WEBINAR - Plug‑In AI for Retail Intelligence

https://vimeo.com/1170408281

Plug‑In AI for Retail Intelligence: A Practical Approach for Retailers

Top teams in grocery and retail are under pressure to “do something with AI”, but most don’t know where to start or how to turn it into real commercial outcomes. This webinar is designed for CIOs, technology leaders and their business counterparts who want a clear, low‑risk path to AI‑driven retail intelligence.

With the help of Garth Sutherland (Chief Executive at Farro Fresh Supermarkets) who has recently implemented AI-powered Retail Intelligence across his 7 stores, we’ll walk through a practical approach for overlaying AI on the data you already have – POS, product, store, inventory and loyalty – so you can support better category, merchandising and marketing decisions without rebuilding your platform or hiring a data science team.

You will see proven patterns for sourcing the intelligence layer, using natural‑language “virtual analyst” capabilities, and managing security, governance and total cost of ownership, so the C‑suite can move from vague AI ambition to measurable results

Wed, 4 Mar 2026 10:00 - 10:45 (UTC+12:00) Auckland, Wellington

Hi everyone, thanks for joining today.

I'm Charles Blumfield, Customer Experience Director at 11Ants. If you're here because you're feeling pressured to do something with AI, you're not alone. A lot of retail teams are being asked to act, but it's not obvious where to start, what looks good, or how to turn it into real commercial outcomes.

Today we'll talk through the reality of what makes AI hard in retail, then share a practical, low-risk way to layer AI on top of the data you already have — so you can make better decisions across the business without turning it into a major build.

Things are moving fast, and retail is right in the middle of it. The biggest retailers are investing heavily in in-house tech and AI teams with big budgets and long roadmaps. That gives them a head start. For everyone else, it can get expensive quickly if it's not tightly connected to value.

At the same time, boards are asking for AI strategies — often without clear direction on what to do first, or what success should look like. So a lot of small and mid-size grocers end up feeling stuck.

All of this is happening after years of investment in operational platforms and purpose-built data. The question has shifted. It's no longer about collecting data — it's about getting real value from the data you already have. If that sounds familiar, you're in the right place.

Over to you, Tom.

Hi everyone, I'm Tom, CEO here at 11Ants. I want to talk a little about where we'd be deploying our AI focus.

I think there's not a board anywhere in the world that doesn't think their company should be doing something with AI. And most of what they're reading and hearing is — by design — made to sound like a very easy thing. Then it gets put to the implementers and the doers, and like most things, the devil's in the detail. We start trying to do things that are a lot more complicated than they sound.

So I think the first question everyone needs to be asking is: where could we deploy AI where it creates value quickly, without becoming a science project — or worse, a failed IT project?

A useful way to look at this is with a grid: value creation on the x-axis, and time to value on the y-axis. There are lots of worthwhile bets in the long run, but we need to narrow down where to start. And obviously the smartest place to start, if at all possible, is in that top-right quadrant.

On the next slide, you can see this quadrant populated. There are literally thousands of things that could sit here — we're just illustrating with a few examples.

Something with short time to value and relatively low value creation might be wait-time optimization, or issuing co-pilots to the team. Things that take a long time to deliver and still have relatively small value creation — like a supplier AI scorecard or personalization — or really any IT project that doesn't work out, end up in that bucket.

Moving across to the bottom right: things with long time to value that do create quite a lot of value — like AI demand planning and ordering. Retail intelligence, traditionally, has sat in this bucket.

In the top-right quadrant, we're looking for quick wins that make a big difference. One of them — probably little known — is retail intelligence done in a slightly different way than people have been traditionally doing it. Where you can actually deploy something very quickly and have a meaningful impact on the whole business, because you're touching the major decisions the business makes.

Here's the reality we seem to hear everywhere: it's just not practical to be evidence-based for every decision made in a retail business. If someone's building a new store, lots of analysis gets done — spreadsheets, analysts, a robust and defensible process to decide where that store goes. But arguably, there are thousands of decisions being made every month which, cumulatively, have a much bigger impact than where you locate a store. And those decisions can't be evidence-backed — it just takes too long, it's too hard.

So the question becomes: how do you give everyone in the business their own personal analyst, without spending a fortune or waiting a lifetime? Where they can put an evidence-based lens on every decision they make, because it's that easy and that simple.

