Pragmatic AI
Considerations for Leaders to Maximize AI’s Promise and Avoid Repeating Mistakes from the Past
November 28, 2023 | Briefing
Artificial intelligence (AI) has dominated headlines over the past decade. Article after article and industry expert after industry expert have hammered home its potential for revolutionary world-changing impact.
The accelerating pace of improvements in AI-enabling tools has shocked even the most optimistic technology proponents. Optimism about the potential of AI to unlock future profits has been powering the tech-heavy Nasdaq to significant gains.
At the same time, there are growing concerns about potential risks and dangers with AI. These include worries about automation-spurred job losses, deepfakes, social manipulation, privacy violations, financial market volatility, and uncontrollable self-aware AI.
We now face a tug-of-war between emerging restrictions on AI development in the name of privacy and individual safety – and the experimental freedom others say is needed to unlock its full promise.
Is the intense media, promotional, and regulatory focus on AI appropriate? Yes, to varying degrees. Should every business leader be considering AI’s potential uses across their organizations? Absolutely! But we’ve been down similar roads before with the emergence of the commercial Internet.
This leadership briefing will lay out a few core considerations for AI decision-making. We’ll examine what can be learned from prior technology adoption cycles as the commercial Internet emerged. We’ll look at a few examples from organizations implementing AI now. And we’ll suggest considerations for leaders seeking to maximize their AI investments while mitigating potential risks.
The Emergence of the Commercial Internet and Fear of Missing Out (FOMO)
Internet commercialization started in the mid-1990s. Cheerleaders in the venture capital and private equity sectors effectively drove market hype. An enthusiastic buzz on the Internet dominated business communities. The dot-com investment bubble seemed to validate the wisdom of any Internet investment, no matter how far-fetched.
CEOs at that time feared being left behind by their competitors. They could sense but not put their finger on the specifics of the Internet’s transformational nature. Their intuition told them the Web would foster significant opportunities to change customer behavior and business-to-business relationships.
However, these business leaders often launched Internet-enabled projects without establishing clear business cases. Instead, they prioritized their goals on brand building and networking alone. They believed that by capturing enough “eyeballs,” they would eventually achieve success and figure out profitability later.
A few companies, such as Amazon and eBay, survived the speculative bubble and have become highly successful. But in most cases, tangible returns on Internet-initiative investments in those early years were elusive.
IBM’s Lou Gerstner confronted this widely adopted approach in his 1995 COMDEX keynote. That theme can be summed up succinctly. When executives consider investing in Internet initiatives for their companies, “It’s about business, not technology.”
Today, in the words of the great philosopher Yogi Berra, “It’s déjà vu all over again.” An AI frenzy pervades. VC and PE firms are championing AI mantras. Technology stocks are booming, powered by investor AI optimism. AI features are popping up in applications across much of the software-as-a-service (SAS) spectrum.
Accordingly, we suggest a note of caution. Business leaders should anchor their efforts when adopting AI with clear business objectives that justify their investments. Set realistic expectations, intermediate business goals, and rational budgets, even if your AI efforts are only experimental. Drive your decisions by recognizing that AI is an enabler but not an end unto itself.
An AI Case Study – Healthcare
A sizeable medical specialty practice that we work with values its ability to thrive independently. They constantly seek options to enhance clinical productivity while maintaining their superior patient experience reputation.
This medical practice is considering an AI system that automatically generates clinical notes from patient-clinician conversations. The business goal is to reduce clinicians’ time on documentation – freeing up more time for patient engagement.
While we are all excited about this AI technology’s promise, we are cautiously moving forward. Although this system is HIPAA compliant, some patients may not be comfortable with having their conversations recorded. It will also take time to see how effectively the system will parse those conversations into appropriate clinical notes.
This system is designed so clinicians review, edit, and approve the notes before they are finalized. Over time, the system’s deep learning algorithms should help it achieve significantly greater accuracy when generating the notes. Time will tell.
Why is implementing this system part of our marketing focus? Because providing distinctive patient experiences is interwoven into medical practice growth. Extrapolating from a recent McKinsey analysis on customer experience, or in this case, patient experience:
- One lost patient can require adding three new patients to compensate for lost value.
