Everyone building software has read similar headlines this year. About US$2 trillion came off public software valuations between January and April. Software multiples fell below the S&P 500 for the first time in history, from 84x at the 2020 to 2022 peak to under 23x by March. Big names like HubSpot, Atlassian and Figma lost 70 to 80 per cent from their highs.

The driver behind the selloff is simple, and the outcome was not particularly surprising. If an AI agent can do the work, you don’t need a human sitting in front of a screen, and if you don’t need the human, you don’t need per-seat licenses. The Tech Insider article sums it up in one sentence: “Once the human user is no longer in the workflow, the per-seat license is no longer in the budget”.

meed is, on paper, a SaaS company. A retailer signs up, builds a loyalty program in a web portal, and pays us a monthly or annual subscription. So, where does that leave us? Does it present a risk?

We don’t have a per-seat license; we charge per location and track where people join, earn, and redeem.

But we are building an agent interface for meed, and I expect it to replace most of what the dashboard does today.

However, the dashboard is about how meed works. It’s not about why meed works, and this is the important point I want to explain in this article.

Yes, the interface is going away

Let’s start with the arguments not in our favour.

Most of what a retailer does in meed today is the kind of structured work agents are good at. Set up a stamp card, choose a reward, look at who joined this week, and notice that a regular hasn’t been in for a fortnight. Today, that is a human navigating through our portal screens.

Soon, it should be an agent that onboards you through conversation, watches your program for you, and brings you the opportunity rather than waiting to be asked. Proactive, not reactive.

I want that. A café owner at 6 am doesn’t want a dashboard. She wants to be told, “Three of your regulars are one stamp from a free coffee, and one of them hasn’t been in for ten days. Should I send them a nudge?” and to be able to say yes.

So if your definition of meed is a dashboard hidden behind a login, then agents are coming for that, and that is perfectly fine with me because...

The interface was never the moat

Bryan Goode, Corporate VP, Business Applications and Agents Marketing at Microsoft, made the point better than most of the people predicting the death of software as we know it. “In an agent-driven world, SaaS will no longer be what people log in, navigate, and toggle,” he wrote in Fortune. “Their value was never the interface - it has always been the data, relationships, governance, business rules and logic that lie underneath”.

The market agrees with him. The companies punished the most are “horizontal SaaS platforms that primarily organise, route, and display data without owning it.” The ones holding up under this pressure own something the agent needs: a system of record, a proprietary dataset, or a network.

So the real question for any software business isn’t, will agents replace my UI? It’s: when the UI is gone, is there anything left?

For most loyalty software, the answer is, sadly, no. A stamp-card app for a single retailer is an interface and not much more. Magic away the screens, and you have a database table with ten rows. An agent can rebuild that in an afternoon.

meed was built on a different premise, and the premise is the part agents can’t reach.

The actual problem with loyalty

Here’s the problem we set out to solve, and it has nothing to do with UI.

Loyalty is a supply-side solution. Every retailer constructs its own program, around its own brand, with its own card, app, login, and rules. Then it expects the consumer to come and visit. The consumer is the integration layer. They carry the cards, remember the passwords, download the apps, and keep each relationship alive by showing up inside each silo.

They don’t, of course. Loyalty & Reward Co reckons most consumers hold between five and fifteen memberships. More than half of loyalty relationships go dormant, and my back-of-the-envelope calculation tells me that leaves a US$1.7 trillion hole in global retail from lost potential upsell. That isn’t a UI failure. It’s an architecture failure. We asked one person to maintain 15 separate relationships on 15 separate islands, and they quietly stopped.

meed inverts it. The consumer joins once. They get one identity and one wallet. Every program they join - the café, the barber, the gym, the lunch place - sits inside it. The retailer still gets its own program, its own branding, its own rules, and never sees the consumer’s personal data; what it sees is behaviour: someone joined, earned, redeemed, or has been gone for two weeks. One platform serving both sides.

That is not a feature. It’s where the data lives and who it belongs to. Which is exactly the layer Goode says survives.

Agents don’t fix the silo problem. They inherit it.

You would think that agents dissolve silos. Surely my agent can just talk to all 15 programs for me?

