Using AI to drive business success

Integrating artificial intelligence (AI) can be a game changer for scaling your business. Although AI has been used for years by businesses of all sizes, a number of factors have pushed AI to the forefront of technology adoption considerations. These include the proliferation of data generation and collection, risk mitigation, compensating for workforce and talent shortages, responding to consumer demand for better customer service, and a host of other business challenges.

Business leaders believe AI is arguably the most disruptive megatrend today. It may offer more possibilities than any technology in recent history. However, when we talk to our customers, we’re hearing that all of the possibilities make AI seem overwhelming. They’re asking:

• How do I even get started with AI?
• What are the risks and how do I manage them?
• How is it going to impact my business?
• How do I find experts to guide me?
• How do I get buy-in across the business to integrate AI into day-to-day operations?

Andy Markus, our Chief Data Officer (CDO), is keen to answer these kinds of questions. AT&T has been a pioneer in AI for many years, and we’ve since embedded AI expertise throughout the company. In a recent interview, Markus shared that AI is so vital to our business that turning it off now would be like turning off the electricity.

“Because of AI, we’re able to be much more proactive in effectively protecting our customers from fraud. We’re able to use data to deliver marketing campaigns our customers care about. We execute on higher standards to plan and administer our network reliably and securely. Everything is impacted.”

Markus knows firsthand about taking advantage of AI to drive business success and avoiding the pitfalls. In our time together, he shared his knowledge about how you can create a roadmap for integrating AI—and take your business to the next level. 

AI is playing a pivotal role in ushering in the “Fifth Industrial Revolution,” also known as 5IR and Industry 5.0. It’s defined by a deeper integration between humans and technology. This includes more automation and use of data to “educate” technology so it can help us make decisions. It even integrates environmental sustainability. 

Markus said that every industry is jumping into AI and the benefits of doing so are many, especially in automating even routine tasks. Examples include using chatbots to answer questions through your website, using IoT devices and data for predictive maintenance to reduce equipment downtime, and helping detect critical errors in everything from accounting to health records for quality control. 

“The magic happens when you connect automation and AI to your data,” Markus said. 

Markus regularly attends conferences and events that give him tremendous insight into what others are doing—or not doing.

“It’s interesting to see the companies that are leaning in, like us, and the companies that are taking a wait-and-see approach. I know I’m biased, but it feels like the companies that are taking a wait-and-see approach are going to be left behind to a degree that it’s going to be very hard to catch up.”

Although AI has been used for years, implementation at scale across the business community is new. What it is, how to use it, and how to integrate it into a business is still overwhelming for many. 

“With traditional AI, there is a barrier of entry to find the data scientists who can really bring that technical acumen to the solution,” Markus said. “Data scientists that have that level of knowledge are a bit like unicorns. So where do you find that right set of unicorns to help you solve the problem? That is a challenge.”

If you’re looking for something that has a low barrier to entry, generative AI might be the ticket. Generative AI is a type of artificial intelligence technology that can produce various kinds of content, including text, imagery, audio, and synthetic data. 

OpenAI, a private AI research company, allows you to use some generative AI for free. For example, GPT-3.5 has advanced the progress of natural language processing (NLP), which is how programs interpret spoken and written human language. Because business is becoming more data-intensive through scale and technologies like the Internet of Things (IoT), there is a need to better process these massive volumes of data in a way that’s consumable and action-oriented. The increasingly efficient advancements of NLP in AI enable leaders to keep pace with data proliferation. And it’s getting even better. More advanced versions such as GPT-4 are available to paid subscribers. 

“I do think with generative AI it can become easier,” Markus said. “About 95% of Generative AI appropriate use cases can be solved with a traditional generic large language model such as GPT-4 from OpenAI, Gemini from Google, or even the leading open-source LLMs. Those solve a tremendous number of use cases.” 

But generative AI can stumble when you get into company-specific questions and answers. There’s even a term for it: hallucinating. “AI hallucinations” are false or misleading information presented as fact. While it’s clear that the benefits of AI are many, hallucinations and the associated risks are included in the challenges that are giving some decision makers pause.