- Eighty percent of the value creation achieved by the most successful growth organizations comes from their existing core businesses – in this case, unlocking continuing revenues from the patients they already have.
An AI Case Study – Business to Business (B2B)
Marketing automation has been used for years. It can help automate repetitive campaign tasks and enable more productive engagement across multiple channels.
The problem is that the messaging used in marketing automation campaigns is usually generic. Marketing targets are often treated as one homogenous group.
This approach proved to be ineffective for one of our B2B clients. Their customer base is diverse, each with different needs across this client’s spectrum of services. Marketing messages and special offers relevant to one sub-segment fell flat or were counterproductive with another.
The short-term fix was for their sales team to customize marketing communications individually. However, this burden took away from the team’s deal-closing priorities. It didn’t take advantage of the significant scale of activity that marketing automation typically provides.
Our answer was to repower their marketing automation with AI components that target each sub-segment according to specific needs.
We leverage information related to purchase histories, company and individual preferences, regional and local idiosyncrasies, and market segment trends. We factor in historical and predicted personal preferences to engage – or not – through specific channels. AI evaluates that – and more – to personalize marketing messaging and suggest and automate what is delivered across multiple channels.
A Pragmatic AI Path Forward for Leaders
In the short period since late 2022, AI has transitioned from a topic primarily handled by tech employees to a significant focus for company leaders. At the same time, governments are acting on growing calls for regulation that may make AI implementation more complicated for decision-makers.
Moreover, we are still in the very early stages of AI. Most organizations are just starting to sort through what AI means to them. It will be a bumpy road for quite some time as AI capabilities mature and evolve.
Against this backdrop, four recommendations for leaders to maximize their AI investments while mitigating risks follow.
1. Select Your AI Enablers Carefully
IT providers are highly motivated by the emergence of AI. From the software industry perspective, AI presents an excellent opportunity for them to break out of their current revenue slump. For IT startup founders, AI provides an anchor to establish a new generation of services against exploding market demand.
We know what to expect from past early-stage tech development cycles. Some of these AI providers will emerge as leaders in the field, some will enjoy moderate success, some will survive but limp along, and many others will fail. It is a market comprising providers without long-term established AI track records.
You should carefully choose who you decide to work with for AI. Evaluating risks, reliability, and the impact on your organization requires you to think beyond technical considerations alone.
Even if you use AI components currently offered for free, implementing AI involves significant financial and opportunity costs. Unwinding an implementation would also involve additional burdens if it doesn’t work out.
This also means identifying an additional layer beyond a provider offering you an AI solution in many cases. Many providers power their solutions with AI technologies from others. You need to know the identity of those others. Clarifying the quality, trustworthiness, and longevity potential of the AI enablers you’ll depend on is an essential first step for evaluating your best options.
2. Understand the Benefits and Risks Between Public and Private AI
There are two categories of AI-enabling resources: public and private. Each has its pros and cons.
Public AI enablers are services, programs, and algorithms that are available to the public at large. Many public AI resources are free to anyone who wants to use them. They can be used as is or modified to serve a user’s specific needs. Public AI enablers are already integrated into many search engines, social media platforms, software, and other online tools.
ChatGPT is an example of a public AI enabler. It is a freely available chatbot developed by OpenAI that enables users to engage in conversations, gain insights, and automate tasks.
A significant concern about public AI enablers are potential risks with data handling, privacy, and sensitive business information exposure. AI systems harvest a considerable amount of user data to learn and improve.
Given the lack of control of that data, some organizations have banned using generative public AI systems when using company hardware and networks. One fear is that employees may reference trade secrets or client information when querying public AI. Or closely guarded proprietary business practices may be exposed.
Private AI enablers are trained in-house with proprietary data for an organization’s specific needs. The data behind these systems can be directly controlled. This minimizes the risk of data breaches or unauthorized access. It also enables the data to be fine-tuned so it doesn’t reflect the biases in some public AI models.
