Technology is changing our lives faster and more completely than any time in the past 150 years. Generative AI really is revolutionary, poised to reshape how we live and work. And that of course is where a lot of the fear comes in. Doomsday headlines abound: AI Poses ‘Risk of Extinction,’ Industry Leaders Warn. AI Won’t Really Kill Us All, Will It? AI Chat Bots Are Running Amok – And We Have No Clue How to Stop Them.
Actually, we know how to wield the tools and tame the technology. In fact, many of the old rules still apply, including the principle of garbage in, garbage out – which was (probably) coined way back in the 1950s. GIGO applied here: your Generative AI solutions, and the outcomes they deliver, will be only as good as the data you start with. So here, six steps to set yourself up for success on the journey ahead.
Six questions to set your organization up for success in Generative AI development
1: What is our AI ‘why’?
Generative AI is exciting! But as the old adage goes, if you don’t know where you’re going, any road will take you there. Success starts with defining what you expect to get out of your work with GenAI.
Consider use cases from an outside-in lens versus inside-out. Build in control metrics to judge success. Say, for example, you want to use Generative AI to improve productivity and reduce cost. Great! How will you measure success? What if productivity goes up and cost goes down but customer experience falls, too? With GenAI there will be many knock-on effects. Be measured and deliberate to ensure you’re not improving one area of the business at the expense of another.
2: Is our data fit for purpose?
Are the data you’re planning to feed into the model fit for Generative AI? (Spoiler alert: Most data are not.) Ensuring the data are available and fit for purpose for GenAI consumption requires deliberate engineering work. Data modernization is about planning how you’re going to manage the data. How will it be consumed? By whom? Who gets to act on the data? These questions aren’t rocket science, but you’ll run into trouble if you don’t ask and answer them – laying out the rules of the road for your organization.
One of the most exciting aspects of Generative AI tools is their ability to ingest a tremendous amount of information and make it easily consumable and actionable. The more good data, the better, and you probably have more of it embedded in your organization than you realize. (Most companies do.) It’s important also to think beyond your own data. Multi-sourced, multi-dimensional data is like a superfood salad for GenAI.
3: Do the economics work?
Building and operating successful Generative AI models costs money, of course. Infrastructure cost, in particular, is a common tripwire for organizations as they start their AI journeys. I’ll break the news to you now: Processing and storage costs associated with GenAI can be quite significant. Yes, even in the cloud. And while high usage is the end goal, it’s important to remember that the more people using the model, the higher your infrastructure costs will be. Rationalize the cost relative to the business value being generated – at the proof of concept phase and at full scale.
4: Is our organization ready to leverage Generative AI solutions?
Generative AI is AI democratized. You don’t need to be a PhD data scientist to understand and articulate what is coming out of the model. Yes, modernizing the data and building the model is a tremendous amount of very sophisticated work. But once it’s done, for the first time ever you have an AI that your whole workforce can interact with. No Bayesian equations required.
Alas, that means change, which is notoriously hard for humans. Are people willing to make decisions differently? Do they trust the AI? Do they feel threatened by it? All tough questions that will determine whether people adopt the new way of working, and you succeed – or not. That’s where Cognixia, an Ascendion company, comes in. With a roadmap for recoding your organization’s DNA they help you take advantage of GenAI’s power while also minimizing its risks.
One way to increase adoption is to allow people to interact with the Generative AI tools in whatever way works best for them. Key is to not create yet another noise machine; people are already buried in information. Leverage the power of AI to enable individuals to differentiate noise from valuable signal. Close the gap between insights and decisions and empower people to make better decisions faster.
5: Have we built ethics and governance into our Generative AI development?
One of the very real fears about advanced AI is the risk of unintended bias, privacy violations, and security breaches. The risk starts with the data, so ethics and governance need to start there, too. At Ascendion, our Ethical AI Framework governs how we develop Generative AI solutions preserving ethical standards and ensuring security, privacy, and compliance. A layer of instrumentation and question analysis on the prompt side creates a pre-emptive early-warning system.
6: How are we monitoring and measuring outcomes?
True to its name, Generative AI should get better as it interacts with you. But what does ‘better’ mean in your context? How do you measure and monitor the inputs and outputs of the GenAI model? When do you inject new data? It often works well to design experiments to see how small changes affect outcomes at small scale. Learn, test, tweak, repeat. Scale up what works well and turn off what doesn’t. It’s a virtuous cycle of learning and feedback.
Ready?
You’ve figured out what you’re aiming to achieve with GenAI. You’ve modernized your own data and added multi-sourced, multi-dimensional data. You’ve figured out the economics. You’ve recoded your organizational DNA. You’re confident that if bias, security, privacy, or compliance issues arise, you’ll catch them early. You’re learning, testing, tweaking, repeating.
In other words, the car is packed, fueled, and everyone’s gone to the bathroom. You haven’t left the driveway yet but you’re ready to head down the road on your AI experience modernization journey, accelerating the path from data to decision.
Destination: Exponential impact.
