Potential Of Innovation

Generative AI (Gen AI), with its ability to create new and realistic content (often indistinguishable from human authored content), holds immense potential across all industries. Anyone can go online and start generating ideas using Open AI's ChatGPT or any of the other Gen AI tools out there today. Every day more Gen AI tools are being deployed and developed.

To see a real-life example of the impact of Gen AI, look at presentation creation. Previously, the process of crafting presentations required considerable time and effort, often spanning hours or days. Now, Gen AI powered platforms can swiftly process source documents like meeting notes or call transcripts, distill key points, and construct professional presentations. And all of this can be done in a matter of minutes, significantly enhancing efficiency and time management. While productivity hacks like this are undoubtedly exciting and can be valuable, we shouldn't confuse them with true innovation. Real innovation is hard, and it takes more than a 'corporate innovation team' to achieve. It requires a willingness to look openly and honestly at your business and to embrace change; these values must be part of the DNA of an organization.

While there's no recipe or roadmap you can follow to make your organization embrace innovation, there are some key things to keep in mind as you attempt to innovate. And over time, through repeated application of effort, your organization may gradually develop a true innovation culture. But you have to start somewhere; these are lessons I've learned that can help you drive innovation at your organization.

Path To Innovation
(or Pick the Right Starting Point)

One of the most critical early decisions is to select the appropriate use case initially, avoiding overly ambitious goals. If your organization is new to using analytics then take your time and go slow. Consult genuine experts that understand data engineering, engineering principles, programming, analytics, organizational structure and most importantly, people. Engaging stakeholders to embrace the new technology will make or break success. Despite seeming straightforward, these principles are often overlooked, hence the need to emphasize them in this discussion.

Innovation, not Destruction

It's crucial to select project leaders with the right expertise. And while you may have an “innovation team” in your company, they may not be the best qualified to lead these efforts. Likewise, it is probably not your recently-hired developers with cutting-edge technical skills, but little institutional knowledge. Engage individuals that understand how your organization works, and know what has been tried in the past and why those efforts flourished or failed. The ideal leader for an innovation effort is humble and open to learning, and has battle scars from deploying complex, innovative technologies. This corporate memory is important, since your goal should not be to tear down all of the existing company and build anew. You're looking to evolve and improve your organization, not destroy it.

Engineering Expertise Matters

Those are the easy problems but when it comes to trying to solve the difficult problems confronting a modern business; engineering expertise is critical. If the CEO wants the organization to have its own custom language model for a specific business domain (e.g. think finance) that takes some serious engineering. Understanding, managing, joining and transforming corporate data from a plethora of various data architectures takes some serious engineering chops. Not exactly for the faint of heart. Depending on the complexity of corporate data, wrangling datasets into usable information can be a difficult task that involves many pipelines, large clusters of servers and serious system and domain knowledge.

Once the data has been wrangled into the right structure comes the training of the LLM which is a complex engineering task. There are simpler ways to train an LLM and get decent results using corporate data but if you want accurate output that requires real data engineering skills. Then there is the productionizing and maintenance of these models, none of which is trivial and requires talent that actually understands the nuances and complexities of this technology. Your developers will not be learning those skills on Coursera. These are hard earned skills acquired by individuals that love their craft and understand the challenges, borne through previous implementations, associated with engineering large extremely complex systems.

Moreover building these systems in a cloud environment without the proper controls and understanding can drive massive computing costs quickly. A former colleague worked with a developer that managed to spend $250k in cloud charges in one month. The developer wanted their data pipeline to go faster so they went ahead and allocated several thousand CPU cores on the largest VMs they could find. It sure went faster and they were almost out of a job. Talent matters.

Organizational Challenges
of Generative AI

Rarely has a technology this revolutionary been developed that will have such a broad and deep organizational and societal impact. Let me say that again - Gen AI will change how we perform modern business. It will take time to internalize the changes Gen AI will bring about but, it will undoubtedly have a deep impact on the organization and its culture. Organizationally you cannot have a passive wait and see approach to this technology. It is readily available, ubiquitous and simple to use. Most likely everyone in the firm is already using it in some form or another to make their jobs easier. Maybe something as straightforward as writing better email using ChatGPT or helping their design process with their own account on MidJourney but make no doubt they are using it.

Organizations must think about their talent pool and having the right people to enable this technology in a way that helps and not hinders business. There will be pushback from employees whose jobs are now at risk and they will do everything in their power to passively or actively inhibit its adoption. Corporate culture matters and most, especially small and mid-size, are not prepared to grapple with the complexity and changes that will be brought forth by Gen AI. The vast majority of mid-sized companies have little to no technology experience and now they are being forced to embrace AI. They are basically at ground zero with technology management and now they have to try and embrace one of the most complex technologies overnight.

