Top Ten Essentials for Enterprise Ready AI

Most of us working with information technology are accustomed to using the term ‘enterprise ready’ to describe the essentials needed for enterprise software applications, but the term is applied more broadly for all types of products, systems, and services. 

'Enterprise ready' generally means containing the essentials necessary to meet the adoption bar, or in other words the rigor within large organizations to determine whether the product or system (AI in this case) meets at least the minimum requirements for compliance, functionality, usability, security, and risk. 

Although most large enterprises invest heavily in internal and external innovation, they tend to be quite conservative when it comes to adoption of new technologies, especially when changing how ‘things have always been done’. Large organization cultures become risk averse over time in part due to their significant impact on customers, investors, and society. 

This conservative approach to adoption of important new technologies like EAI creates special challenges for internal and external innovators, as well as CXOs in organizations, captured well by the innovator’s dilemma coined and defined by Clayton Christensen. Conservative adoption can also frustrate ‘change agents’ within the organization charged with innovation, or employees more broadly who often feel they need to go around existing channels to get their work done and remain competitive, which can cause serious problems.

The rapid increase of AI across the enterprise has created an especially challenging situation for senior managers of large organizations, particularly for those companies at high risk of being disrupted or displaced, whether by startup or increasingly an established company. The recent record adoption rate of LLM chatbots by individuals, and the response by regulators, has raised the question of enterprise ready AI in boardrooms at organizations of all sizes.  

Data structure and governance plays a leading role in EAI architecture, ultimately represented by the board, often via the CDO. Depending on the organization, the full cast of characters who influence EAI include those charged with operations, supply chain, compliance, liability, cybersecurity, sales, marketing, productivity, ethics, safety, human resources, and sustainability, among others. Unions are also beginning to play a much larger role. It’s safe to say that no organization was ‘EAI ready’ in 2022, and very few can be considered so today, if any, so let’s take a look at what it requires. 

In a perfect world, enterprise architects (EA) would lead EAI architecture, but any seasoned EA will tell us they live in a world far from perfection. Every journey begins with assumptions, which should be based on the most relevant information available to us. Our assumptions are better than most due to 26 years of R&D, hundreds of discussions, and dozens of deep dives.

System design

Whether adopting a third-party system like our KOS (EAI OS), which I obviously think is wise as we have invested decades of work on it, developing a custom in-house EAI system from scratch, or in most cases a combination of the two, the goal should be a unified, cohesive, seamless, and highly competitive EAI system across the entire organization and ecosystem. Anything less is unlikely to be competitive within a short period of time. 

It's been my career-long observation that the best products and systems are usually designed and built by exceptional, passionate, and talented individuals, or small teams with a lead designer. Jony Ive and the iPhone is a classic example of a lead designer who worked closely with Steve Jobs, but rather than a universal consumer device customizable to an individual’s needs, an EAI system must do similarly for an entire organization, including each individual, team, and business unit within the organization.  

Another example was Microsoft Word and Office, which came to me primarily through my old late friend and early partner in KYield, Russell Borland, who was an early employee at Microsoft, key contributor to Word, and an early product developer and manager, before moving over to MS Press to write more than a dozen books about Microsoft products. While the technologies and challenges are different today, the complexity and process have similarities. 

Several very large organizations we’ve engaged with over the past few years have far too many chefs in the kitchen to produce a satisfying meal in EAI systems. While some can create masterful side dishes, it’s unlikely the meal will result in a well-choreographed performance that delivers a unified, cohesive, and efficient system, much less competitive.

In defense of even our direct competitors within a few large enterprises attempting to build something similar to our KOS, EAI system design should be as free from bureaucracy, turf wars, and conflicts as possible. That’s very difficult to do inside large organizations that have many conflicting needs. 

“The hammer is about to come down on some of the world’s biggest tech companies”

With the new rules on AI and data in the EU, and other countries working on regulation, many companies are now in a rush to comply, including tech giants, which reveals their lack of enterprise readiness. A recent article in the WSJ by Kim Mackrael and Sam Schechner suggests that “The hammer is about to come down on some of the world’s biggest tech companies”. 

“This is a Glass-Steagall moment for big tech,” said Brian Wieser, a tech analyst and former investment banker, referring to the Depression-era law that reined in banks. “They’re going from effectively no regulation to heavy regulation.”

It couldn’t be clearer that consumer LLM chatbots were not enterprise-ready, safety-ready, or consumer-ready, nor were they likely compliant, yet they are being integrated as rapidly as possible into some of the largest enterprise software products. This is the reality facing enterprise decision makers.

