I’ve been reading A LOT about AI and Retail recently – bringing together two passion points of mine. I’m very curious about the AI journey we’re on, and excited to be on the learning edge of technology as it develops into use cases and applications. Which ones will stick, which ones fade and which ones burn out in a blaze of glory?
Anyone around during the dot-com boom and bust will know the parallels and while there are many factors that make AI different, there’s more than a hint of ‘bubble’ in some of the products and business valuations hitting the market.
Looking at retail and AI at the moment there’s an awful lot of blah blah blah being written about its potential. I’ve come across amazing statements that AI will accelerate digital transformations, offer customers greater value and even support the transition to net-zero retailing. Each one bold, all of them with one thing in common – an absence of detail in how they’ll actually achieve these things.
Understandable to an extent – this is emerging technology and the hype is running ahead of the reality. However, it’s also unhelpfully increasing that hype, setting Board and Shareholder expectations that AI is the new silver-bullet to solve all retail’s woes. That wasn’t true of digital, and it won’t be true of AI… but there is HUGE opportunity to use AI (alongside automation, digital and data initiatives) to address some of the very real challenges facing retail into 2024.
And here’s how.
Set the foundations
Get specific - as Linus from Peanuts so wisely said “to solve a problem, first you need to understand what problem to solve”. Move away from generic ‘tech & data’ statements to desired outcomes and be mindful of the opportunity cost of chasing shiny tech or vanity projects.
Use cases around demand, inventory and distribution are the most commonly seen to date for retailers including Nike, Mango and others – but with the advent of Gen-AI, identifying the right use cases that affect service and experience will be next on the horizon.
Speed to value – many retailers don’t have the time or resources to invest in long-term transformations that take them below the waterline (the dip of doom) in return for promised returns in years 2 and 3. Apply business agility and Agile to deliver achievable value increments and experiment with AI and other tooling along the way.
Bring in the architects – there are armies of SAAS and platform vendor sales teams beating a path to retailers’ doors, promising ‘AI driven solutions’ (or having recently put AI in front of their product name). Experimentation is important and external SAAS can provide a quick route to action but beware today’s innovation becoming tomorrow’s technical debt. Think Jenga on an industrial scale.
Create a culture of curiosity – although focus is on the Big-Ticket items, one of the best use cases for AI is the removal of organisational tasks which are prone to errors, repetitive and high volume (or all 3). Your team are often the best source of identifying problems and opportunities – from customer-facing experience improvements to back-office operations.
Unlock AI by engaging your teams who may be fearful of being replaced in the AI conversation, with permission to play. Walmart recently announced they are experimenting with Gen-AI, with proof of concepts underway in contextual-search as well as voice-based AI-bot interactions to manage delivery. By embracing experimentation, they benefit from signalling innovation to the industry whilst iterating concepts in real customer scenarios.
Ethics, Risk and Governance – it sounds counterintuitive to propose agility and experimentation and then drown it in governance but the nature of AI demands that the foundational conversations are had up front. The technology is emergent and experimenting at the bleeding edge always contains risk. Be open and clear on elements including:
Approach to generative AI sources and traceability of ideas / content
Transparency of the AI machine. Are you comfortable with ‘blackbox’ AI that cannot be interrogated, or do you want an audit trail of ‘why’ alongside ‘what’?
Fail fast and opportunity cost – this is emerging technology and not all solutions will have longevity. What is your corporate culture for being a technology leader vs a fast follower?
Explore & experiment – with a focus on Speed to Value
So you’ve got the basics in place. What next? There are great examples of AI already being used in a retail context and in the spirit of avoiding more blah blah blah, let’s explore a few of those.
Going back to the first foundational point above, we think it’s important to anchor AI (and other digital / data initiatives) in clear business outcomes – aligned to your value chain. Every retailer will have a different value chain, but looking at the aggregate view there are a wide variety of opportunities to consider.
The chart below shows an array of strong opportunities. Some of these will be easy to implement coming through upgrades to existing packages, others are more difficult requiring new and potentially complex integrations.
The map of opportunities will differ from business to business, and to be pragmatic, substantial progress can be made towards the above outcomes through sensible improvement to processes and use of algorithms that are already well understood and proven.