If we could solve that, we have what we define as retail intelligence — not just a dashboard dumping ground, but starting with a business problem, not a report. Instantly accessing the right data to answer it. Understanding why something's happening, not just what's happening. Getting clear guidance on what to do next. And ultimately, making better decisions faster across the organization.

Before we talk about analytics or AI, I want to acknowledge a reality that most retail teams live with. You don't have a lack-of-data problem. You usually have a "where is it, who owns it, and can we trust it" problem — and that's normal.

We use a simple way to get clarity. First: what problem are you trying to solve, and what data do you actually need to solve it? Second: what data do you have, and where does it live? For most retailers, the core set is POS and transaction data, plus inventory, promotion, assortment, budget, and wastage. Loyalty or customer data is great if you have it, but it's not always the starting point. Third: what shape is it in? Is it complete, consistent, and timely enough to support business decisions? Most of the time for retailers, it's usable — it's just locked away in operational systems, split across teams, and often only available in aggregated form. And finally: what's the minimum you can start with? In practice, you start with transaction data and add the next most valuable pieces. You don't need everything on day one. The goal is to start small, but still be able to explain what happened and why.

When you look at your organization through that lens, you can usually place yourself on a maturity curve pretty quickly — and that helps you pick the next step without turning it into a big program of work.

What we're looking at here is an analytical maturity matrix. Some organizations are in a data fog or data lockup. Others have dashboards and a few power users. But very few have true, organization-wide data fluency.

The big point here is: you don't necessarily need to be climbing this ladder one painful rung at a time. Traditionally there's been a very clear path you needed to follow to go from a starting point to being an incredibly data-led retailer. I think that's less true now. There are ways to jump from quite low on this curve to very high. We call the pinnacle of that being an intelligent retailer.

Of course, there's a right way and a wrong way to overlay AI — and that's what Charles will cover next.

Back to Charles.

We've all pasted data into ChatGPT and been impressed. So it's completely reasonable to think you can point a large language model at your transaction data and start asking questions. In practice, that falls over pretty quickly.

First, there's a hard technical limit. LLMs can only work with a fixed amount of information — a context — at any one time. Retail data is way bigger than that, even if you zoom in on one category or one part of the business. And that constraint still matters even as context windows grow.

Second is the nature of retail data. It's huge, detailed, and deeply connected. To answer real retail questions, you usually have to join millions or billions of rows across products, stores, calendars, promotions, customers, and inventory. If you push raw data into an LLM, it can drop relationships, miss key drivers, and give answers that sound confident but aren't decision-ready.

Even if you could cram more data in, the model still doesn't understand retail. It doesn't know what normal looks like, what counts as meaningful variance, or what matters commercially — unless you give it structure and guardrails.

And when you try to make this work for real, cost becomes the big issue. You end up building pipelines to pull the right data, join it, govern it, and keep it running. That gets expensive to build, run, scale, and maintain. You can try to solve the context issue by aggregating the data, but that's a false shortcut — you lose the detail you need to explain what happened and what to do next. And you still carry the risks around performance, security, privacy, and governance.

So this isn't a prompt problem, and it's not a model choice problem — it's a structural problem. Before AI can add value, retail data needs to be shaped into a retail intelligence layer that understands the data and the decision logic. Only then does it make sense to put AI on top.

So here's something that we'd argue sits squarely in that top-right quadrant — rapid to deploy, with an incredibly large amount of value delivered to an organization. That's turnkey retail intelligence. We're going to introduce you now to ANT.

ANT is designed to be plug-and-play with instant time to value. What we mean by that is: load your data into the system, and everything we show you comes to life instantly. It's accessible to everyone, with zero training overhead. Everyone's accustomed these days to asking questions and interrogating large language models — and we've certainly seen that as we've been deploying this with customers. We've seen them start running with it without any training at all, asking increasingly sophisticated questions. It's actually surprised us how quickly people adapt and learn to use it.

ANT is pretty simple at its core. Instead of starting with a dashboard and slicing and dicing until you find something, you start with a problem you're trying to solve — and it brings back the evidence, context, and next best actions.

At a surface level this looks simple — type something, get something back. What's happening under the hood is actually a lot more complex. Fundamentally, two things are going on: the large language model interprets the user's question, understands their intent, and turns that into a series of queries of the 11 Ants platform. The 11 Ants platform is exceedingly good at extracting exactly the data needed out of billions of rows, very quickly. Then the LLM interprets the results and provides a narrative. It's a system working in concert — not just you interrogating your database with an LLM.