While public domain AI like Chat GPT-4 may be a reasonable entry point, it’s important to have best-in-class techniques such as Retrieval augmented generation (RAG), API driven data connectors, fine-tuning processes, accuracy monitoring and even custom LLM development on your roadmap. This mitigates the risk of the 30% inaccuracy found in using public models on proprietary information driven use cases.

What does the AT&T architecture for AI look like? We use a foundation of open source and commercial generative AI to build a private, proprietary, ever-evolving Generative AI ecosystem that learns through the data it consumes. How does that work?

Let’s start by clarifying some terminology. RAG is an AI framework. Its purpose is to improve the accuracy of a Large Language Model (LLM) through intentional focus. An LLM uses deep learning techniques referred to as transformers to understand information better through enormous amounts of curated data called a knowledge corpus. And deep learning simulates how humans think to perform tasks with increasing accuracy as it learns. This is what AT&T relies on to power our use of Generative AI.

“We use Azure OpenAI Services as well as open source and commercial offerings from companies we collaborate with to create the core LLM interoperability.  We layer that with other GenAI enhancing functionality that uses retrieval augmented generation (RAG) methodology that helps generative AI become really smart and hallucinate a lot less on AT&T information,” Markus said. 

When generative AI is optimized for specific company uses, the data it learns from can be more targeted, which yields very accurate results.

Markus continued, “We met with Open AI recently in San Francisco. Their products are an important part of Ask AT&T and we’re very encouraged on the future promise of their product set. That said, it’s paramount to keep the architecture interoperable. We have access to many open source solutions as well as other commercial products and often use more than one LLM in executing a given use case.”

AT&T built “Ask AT&T,” a generative AI tool for its employees and AT&T use cases. Ask AT&T is helping coders and software developers be more productive. It’s also helping to translate customer and employee documentation from English to other languages. We’re training it on our contracts and financial materials, it assists with our human resources questions, helping to make our customer service agents even more efficient, and we’re training it to aid in identifying and fighting off fraud attempts. Learn more about “Ask AT&T” here.

Needless to say, taking full advantage of AI requires solid preparation, as well as a reliable network-based foundation. Part of our success in our AI architecture is because our foundation is on the reliable, secure AT&T network. With this foundation in place, we can focus on the architecture.

“You have to invest in the right architecture, and I think one thing that's really important is to not only invest in the architecture but to align around architecture,” Markus said. “One of the things a lot of companies have in the AI space is what I would call the Wild West. It’s not that they aren’t doing a few good things, it’s that they’re doing a few good things in a disconnected way that doesn’t lend itself for reuse and expansion.” 

Can any business of any size benefit from AI? The simple answer is yes. It can help everyone from the business owner who operates from their kitchen table to Fortune 500 companies.

“I think it’s for every person, every business, small, medium, or large,” Markus said. “When I talk about a game changer, it will change the way we work, and we live in a way that’s hard to even imagine.”

Business leaders have begun the journey. Forbes Advisor commissioned a study of 600 American business owners who currently use AI or plan to incorporate AI in the next 6 months. The study, conducted by OnePoll, shows that the most popular applications include customer service (56%), cybersecurity and fraud management (51%), digital personal assistants (47%), customer relationship management (46%), inventory management (40%), and content production (35%).¹

And in specific industries, AI is transforming business operations through agility, efficiencies, and revolutionary positives for end beneficiaries like consumers and patients. Three of the industries being positively impacted by AI adoption are healthcare, manufacturing, and transportation.

Healthcare – The use of AI in healthcare is undoubtedly revolutionary. Use cases include augmented care, the discovery of new medications, disease diagnosis and prognosis, and many other benefits on the horizon. Because of the opportunities, much research is focused on exploring the possibilities in healthcare through AI. For example, one study The National Institute of Health is funding is the development of smart clothing to reduce lower back pain.²

Manufacturing – The integration of AI into manufacturing is enabling Industry 5.0 on the factory floor. It’s a natural progression from Industry 4.0, which was defined by Internet of Things (IoT) devices. Although IoT data in manufacturing tends to be localized and specific to operations, it may be possible to use the data that’s available to build an AI-based architecture that’s specific to your company’s needs. Your factory operations might become leaner, more agile, more efficient, and cost-effective.

