Humanizing Retail AI - Our Design Philosophy for Trust in Retail Tech

Humanizing Retail AI - Our Design Philosophy for Trust in Retail Tech

The retail space is no stranger to solutions powered by artificial intelligence, machine learning or any equivalent buzzword in and around this realm of technology.

Search up ‘AI tools for retail’ and you’ll see plenty of sponsored results and lists like ‘Top 10 startups developing AI for retail’. There’s plenty of players in the space, there’s no doubt about that.

What there is doubt about, however, is how this works in practice. How exactly does AI help retailers? What exactly does the experience look like? When does AI work well and when does it not?

It’s unlikely that you’ll find the answers to these questions inside the very landing pages that are ushering in this new age and marking a ‘new dawn’. The reason for that is understandable.

Building experiences powered by Artificial Intelligence is not an easy business. Large companies have devoted years and thousands of people to create usable, understandable and most importantly, truly augmentative experiences for the masses.

Within retail though, this challenge is slightly different. Most of the B2C applications of AI we see out there operate within a completely different set of bounds compared to B2B. In B2C, no one is forcing you to use the technology, you are predominantly responsible for adopting a new product. Within B2B, your boss, who you might not particularly like, is being told by their boss, who they might not particularly like even more, that a new piece of software is now in use.

That’s a very different set of circumstances to downloading an app after your friend shared a link.

As a direct result of these different circumstances, designing the experience for these users and their goals poses a new challenge.

A challenge that centres around trust.


Despite the complexity of AI systems, the premise behind their adoption is extremely simple.

Users, who are usually asked to adopt a system, rather than choose to adopt it, are giving up some level of control to a faceless, non-human piece of software.

Ask yourself when you’ve given up control of something, what things did you think about? What questions did you ask yourself? What made you nervous?

This is the exact journey a user who’s adopting AI software is thinking about themselves. When you gave up control over those things, whether it be taking your child to daycare, letting a dentist take out your wisdom teeth or even letting someone else drive the car, you will have asked yourself at some point:

“Do I trust this person”

The same question will be asked of any AI software. The only difference is that instead of not trusting a person, you're not trusting a piece of technology, more fundamentally, a piece of code.

It’s our job as designers, engineers and salespeople, who deliver this code to users, to ensure that they trust us.


A group of UC Berkeley researchers created a fantastic resource AIUX, which I stumbled onto while asking another AI tool (GPT) to help me find resources around data-heavy experiences.

There’s an excellent concept in this resource called explanation types:

  • Global Explanations - Global explanations help users understand and evaluate the system.
  • Local Explanations - Local explanations help users examine individual cases, which can help with identifying fairness discrepancies and calibrate trust on a case-by-case basis.

That last sentence there is key:

Local explanations help users examine individual cases… and calibrate trust on a case-by-case basis.

This is what I call the holy grail of AI powered experiences. It becomes more important the more control the system wants to have, and therefore the more control a user gives up.

In the current setup for retailers, it’s fairly easy to trust Excel, because Excel is a mirror of you. Excel does basic calculations, yes but those are never in doubt, and you input everything else, so it's easy to feel fully in control as a user in that situation.

But with an AI-powered system, we’re now moving to a better world than Excel, or so we as the creators deem. The challenge is to convince users of this too.

The more local the explanations within the system, the easier it becomes to calibrate trust, and the more likely a user is to adopt the technology.

There is a direct correlation between trust and adoption in AI-powered systems.

We need to do as much as is necessary to get to the tipping point with users, and to make them trust our system.


Retail can have some very messy and specific problems. This poses a challenge to AI-powered systems, but it also means those that take on the challenge, might have a tasty reward waiting for them on the other side… users eager to adopt it.

For us at autone, we’re putting the explainability of our powerful calculation engine at the heart of the experience for our users.

We’re finding the balance between the tango of human and machine. When to proceed with the reasoned recommendation from 100s of data points, and when to make adjustments based on your specific demands and expertise.

We need this balance to always exist, and we need to ensure we listen to why our users make adjustments when they do decide to make them. We frame the experience as a review and adjustment process. It’s not a battle of you vs. the system, but it's a tango between the two.

We crunch the numbers, save you time, and present you the data in a way that’s digestible and reasonable, and the user, akin to a pharaoh in Ancient Egypt being fed fruit, picks out the grapes that take their fancy:

We want to explain how we derived our recommendation, and what factors we took into account, so that you as the retail expert, can judge accordingly.

We’d love for you to always take the fast route and proceed with our recommendation, but that’s unrealistic, and any system that prevents that is doomed to fail its users through a lack of trust.

So we present to the user what needs to be seen, make it easy to review, adjust where necessary, and proceed to take the action that matters, moving the right amount of stock, to the right stores, at the right time.

Our user's job at the end of the day is very simple:

Select a certain number of units of stock.

That’s it.

That’s the tweet.

autone helps you get there fast and with trust.


Retail is messy, but autone keeps things simple. ??

Are you attending NRF '24? Be sure to visit autone at booth #403, where we will be showcasing our innovative AI-powered retail solutions, offering a unique blend of usability and trust.

Discover how we're redefining the retail experience and learn how our technology can benefit your business.


要查看或添加评论,请登录

社区洞察

其他会员也浏览了