It is much simpler and wiser to design-in governance and security from inception. Attempting to do so once scaled is very challenging and not very effective, unless they start over from scratch with new and better architecture, including data stores free of copyrighted content, which would greatly reduce the value of the consumer chatbots. 

Bottom line: It’s up to each organization to ensure their EAI systems contain proper governance, security, and compliance. Fortunately for KOS customers, we did so from inception. So far it appears the KOS is compliant with the EU regulations as well as proposed rules being discussed in the U.S. 

Top 10 essentials for EAI readiness:

1.  Enterprise-wide governance. Every EAI system should have strong governance so the organization can manage the system to pursue goals, achieve compliance, manage corporate policies, and deliver operating principles (see our EAI principles). A well-designed governance system (CKO Engine in the KOS) should enable simple-to-use admin over the entire EAI system. 

2.  Safety and security, including protection, ownership, and control of the organization’s data. We offer multiple types of security in the KOS. 

3.  Functional applications. We call them functions in the KOS, but they are really data-centric software applications that integrate and run ML algorithms. The highest value functions in the KOS run in DANA (digital assistant for each individual), including personalized learning, prevention, and productivity enhancement (we recently introduced the GAI function).

4.  Accuracy. Most organizations require a very high rate of accuracy, which is a serious problem with consumer LLM chatbots (see our recent exec briefing and video on how to provide safe and accurate generative AI).

5.  Compliance, including legal, regulatory, and corporate policies. Many companies are currently scrambling to prepare for new rules while others are worried that if they act too soon they’ll need to rebuild after the U.S. finally decides what to do about regulation. 

6.  Scalability was one of the early challenges in EAI, discussed for example by Vint Cerf and myself about two decades ago via email. Fortunately, our systems are much improved today as scalability across the ecosystem will be increasingly necessary (via DANA’s network function in the KOS).

7.  Sovereignty protection. I think this situation will improve over time for a variety of reasons, including more functions run at the ‘edge’, on personal devices, and more hosting options, but in the near-term lock-in to top cloud providers represents a serious risk for enterprise customers. 

8.  Intuitive UI and simple to use. Since EAI systems are run on data, and the people and organization are the source of much of the high-quality enterprise data, intuitive design and simplicity are critical. 

9.  Interoperability. Software APIs help a great deal, but precision data management requires interoperable data as well. We structure data in the KOS so that it can be moved easily and reduce dependencies on any single database or cloud provider. 

10.  Popularity. High-quality inputs create high-quality outputs, and people produce the high-quality data, so it’s essential the EAI system is the preferred system used by people. Our digital assistant DANA is provided to every employee in the organization and can be extended to partners and end customers. We designed DANA with the intention for the digital assistant to be the preferred app for knowledge work. 

Leadership

Given the importance of EAI to the modern organization, in particular businesses and some types of government entities, I have consistently recommended that the ‘EAI architect’ should be the CEO of the organization. Our KOS, for example, is enterprise-wide by necessity to capture and deliver the most value for productivity, risk management, prevention, and personalized learning. Physics requires enterprise-wide use to capture the most value in critical functions, which I’m convinced will be necessary to remain competitive moving forward. 

The KOS impacts every aspect of the organization, and only the CEO has that responsibility from a legal and board governance perspective. He/she (CEO) must delegate but should learn enough about EAI to be a functional champion and lead the effort, including direct reporting from the team tasked to execute the strategy, typically including the CIO, CDO, CTO, COO, EAs or others. 

A few companies now have Chief AI Officers, but what’s most important is the leader is respected, the team is highly competent, not suffering from self-destructive bias like NIH syndrome, they lack unhealthy relationships with vendors, have strong support from the CEO and board, and the teams executing the strategy have the skills and talent to develop a competitive system. 

Business unit leaders that benefit most from EAI are often the initial champions, and when operationalizing and scaling should lead the initial project team, but the CEO (and CDO/CIO) should still sit in on regular updates, providing the enterprise-wide strategy. Adopting EAI is a transformative process with enormous implications for the future of the company, and includes a steep learning curve for all concerned. As CEO I would certainly want to remain on top of it through the process.

Will the AI revolution lead to the return of creative destruction, which has been in a sharp four-decade decline, mass extinction of incumbents as some have suggested, or will AI reinforce market power and accelerate market consolidation and monopolies?  The outcome depends on our individual and collective decisions and actions, not least regarding regulation and the adoption of EAI systems.