Securing the appropriate talent is essential, particularly given the limited availability of individuals with the requisite experience in this field. There are many people out there that are selling their expertise who have limited to no experience. Education itself is not an indicator of depth of knowledge and in fact can be highly misleading. A PhD is not a prerequisite for deploying and using this technology but deep, broad experience is mandatory.

Digressing for a moment, many companies have moved towards centralized recruiting where small divisions within the organization have enormous control over who gets interviewed and hired. Previously managers were directly responsible for hiring their own teams and owning the search process. Although it is time consuming it generally yields better results since most managers would typically review non-standard candidates. I have reviewed thousands of resumes that a recruiter would have passed over and have always found fully qualified individuals in that reject list. A manager with experience can see skills in a candidate that are not explicit in a resume that central recruiting will miss.

Centralized recruitment processes have significantly complicated hiring procedures, leading to widespread frustration among managers who often find themselves exasperated with the system. If you want the best candidates, put control back into the hands of the managers. A manager with deep company experience has a higher probability of finding a candidate, even if they have skills that are different than a job description would have. More than ever this matters with hiring for AI jobs where the search needs to cut across domains and there are many, not at first obvious, candidates.

And lastly senior leaders need to have a deep understanding of the culture and personality of their organization. An honest, open review of the culture is not only necessary but a requirement. Is the culture closed off to outsiders? Do you easily embrace change? Are you setting up a outsider to fail? Are existing employees who are threatened intentionally or unintentionally sabotaging them? What is the average tenure of your employees? Chances are if your average employee tenure is long, change is going to be difficult. In order to facilitate change you need a brutally honest assessment of your corporate culture to make the right hires.

Change is going to happen whether you embrace it or not and it is better to be prepared and manage it.

Ethics and Responsibility

F. Scott Fitzgerald famously said, "The test of a first-rate intelligence is the ability to hold two opposed ideas in mind at the same time and still retain the ability to function." After sitting through numerous panels and lectures on ethics and responsible AI I am convinced that there are few that possess first-rate intelligence. Simplistic and apocalyptic arguments are in abundance otherwise the world is headed down a dark path. Rarely is the world quite so simple and it is full of complex nuance and everything has a balance. When building these models it is critical to try and achieve some level of balance that satisfies business needs without facilitating unethical decisions.

Many professionals engage in discussions about ethics in the context of Gen AI, yet their narratives often contain contradictions within the same statement, revealing a fundamental misunderstanding of how Gen AI functions. They often confuse AI and Gen AI treating them as the same. The comment that always entertains me, and strikes me as naive, is that many of these arm chair ethicists go on about how Gen AI shouldn't be generating inaccurate information. Of course you wouldn't want a medical LLM creating incorrect answers, but the current class of LLMs and Stable Diffusion models are statistical models meant to create output that is unique. They are not dictionaries meant to be fact based and then again my facts aren't necessarily your facts. This is why creative individuals use these models to give them ideas. They create highly unique output highly efficiently. There are many domains where you should not yet deploy an LLM because of accuracy issues - medicine being one of them.

My ethics are not your ethics and my biases aren't your biases. Think about that for a minute and then think about how an engineer would train a model from that statement. In a world where we have become hypersensitive it is almost impossible to get this correct. But then again what is correct? The simple answer is that there is no right or wrong when it comes to ethics and responsible computing but there are some rules an organization should follow and every organization will have their own set of guidelines. Depending on the domain there are examples of when bias in data makes rational business sense and data reflecting those biases should be included. I like to say that math isn't biased but the data going into it might not accurately reflect reality. This is where having the right talent that understands both the data and the business domain is critical.

It is also important to understand that models will decrease in accuracy when you remove data that someone thinks reflects a bias of some sort. The only way you have no bias is to have no data. Comprehension, and formalized data governance, of both the data included in the model and the model itself is necessary otherwise incorrect conclusions can be drawn from the model output.

Embrace Innovation

Gen AI offers transformative opportunities for businesses willing to embrace change and harness its capabilities. However, understanding and leveraging it is not just about adopting new tools; it's about cultivating a culture of innovation, embracing ethical considerations, and recognizing the complexities of technology integration.

Leaders must navigate these changes with a clear vision, grounded in the reality of their organizational culture and the global technology landscape. The journey toward innovation is continuous, requiring a commitment to learning, adapting, and investing in the right talent. By thoughtfully integrating Gen AI and fostering a culture that values creativity, ethical responsibility, and technological fluency, organizations can unlock new realms of efficiency, creativity, and strategic advantage.

In the face of these advancements, we must remember that innovation isn't just about the tools we use but how we use them to reshape our business models, operations, and ultimately, our future. Balance the potential with the wisdom of experience.