMG LLP’s Technology Enablement Practice, remarks, “They view it as a technology to solve a specific business challenge. They don’t treat AI as a hammer looking for a nail.”

He encourages companies to ask specific questions to evaluate a specific use of AI:

What value are you creating for stakeholders?
What strategic propositions are you trying to accomplish as a business?
How can you use AI to pivot or accelerate that strategy?

“Top-performing organizations stay true to their business strategy and use AI as an accelerant.” – Todd Lohr.

Not doing so can lead to wasted resources, delayed priorities, and, sometimes, outright failure. Roboyo’s Chief Technical Officer, Frank Schikora, advises mapping AI to clear value for the business.

He says, “I’ve seen failures time and again where the time-to-market was underestimated and the place where it would fit within the rest of the business process just wasn’t clear.”

4. Pick the right use case.

Starting in the right place with AI is easier said than done, but experts gave a couple pieces of advice here. Piyush Kumar of Wipro suggests picking use cases that showcase value. He adds:

“Focus on early wins with low-hanging opportunities while you build and invest toward the bigger vision.” –Piyush Kumar.

Then, once you’ve initially selected an AI use case, ensure you’re working in tandem with your legal and security or risk teams.

Xebia’s Managing Director and Co-founder, Intelligent Automation, Akhilesh Natani, reminds enterprises of the risks: “Organizations must be able to continuously track and adhere to a continuously evolving regulatory landscape. To navigate these waters, there must be strong alignment between AI practitioners and the legal and security or risk teams to evaluate AI use cases and feasibility with regulations in mind.”

5. Fail fast.

As in all new initiatives, creating an environment where teams can fail fast breeds more creativity and enables quicker progress.

KPMG LLP’s Todd Lohr advises that with technologies like these which are reasonably newer and not yet perfect, “You have to have that pioneering spirit to move into new technology areas and keep the focus on innovation.”

AWS’s Piyush Bothra encourages teams to start with the basics: “Give teams enough time and resources to experiment. Fail fast, adapt, and re-experiment. That’s the best way to learn.”

6. Create an AI center of excellence.

Several experts suggest enabling AI by creating a center of excellence.

Wipro’s Piyush Kumar notes two benefits: “An AI center of excellence (CoE). . . helps them set governance and drive evangelization around what AI can do across the company.”

“A pivotal factor in achieving success is the formation of a cross-functional team to tackle the project.” –Hasit Trivedi.

These centers of excellence should include more than just technical experts.

Tech Mahindra’s Hasit Trivedi warns businesses against this pitfall: “At times, technology becomes the most important thing in projects, and that often leads to exclusive involvement of technical experts: for instance, AI-driven projects where data engineers and data scientists dominate the scene. However, this may lead to disappointing outcomes.”

7. Identify threats.

In addition to the regulatory landscape, organizations must identify other hurdles that could get in the way of incorporating AI into the business.

Xebia’s Akhilesh Natani adds: “The key is to identify the threats, then find solutions that neatly define the responsibilities and preventive actions for these threats.”

He prioritizes the common threats into these categories:

Governance.
Regulatory landscape.
Internal issues like misuse of data.
Alignment of AI strategy with wider technology strategy.

Akhilesh explains these threats in more detail in the full Enterprise AI eBook.

8. Provide for AI maintenance.

Once AI is up and running, you’ll need to maintain the models. Roboyo’s Frank Schikora encourages building “AIOps functionality.” He notes, “Just because the model gives the right answers in a user acceptance testing (UAT) environment, you still need to train again and then validate. I think this often isn’t taken into consideration—people think they have nine months of highly trained and expensive people to build a model that will save them enormous amounts of money without having to look at it again.”

Why use AI in business?

At the end of the day, this groundwork pays off.

IDC reminds us of the impact AI can have: “GenAI has the potential to put all [the vast amounts] of internal and external data to work in driving still greater scale and speed for digital businesses and further accelerate business impacts.”

¹IDC, Generative AI: The Path to Impact, EUR151153223, Aug 2023.