Transportation – For over-the-road and other modes of commercial transportation, AI is helping to improve safety. In vehicles, telematics systems use IoT sensor data and AI algorithms to analyze vehicle data, driver performance, and maintenance needs. The result can be lower costs through predictive maintenance that’s even more efficient than is realized by IoT alone, better safety for drivers and the public, and lower costs.

Changes in the workforce are helping to fuel the rapid evolution and adoption of AI. These include the shortage of skilled workers and the expansion of the hybrid workforce. And, since consumers continue to demand better, more personalized experiences and products, companies are feeling pressure to be innovative in how they deliver quality to their customers.

AI is allowing companies to become more agile and productive, which leaves more room for innovation and to make smarter decisions in addressing customer needs. In fact, 64% of business owners think AI will increase productivity.³

As AI becomes more commonplace in business, Markus says, “We’ll have a research assistant at our fingertips. Eventually, we’ll have a personal assistant that can help us with everything that we do digitally. Writing emails, creating PowerPoints with just a few prompts, doing our shopping for us, curating our music taste, finding the social content that we really want to see.”

Building a business case to gain support in your organization for an AI architecture is essential to AI adoption. Because of the broad implications of AI, having consensus across key stakeholders may make it easier to see where AI can have the most impact

“Even before you start solving the technical problems, you have to have the right providers for the business,” Markus said. “Once you’ve identified those who are willing to help your business implement meaningful change, who are willing to think about things differently with AI, who are willing to put these great ideas to work, then everything changes.”

Consulting third-party subject-matter experts can provide you with an objective point of view and an objective assessment about how ready you are for AI adoption, where it can make the most sense in your business, and how to get started. From there, you can roadmap your plan and continue to work with them to innovate for the specific AI-related needs of your business.

“Make sure your subject matter experts are involved, that this isn’t just a science project, that they have a legitimate seat at the table. Then you’ve got to hire the best talent you can, take all this incredible technology and make it work for the given company. Because what works for AT&T may or may not be exactly what’s needed for say a large manufacturer or a small B2C company.”

One of the reasons that AI seems overwhelming to businesses is because its development and how it can be used is rapidly evolving.

“Every week something kind of monumental changes with AI—especially with generative AI,” Markus said. “Having providers who can grow with you, who can adapt to these changes, who will be on the leading edge so that you can take advantage of it is very key.”

Artificial intelligence is data intensive. It requires fast access and speedy transfer of data; in other words, a modernized network built with fiber connectivity and security. An outdated network (often built on copper) with slow speeds, high latency, or unreliable connectivity reduces processing times. What’s more, most AI applications need to provide real-time or near-real-time data. Reliable, secure, high-speed connectivity is a must.

To move forward, it’s important to go beyond the buzzwords and rumors. Begin to dive deeper into educating yourself and your key stakeholders with AI and how it works. That basic understanding can open the door to AI possibilities in your business. Next, conduct an audit to see where it may help the most. From there, lay a roadmap for AI adoption. As we’ve covered, the topic of AI can seem overwhelming, however businesses of all sizes can benefit from it and that first step may mean using a public domain option. 

“I’ve rarely seen technology that has this much potential,” Markus said. AI, compared to the introduction of the internet and PC, “has the potential to have as much impact on how we work, how we live, and how we just do things, as those huge changes.”

See how ultra-fast, reliable fiber and 5G connectivity protected by built-in security give you a new level of confidence in the possibilities of your network. Let our experts work with you to solve your challenges and accelerate outcomes. Your business deserves the AT&T Business difference—a new standard for networking.

Learn more about AT&T Business solutions at business.att.com. And to connect with an expert who knows business and can help advise you how to set your foundation for AI adoption, contact your AT&T Business representative.

¹Katherine Haan, “How Businesses Are Using Artificial Intelligence in 2023,” Forbes, April 24, 2023, https://www.forbes.com/advisor/business/software/ai-in-business/

²“Artificial Intelligence and Medical Research.” National Institute of Health. January 2023. https://newsinhealth.nih.gov/2023/01/artificial-intelligence-medical-research.

³Katherine Haan, “How Businesses Are Using Artificial Intelligence in 2023,” Forbes, April 24, 2023, https://www.forbes.com/advisor/business/software/ai-in-business/