Private AI is generally more costly to implement than public AI. However, some vendors offer private AI platforms. These platforms provide data security advantages while not requiring the costs of a complete AI infrastructure and staff to be built in-house.
3. Be Vigilant with AI Data and Content Legal Issues
Most organizations are aware of their current legal obligations concerning personally identifiable information. These already apply, to varying degrees, to the use of AI.
A patchwork of existing federal, state, local, and international laws add complexity to this challenge. Healthcare providers also must deal with HIPAA, and there is an extensive list of industry-specific regulations governing the handling of sensitive data.
There are also AI-specific regulatory requirements that are currently in discussion and emerging on different political levels. Leaders must make themselves aware of the concerns driving the AI debates. That way, they can proactively position their organizations appropriately. Perhaps more importantly, leaders can consider societal cues on AI and ensure the customer experiences they provide adapt.
One area of focus should include situations where you are using content that AI has completely or significantly created or edited. This potentially may become a problem when AI-generated content is based on the copyrighted work of a third party.
What specific actions should you take to be vigilant about data and content legal issues? Disclose the use of AI-generated content to mitigate potential legal and reputational liabilities. Monitor how AI content is created in your organization and set appropriate internal policies. And seek legal indemnity for copyright issues with content from any generative AI vendors you engage.
4. Be Realistic About Your Expectations for AI
Artificial intelligence (AI) is a term that encompasses many technologies. Each of these technologies has different uses and reflects different innovations with potential transformative impacts.
Gartner breaks this AI landscape into 29 technologies on their AI hype cycle, with none yet considered mainstream. This research firm projects only three – computer vision, data labeling and annotation, and edgeAI – will reach productive mainstream adoption within the next two years. They also forecast that generative AI and decision intelligence fall within a 2- to 5-year probable mainstream window.
However, we know from past new technology cycles that part of the market will adopt early. Innovators will overlook any glitches and product limitations in an early-stage AI service offering if they can be on the cutting edge.
Early adopters follow the innovators. They are highly motivated visionaries looking for breakthrough technology to help transform their organization. However, they may pay a premium for AI offerings compared to expected costs once the market becomes mainstream.
Early majority purchasers will wait until the market matures. They are risk-averse pragmatists who want the most significant issues in an AI offering to be fixed before moving ahead. They are also traditionally cost-sensitive with their technology purchases.
The innovator and early-adopter leaders implementing an AI offering in the early market may capture first-mover competitive advantages. They should recognize, however, that their initial technology experience may be more buggy, costly, and risky than if they had waited until later.
We’ll finish with one final matter to look out for with generative AI.
Current generative AI systems are prone to what is known as “hallucinations.” Simply put, the output from these systems is sometimes inaccurate or wrong. Accordingly, generative AI initiatives continually require human review to catch issues that may harm your reputation and customer experience. If you opt to deploy generative AI, you must devote sufficient resources to monitor it and protect your brand actively.
Wrapping It Up
For technology providers of all levels, shapes, and sizes, AI represents their key to salvation. The sales pressure on leaders and decision-makers to try the next best AI-enabled thing is building and won’t let up. The challenge for leaders will be navigating optimal paths for their organizations when using AI, particularly in the early stages of the developing landscape.
About the Author
Ted Bream is the Managing Partner of Core Agenda, a consulting firm specializing in fractional chief marketing officer + services. Ted’s digital experience began with leading the development of some of the first web-based initiatives at the dawn of the commercial Internet. He’s had executive, founder, partner, and consulting roles across dozens of digitally enabled ventures ever since. Ted has been an executive at IBM, Gartner, Scient (one of the first large Internet consulting companies), as well as a myriad of technology-enabled early-stage ventures.
Ted is known as an adaptive leader, innovator, market maker, and customer experience evangelist in growth marketing and digital transformation. He has an extensive history of helping clients discover opportunities and achieve standout results, particularly with challenges across the healthcare and B2B segments. Ted is a passionate and compelling speaker on a variety of digital transformation topics.
When not focused on clients and innovations in marketing-led growth, Ted enjoys spending time with his growing family and good friends, veterans causes, and a golf game that provides him far more joy than his elevated scores would indicate.
