#6: AI Made Me Do It — Goal-oriented AI is better than AI-oriented goals

By Bernard Kiyanda

A monthly take on what’s new in AI — personal stories, opinions, and inspiration from AI thought leaders. This edition is written by Drift’s VP of Engineering, Bernard Kiyanda.

What’s On My AI-Mind

Like any tool, AI must be applied with precision and forethought. The first step to building a house isn’t to buy a hammer — it’s to plan out your project and design your blueprint. Similarly, as we introduce AI-powered tools into our work and lives, we need to identify our needs and goals, and then plan out how AI can help accomplish that vision. The startling speed at which AI is accelerating means that new AI-enabled tools are being launched every day. It’s thrilling to see so many people around the world apply their uniquely human creativity and ingenuity to shape the ways AI can improve our lives. But this also means that there are now countless AI models, and one form won’t be the fix-all for every problem. For this reason, it’s necessary that companies do their due diligence in research upfront because those who rush to employ AI tools without their blueprint on hand may find themselves disenchanted, rather than empowered, by AI.

In the AI era, balanced and informed decision-making is key. At Drift, our engineers, product managers, and designers have been shipping AI products since 2019. We’ve deployed sales and marketing products augmented with GPT and other large language models to transform the buyer journey. On our team, we ensure that every team member leverages AI tools so that our decisions are diverse and expansive, not siloed to only data scientists. That’s why we’ve held two AI-focused hack-a-thons — collaborative events where the entire product team works together to learn and get creative with AI. This way, when we’re building our product roadmap, we can all think about how AI may help us solve a specific problem more efficiently. Because we continuously measure the influence of AI on our customers, our decisions are informed by the customer impact of specific AI capabilities like performing an intent classification, detecting a customer topic, summarizing large amounts of data, or generating AI answers based on previous similarities rather than strict rules.

The Intelligence Revolution is a marathon, not a sprint. AI is here to stay, but that doesn’t mean you should rush to adopt everything all at once. Fight the urge to play catch up. Instead, consider what the transformational journey looks like for your work and your team. Start with your goals — and your small ones at that. Take a trial-and-error approach, and always include safeguards (particularly for data privacy) and rigorous feedback processes. Even if you think you’ve found the perfect application for an AI tool, managing the challenges of AI biases, hallucinations, and response accuracy needs to be incorporated into your project roadmaps. Don’t assume AI is always the answer: In many cases, you may find that your business experience creates an effective customer outcome. But by thinking strategically about where AI can fit in within your broader project, you’ll find that AI can drastically accelerate goals for your team.

Starry-eyed zeal won’t give you an edge, but customer value and smart expansion will. Ultimately, it will be iterations that refine and fortify our understanding of the growing relationship between humans and AI. As with any major innovation or discovery, some of the glimmer and changes of the generative AI boom will turn into new challenges. Nevertheless, AI is one of the most revolutionary technologies ever created, and it has the power to streamline our work, improve our lives, and advance our visions for a better world. The companies that will thrive in the future are those that will adapt rapidly and transform their product with it, while still creating human connections between buyers and sellers.

GPT Made This for Me: Browsing Without a Browser

What I love about GPT is that when I have a question, I can always get the answers I need. Search browsers have become time-consuming, and it’s difficult to find the actual information I want. It’s tedious to have to navigate all available information. With GPT, however, I can ask a question and get a straightforward answer from research or code — and if I want to follow-up or iterate on the responses, it’s easy to do so.

The Single Best Thing I’ve Watched About AI Recently

This intro to Large Language Models by Andrej Kaparthy is an excellent resource that provides the fundamentals to understand LLMs, and I recommend it to everyone interested in learning more about AI.

Drift leveraged its own AI to analyze the most common topics of conversation FROM the 30M+ conversations that happened in Drift last year. Read our new Conversations Trends Report to find out what we learned.