Every client we work with these days wants to incorporate some form of AI into their product experience. Whether driven by the C-suite, financial markets, competitors, industry analysts, SMEs, or likely some combination of all of these, the goal is the same: Get some compelling and useful AI capabilities into the product as quickly as possible.

Unfortunately, this feverish demand can clash with two realities: getting users to trust AI  is challenging, with many having been burned by vendor overpromises; and internal development teams don’t always have the talent, resources, or time to deliver the AI powering new experiences.

How Product Leaders Should Think About AI

While many slide decks dream of an autonomous, self-healing, AI-driven SaaS app world where humans occasionally audit but largely let AI run the show, that reality is likely much farther off than “Q12027” for a variety of reasons. That said, it’s clear that people really want some outcomes that a realistic AI can deliver on in the near term. This includes two key needs:

• Taking Over the Mundane. People still see their fellow humans doing complex or creative jobs, but they definitely dream of AI helping to alleviate them of the boring, repetitive, and relatively simple things that devour most of the hours of their day. A prime example is higher-level technical support people who want to be free from fighting small fires all day. They will happily outsource these tasks to less skilled, AI-augmented colleagues, or even to AI conversation bots.
  
  

• Making Sense of Complex Data. Almost every organization collects way more data than they can ever make sense of or use. And many struggle with Tableau dashboards or try to use Google Sheets to figure out the signal from the noise. AI’s analytical capabilities are a perfect match. Of course, like the complex algorithms that preceded AI, the underlying system has to be solid and free of bias to provide value. Distilling useful insight leads us to our next topic: AI principles.

Key AI Principles for Current Products

Based on extensive end user research, talking to stakeholders at a variety of vendors in many industries, and tracking AI in general, we believe most organizations would benefit from using these key principles as guidelines:

• Keep the Team From Overpromising. In most cases, AI is simply not ready to do everything the CEO or marketing teams dream up. For example, most AI-driven script generators can deliver lines of code, but they don’t have critical features needed to make them actually useful, such as the ability to test the code and iterate. And that’s before adding cool features like “what if” scenario generators to see how it will impact a larger system.
  
  
• Offer Multiple AI Recommendations and Efficient Feedback Delivery. Most users we talk to just don’t think AI will come up with the single, best course of action. If there’s just one option, they report that they are likely to tune out the advice. However, if there is a range offered (three is often cited), and they can provide feedback to improve it (a simple up/down vote is often easier to get than text feedback), they are much more willing to at least review what is offered.
  
  
• Let Users Control the Degree of AI Influence. The product team may want AI front-and-center, but given it is likely to underdeliver, users will want a way to adjust its presence. For example, a product could offer consistent but somewhat subtle interaction moments and controls. As users get comfortable, make it easy for them to increase trust in AI recommendations and have more prevalent AI prompts on screens.
  
  
• Allow Users to Peek Under the Covers. The models and code that drives AI won’t make sense to most people, but some insight into the source of the data and why, at a high-level, the recommendation was made can increase user trust, especially during early use and even later, during complex interactions with significant impact.

Figuring Out the Secret Sauce of AI for Your Product

Getting AI in your near- and mid-term product rollout plans is not easy. With so many inputs and ideas, often the chosen path can look like the least common denominator approach, one that tries to give something to everyone but ends up missing the mark, particularly when it comes to end users.

So how to improve your chances of success? What we have seen work is to blend strategy refinement with user-centric research and creative design exploration. Each informs the other, and when woven together correctly, they can deliver products with AI that are useful, trusted, and capable of being built.

Getting Started

AI is no longer just a buzzword—it’s a reality shaping product experiences today. But figuring out the right balance of AI-driven features is a challenge. Whether you’re struggling to meet ambitious goals or navigating user trust, we’ve seen firsthand how combining strategy, research, and design exploration leads to successful AI product rollouts.

If your team is interested in integrating AI into your next project, or if you’re facing challenges with your current implementation, we’d love to help. Reach out to us at hello@proximitylab.com, and let’s explore how we can elevate your product with AI.