And what that enables is some pretty remarkable things. Here's an example based on a real question from a real retailer: "Should I be swapping out my Vita Coco range for the smaller one?" — it comes back with specific suggestions and next actions, and tells you what to monitor once you've done that. Between the question and the answer, there are also two to three pages of robust analysis explaining why it reached that conclusion. We'll share that as a PDF with everyone after the webinar.

It's really difficult to showcase ANT's capabilities with only three questions, but I'll try.

Question 1 — the kind a category manager or analyst might ask: "Put together a supplier scorecard covering 12 months, 6 months, 3 months — compare performance versus last year — prepare for Coca-Cola."

You can see it interprets the question, pulls data directly from the platform, and returns the interpretation: detailed KPIs, analysis of what it all means, a summary, recommended next steps, and guidance on the sorts of follow-up questions you might want to ask.

In this case, we break it down by brand. Again — it understands the concept of brand, understands your data, queries it, brings it back, and gives us a detailed summary of Coca-Cola's brand performance.

Question 2 — a more complex one for most businesses: "What products are important to my high value customers?"

To answer this, you need customer data and transaction-level data. In this case, ANT asks for a bit more guidance: how do you want to define high-value customers? Which time periods should we analyze? And what product level — category, subcategory, or individual SKUs?

We specify categories, and it goes away, looks at our high-value customers based on RFM, and identifies the products most likely to be purchased by those customers. It comes back with detailed information, interpretation, and — again — suggestions for what to ask next.

We then compare these customers' product mix to the rest of the customer base, helping us identify differences between our high-value customers and others. It does a great job pulling out what's more likely to appear in these baskets versus others, with recommended next steps. We can then drill down to SKU-level performance.

Question 3 — looking at an individual store and comparing its performance versus last year, but benchmarked on a like-for-like basis. We've had a change in our network, and like-for-like is the only way to get a true read on performance.

ANT understands the concept of like-for-like and many other retail concepts. It goes away, takes that concept, drills down into categories, and helps us understand what's actually going on in that store versus the rest of our market — so we can dig in and see what's driving performance.

Look, three questions really doesn't do it justice. The only way to get a real sense of it is to try it yourself. But we do have someone here today who has had some experience with the tool.

Today we're joined by Garth Sutherland, CEO of Farro Fresh — one of the first of our customers to experience ANT and our retail analyst, turning retail data into clear, decision-ready insight. Garth is going to share what it's been like from a CEO's perspective, some observations from across the business, and his view on what the future might hold. Garth, thank you — over to you.

Fantastic, thanks so much Charles. Really appreciate the opportunity.

Evidence has been a big part of our journey over the last few years as we've matured as an urban farmer's market grocery business here in New Zealand.

On what impact the technology has had on me personally:

I think, like most people, AI has felt a little daunting. We've all known we need to get into it, explore it, make sure we're keeping up. The experience with 11 Ants has been really rewarding in starting to bridge that gap in our knowledge and understanding of how language models work and how AI can support businesses.

Simply put, it's changed how I start conversations in the business. Historically, I used to begin with "what report do I need?" or "what numbers do I need to pull?" That's now shifting very quickly to "what problem are we trying to solve?" — for our customers, for our team, for the business. And that's evolving the business at quite a rapid rate, which is exciting to see.

It's made me more confident and faster in my ability to review data, comprehend it, and come to solutions. I would have spent a fair amount of time working through spreadsheets, chasing numbers, reconciling versions, or waiting for someone who knows where the data lives to pull information. My ability to access information now has become so much faster. I'm spending more time actually thinking about the data and following up with actions — as opposed to just trying to find and pull the information in the first place. It's really helping me make decisions a lot faster than I've been able to do previously.

The biggest surprise for me specifically around ANT has been the quality of the thinking. It hasn't just been giving answers — it's giving a lot more detail and context around the information you're looking at. It'll give you the likely drivers, comparisons, and — what's really been exciting — it actually gives you options for what to do next. Of course, we have to temper that and make sure we're getting into the detail as a business, but it's driving our ability to ask harder questions as we move forward. There's quite a step change from just data retrieval to overall decision support. I've found that really rewarding.