Some of these AI opportunities will come relatively easily. For example, most retailers will be using some form of packaged software for Associate Scheduling in store. Many vendors have already incorporated AI powered algorithms into their products so as long as you have an environment where product upgrades are straightforward and can hook in some of the additional data sources required to power advanced algorithms, then it should be straight forward to obtain AI benefits.
Different routes to implementation
Not without bias, we would always propose starting with an AI Impact Assessment – short, sharp, broad and shallow, to help identify the biggest opportunities (where can AI help you get or stay ahead?) and highest risk areas (where will AI disrupt your existing value delivery?).
Looking at implementation and operationalisation, there are multiple routes to introducing AI in your retail value chain.
1. AI built in
SAAS / PAAS for most retailers the first exposure to AI will be through existing packages and platforms, investing in their own capability and rolling it out to clients. From Salesforce to Adobe, Shopify to Bloomreach and any number of magic quadrant vendors, AI is already making its way in the upgrade path and releases.
These vary from back-office improvements to addressing retailer use cases. One of my favourites is the use of Gen-AI to produce tone-of-voice appropriate product descriptions, like the one Shopify have launched - Shopify Magic.
If ever there was a high volume, repetitive and unloved task in digital retail, this has got to be high on the list.
At this point these vendors are mostly using AI to keep relevance and edge in their existing products but expect enhanced AI service offerings to come to market with equally enhanced pricing over time.
2. AI Tools
Already in use and being experimented with across retail, this is the first step for many of us in discovering how AI can help us do what we already do – generally faster, optimistically better. Some of the more mature and headline grabbing tools include:
ChatGPT – arguably the fuel behind the accelerated hype, Chat GPT’s open access and ability to collate terabytes of information into summaries or generate ideas based on prompts has made this tool synonymous with generative AI. Currently a bit too ‘general’ but rapidly accelerating towards more advanced use cases, using walled-garden data sets that tap into retail data from stock to SCV.
Grammarly is used to review and proof content from emails to academic papers and ironically in the world of AI generated content, Grammarly is in use in universities as a ‘plagiarism detection tool’, using AI to combat… AI.
Midjourney is one of the leading AI tools that generates images based on text input. Largely being deployed agency-side at the moment to generate initial concepts before adding human creativity, for retail, use cases extend from accelerated and personalised marketing assets, through to visual merchandising.
There are many tools in the marketplace with more being released every week. Generative AI in particular is attracting tools that talk to content, creativity and marketing use cases, but they’re rapidly extending into merchandising and omnichannel user experiences.
As above with the all-important architects and governance it’s crucial that alongside experimentation, retailers have a handle on who’s using what and sets the parameters for safe usage. Those Samsung developers who accidentally shared proprietary code with the world is going to be one of many such stories – avoid them being about you.
3. Developing your own solutions
Although this seems furthest off for many, there’s an argument that the leading retailers will go here first. After all, the packages SAAS and tools are available to everyone, so short-term first mover advantage will be replaced with long-term sustainable change.
The development and launch of ‘plug-in’ solutions will represent the path for many retailers to benefit from advanced focused solutions, integrated into your value chain and channels.
The ability to deploy AI processing and analysis at scale, operating on retailer’s own data sets is a really exciting area to develop meaningful competitive advantage.
However, to do this effectively means:
a. Using AI to amplify your difference, not be the same as everyone else
b. Anchoring your use of AI in your own data and training the models on your brand and voice
c. Composable and Open architecture solutions, enabling AI to be an integrated part of your digital and business automation not a bolt-on
For a case study in developing proprietary AI based solutions, look no further than Amazon Go / Amazon Fresh. But that’s another blog entirely!
Start experimenting now
At Transform, we’re already experimenting with applications in our AI Labs, embracing experimentation (and failing fast!) along with everyone else. In the process the biggest learning has already proved the first point made: Get specific on the specific challenge and opportunity to address.
We’ve developed great AI proof of concepts which worked but added no appreciable value, alongside AI solutions for challenges affecting our clients today – from consistency of job descriptions in a fast-moving media company, to AI driven automated test protocols and execution.
I was privileged enough to be involved in the last retail revolution, working with retailers across GM, fashion, homewares and marketplaces. There will forever be a bit of my heart with Argos, one of my first major transformation projects. The journey from digital to multi-channel to omni and social, and now unified commerce is a never ending one – but its evolution has taken place over decades. The AI retail revolution will be much faster and only time will tell if it has the same impact on the industry. I suspect so.