5 Key Lessons Learned for Maximizing the Value of AI in the Enterprise

AI Has Already Radically Changed Enterprise Thinking

AI has captured imaginations, investments, and headlines around the world, from boardrooms to corporate strategy sessions to the popular media. It has already affected the way we do business, how we shop for goods and services, and the way we live our daily lives. Earnix and its customers have successfully and cost-effectively applied AI in the enterprise and in hundreds of other situations to solve some of the most pressing and difficult problems in insurance and banking.

AI capabilities that were “lab experiments” as recently as five or ten years ago have now made their way into mainstream production systems, and SaaS/cloud deployment and other technology advancements are allowing AI-driven technology to reach the masses at scale:

• According to Gartner, AI is being used by 40% of businesses to improve their customer service outcomes

• McKinsey reports that 70% of executives say that AI has already had a positive impact on their businesses

• And, according to the tech review blog TechJury, 35% of businesses are using AI today

AI Will Continue to Grow Exponentially

And that’s just the beginning:

• Per that same TechJury survey, 42% of companies are exploring AI for its implementation in the future – that leaves relatively few organizations that haven’t at least begun the journey

• The World Economic Forum projects that AI could create 95 million new jobs by 2025

• And, IDC estimates that AI and analytics are expected to contribute $14.2 trillion to the global economy by 2030

Capitalizing on AI in the Enterprise

With all the positive AI outcomes and its seemingly endless potential, there are several questions that naturally arise:

• “How do I and my enterprise maximize the potential benefits of AI?”

• “How do I capture market demand and market feedback with my use of AI?”

• “What lessons can I learn from others’ wrong turns?”

• “How do we achieve rapid innovation and quick time to market, while ensuring quality and dependable functionality?”

• “What data do I already have that’s useful, how accurate is it, and how do I make sure it stays ‘fresh’?”

• “What other data sources might I need to acquire? Where does that data reside? How accessible is it?”

• “And what about some of the challenges that AI poses – potential bias and other ethical concerns, large volumes of personal data being tempting cyber-attack targets, regulations such as GDPR, lack of transparency, and the massive complexity in some AI systems?”

All are certainly good questions. If you’ve started down the AI road, you may have answered some and still are looking for answers to others (it’s doubtful anyone has all the answers at this point in AI’s adoption). If you’re early in the process of utilizing AI, all those questions and more may be swirling in your mind.

To help with the journey, wherever you are now, here are five lessons Earnix and its customers have learned, and that we suggest be part of your thought process in every AI-driven project you undertake:

Lesson #1 - It’s Not About the Model

While building new models and seeing where they lead can be thought-provoking, fun, and exciting, don’t let that sense of wonder become a goal in and of itself.

Like most things in the enterprise, the key to success is to understand the business challenge clearly and precisely, and to establish the desired outcome before starting down the road. As in a quote attributed to Yogi Berra, “If you don't know where you are going, you might wind up someplace else.”

Are we trying to increase revenue for a particular product line, and by how much? What improvement in time to market can we achieve with AI? Is there a target for increased customer satisfaction? Can we use AI for internal purposes, such as increasing employee engagement?

Establish the answers to those questions first, then apply AI and develop the models. This is not to say we’re advocating a return to the “waterfall” world. But even in an agile environment, and even if you’re aiming for a minimally-viable product (MVP) in the early going, establish achievable goals and document them in a transparent way.

Also, remember that no model is an island – always keep in mind that there are existing enterprise systems that must be integrated with, and concepts such as API-driven development can help capitalize on previous efforts, investments, and solutions. Not every problem, AI-driven or not, needs to be solved from scratch. Also, not all existing workflows are candidates for replacement – the current workflow might be on target, and AI can make it better, make it deliver more insight, or operate with more efficiency, but be careful not to “throw the baby out with the bath water.”

And, finally, choose solutions and platforms that are in line with your in-house expertise. Even leading-edge, advanced technology is useless without someone in your operation who can put it to its most effective use.

Lesson #2 - It’s All About Your Data

Many organizations begin the search for an AI “magic bullet” outside the enterprise, whether by accessing expert analysts and consultants, hiring in experienced AI practitioners, or by acquiring access to third-party data sources.