On how it's changing the day-to-day of the business:

It's shifting the organization quite quickly toward more evidence-led conversations. Particularly in our trading meetings — those conversations have become a lot richer. We're not having nearly as many conversations that start with "I think this is the reason why" — it's more evidence-based, more "let's go and check." People are more willing to challenge assumptions because it's actually easier to access and validate the information.

Live examples of this: we'll have a trade meeting, and we won't take a week to work through and understand the questions — we'll do it live in the session. That's really helping our ability to make decisions quicker for our customers.

The other element is just how much it's broadening access to data. Not everyone in a business is data-inclined, and historically that's meant we've been very reliant on a few subject matter experts or super users to access information across various data sets or systems, and then try to work out the interpretation.

We've found very quickly that everyone — from marketing through operations through to our category teams, all the way through to category assistants and support office — is now accessing this information and self-serving. They're coming to discussions better prepared. That is so rewarding, and so rich, in terms of the quality of conversations we're now having.

As a result, it's speeding up our operating rhythm. We're able to answer questions quicker, look at pricing and promo reviews live, and make decisions based on that faster and more frequently. Ranging decisions — there's obviously still a place for in-depth range reviews, but understanding SKU and product performance, understanding customer impact on a particular range — we can access that information so much quicker. That makes us better for customers, and in a fast-paced industry like grocery, so much more powerful.

We're producing more insights in the same time it used to take just to produce reporting — and that's a big one for us.

On how it might change the future of grocery:

Grocery is really fast-paced. AI — and specifically the tools we're using with 11 Ants — is making us faster and smarter. We're shifting from "what happened?" reporting to running the business in near real time.

The reality is that for most of us, we don't have a shortage of information. It's our ability to synthesize it and make sense of it — to make real-time or near real-time decisions. Working with a tool like 11Ants, you're able to do that a lot faster, which means we're able to make customer-first decisions at scale. Instead of debating what we think shoppers want and need, we can interrogate the behavior a lot quicker — understanding shopping missions, price sensitivity, understanding what the real drivers are. And we can do it at scale.

We get to make better ranging and promo decisions — not just "did it sell?" but richer information: who did it bring in, what demographic came in, how did they purchase, where did they purchase, what did it replace, and — very importantly — what margin did it generate, and what were the outcomes for the category long term.

I will say we're still in our infancy in this space, so we're learning as we go. One big learning: you're not going to have all the answers straight away — you've got to be willing to explore and test and push into this space. Our ability to ask good questions has become really important to maximizing the opportunity with these tools.

One thing I'd really call out is taking steps toward more predictive operations. Forecasting and scenario planning has always been particularly challenging — using historical data to inform the future. Although we've started to push into that space, there's a lot of learning still to come. Things like: if we move price, what does that do to volume? What do we need to do from a replenishment standpoint? How do we work with our suppliers to better forecast volumes and margin outcomes? There's a lot in the grocery space that we're going to benefit from with these tools.

So — lots to think about, lots we're working through, and huge support from 11 Ants as a partner to help us develop this. Very excited about what the future holds. Thanks so much for the opportunity, Charles — I'll hand back to you.

Thanks very much, Garth.

Just a few closing thoughts — three questions to leave you with:

How easy is it today to answer the real business questions within your organization? Who can do it, and how long does it take? What would change if the answers were instant and shared — what kind of difference would that make?

The good news is it doesn't have to be a big step to explore this further. If today has been useful and you'd like to dig in more, feel free to get in touch.

We'll now open up for Q&A — click the Q&A button at the top of your screen and type your question in there.

Q&A

Question: Can you ask 11Ants to prepare a series of discussion points for a pending supplier negotiation around sales, margin, and stock opportunities?

Answer: Yes — and interestingly, we're also seeing people go a step further and ask it to prepare a presentation deck, or draft an email to send to a supplier. It's been quite fascinating watching people take the next step and the next step within the solution. It saves enormous amounts of time. Every day we're more surprised by the imagination people apply to what they can get out of ANT, and the further steps they can take without having to do very much work — which is always a good thing.

Okay, I think that's it. Thank you to everyone for taking the time to attend. Get in touch if you have any questions or would like to discuss further. We'll also send out the example ANT question and answer with a bit more detail of the response, for your reference. Thanks a lot, everyone.