While all these are potentially-viable routes to success, don’t overlook the most unique and invaluable asset you have – your data. You’re sitting on perhaps the greatest source of competitive advantage – make sure you use it to its fullest!

When considering industry-leading applications such as real-time rating and pricing, quoting, underwriting, and new product introduction, your data is a key ingredient in gaining new customers and market share. Upselling and cross-selling existing customers relies almost exclusively on data about their transaction histories, payments to other institutions, late payments, and the like.

Your customers won’t reward you for a series of “me-too” solutions, and a way to separate yourself from the pack is to leverage the data at your disposal to power unique capabilities. Invest in the ability to leverage the data you have, ways to automate its collection, and don’t skimp on data maintenance and updating – data is a bit like produce, it’s gets old quickly.

Balance the desire for speed and innovation with compliance resources in privacy, policy, auditing, and updating, to help avoid those issues with bias, cyber-security, and regulation, and to maximize the benefits in your data.

Lesson #3 - Embrace the Open Community

This one can be a “tough sell” in some larger, more traditional organizations, and in conservative industries such as insurance and banking. But the risks are declining every day, and the rewards are increasing at a commensurate or faster rate. It is possible to utilize open-source code and techniques and remain compliant.

The rewards can be immense. The open-source community is the fastest, most accurate innovator in AI, buzzing with activity and innovation 24/7/365, where data usage and techniques are continuously evolving at a rate no individual enterprise can keep up with. You cannot hope to hire or train enough internal resources to beat the community in solving the next challenge.

If necessary, start with a small, “dip your toe in the water” project or digestible MVP, and prove to yourself and your organization that using the community is the way to go. Follow-on enterprise-scale projects will benefit immensely from the experience, and from the sense of familiarity and security that early project will afford.

Lesson #4 – Build for Scale from Day One

When those “real” projects get underway, establish a plan for scaling to enterprise levels from the get-go, even if you start with a pilot or beta to prove the concept and refine the functionality over time.

To yield the maximum results, big data requires safe, large-scale, streaming solutions, and to reach large audiences (that global group of new customers you’re after, for example), standardized model serving is an absolute necessity. Don’t think small - think big right from the conceptual stage.

Most financial services firms have embraced cloud computing, which delivers on the scaling and updating needs of rapidly-evolving AI data sets and models. Protection from cyber attacks is another benefit of SaaS/cloud deployment, as in many cases global cloud providers have the resources and know-how to protect their networks better than do individual enterprises.

Lesson #5 – Use AI Responsibly – Plan Accordingly

“With great power comes great responsibility.” This old saw is especially true of wielding the power of AI. It’s critically important to your customers, your shareholders, your team, the industry, and the wider community.

With their speed and power, AI models can move faster than humans can keep up, and become “black boxes” that yield unintended consequences. The decisions AI models make are not always intelligible to humans, leading to a lack of transparency. Decisions rendered can be inaccurate and/or discriminatory, and can result in embedded or inserted bias. All these outcomes can run afoul of regulators, undermine internal confidence, and expose the enterprise to significant public relations and brand damage.

Just as you should plan and execute for operation at scale, you must take the responsibility to plan and execute for the responsible use of AI. A separate group or function should be established to monitor your AI models for bias, and take steps to mitigate it immediately if detected.

AI can be a double-edged sword – it can lead to safer, faster, more defensible decision-making, but effort must be taken to ensure those are the demonstrable and sustainable outcomes.

Summing Up AI in the Enterprise

AI can deliver massive positive impact to your business, and in a very short period of time. New insights can come fast and furious. You can effect change that will propel the business (and your career) forward. Your technical team can be motivated, challenged, and excited by working with, and growing with, the latest AI tools and technology.

To maximize the business impact of AI, and to maintain focus on what truly matters, keep in mind the five lessons outlined above, and always ask yourself “What’s the problem I’m solving for?”

Uncovering unexpected correlations, building new models, and driving innovation are all part of the equation, but all must be part of a greater whole – delivering enterprise value.

To Learn More

This topic was the subject of a recent presentation by Erez Barak, Earnix Chief Technology Officer (CTO), at the Excelerate 2023 customer conference in London. You can view his presentation in its entirety here, and review all of the presentations from the conference here.