Read what the loyalty industry is actually telling itself, and you’ll see the opposite happening. Here are three examples:

dunnhumby’s long piece on agentic commerce - written for grocers - says the agent era means “loyalty must now be earned twice – once from the customer and once from the AI agent acting on their behalf”. Its prescription: each retailer must turn its program into one of the “data-rich, dynamically personalised systems that AI agents can query, evaluate, and incorporate into recommendation algorithms in real time”.

Fusion5, writing for Australian retailers, makes a wry observation. Every loyalty program “rests on the assumption that the customer comes to you.” When an agent does the shopping, “the seller never gets the session… The loyalty moment never happens. The program was built for a journey that no longer occurs.” Their line for what happens next is perfect: “The programme fires correctly - it just fires into a void”. Their answer is “callable loyalty”; each retailer exposes status, rewards, and earn rules via API so an agent can act on them.

Philip Shelper at Loyalty & Reward Co has the same message. Incentives today are “scattered across ERP pricing rules, POS systems syncing overnight, ecommerce promo logic, and CRM-driven targeted offers running independently. That fragmentation is annoying for humans, but for agents, it’s fatal”. His conclusion: “Loyalty must be centralised, structured, and agent-readable”.

Three serious, well-researched pieces that correctly identify the problem. And all three are addressed to a single retailer, telling it to make its own silo agent-readable.

Wonderful - 15 more APIs!

Stand back and look at what that means. 15 retailers each build their own agent-callable API. 15 APIs, 15 schemas, each controlled by a party with a commercial incentive to put things in the best light (a charitable description). The consumer’s agent now has to stitch 15 islands together instead of the consumer doing it. The integration burden has moved. The architecture hasn’t changed at all.

That’s the deeper problem. Agents are a new interface onto the same broken structure.

The counter-argument, and why it fails

The strongest objection to all of this is that the agent is the wallet. If my agent can query every program, I don’t need a consumer-side network. The agent aggregates for me.

I heartily disagree for three reasons.

1 - Querying isn’t owning.

An agent reading 15 APIs has 15 partial views and no record of its own. The Merchant Advisory Group’s analysis of agentic payments shows where that leads: “If AI platforms control discovery, loyalty shifts upstream… Loyalty may migrate to the platform layer, with agents acting as gatekeepers”. dunnhumby names names. Google, Amazon, and Apple “can offer AI shopping experiences that span categories and merchants, aggregating consumer intent in ways individual retailers will not be able to match”. So the consumer’s loyalty record stops being scattered across retailers and ends up inside Google’s agent instead. That isn’t a solution. It’s just a bigger silo that can respond in sentences.

The consumer’s relationships should belong to neither the retailer nor the agent platform. They should belong to the consumer.

meed’s demand-side solution encompasses the necessary permission-based approach that will enable consumers to own their identity and control its use.

2 - The long tail will never wire up.

Reading all the advice in the above articles: CDPs, decision layers, real-time APIs, “separate incentive decisioning from incentive execution.” That is enterprise work for enterprise budgets. The hawker stall in the Raffles Place basement (that’s a Singapore reference, in case you don’t recognise it) is not going to build an agent-callable loyalty API, and nobody is going to build one for it. The only way a small retailer becomes visible to an agent at all is if it sits inside a network that’s already callable. One integration, every program.

That’s what a consumer-side layer is for.

3 - Agents need something to be loyal to.

From the Loyalty & Reward Co article: “When agents shop, they optimise for a goal function… The brand story matters, but the agent needs it expressed as computable facts”.

Brian Scott of RAI Partners, talking about payments, said the answer is to make loyalty one of the instructions a consumer gives their agent. “Hey, I want to pay with my ABC Federal Credit Union card”.

I agree (I wrote about it on LinkedIn). But declared loyalty needs a home before an agent can read it. “Start with the independent café I love, don’t switch to save 20p” is only a usable instruction if there’s a record somewhere of which cafés I love and what I’ve already earned there. 15 retailer APIs don’t tell the agent that. One consumer wallet does.

So what is meed in an agentic world?

The same thing it is now, just a bit more talkative.

A consumer identity that carries relationships across every business the consumer cares about. A retailer-side program that gives small businesses their own loyalty without a POS integration, an app, or a data science team. And a privacy posture that means the retailer never gets the consumer’s personal data, and the consumer never gets farmed. The agent - ours, or the one the consumer brings - sits on top of that and reads it, with permission.

We’ll ship our own agent for retailers because it will be a better product than a dashboard. But it’s the roof, not the foundation. The foundation is a cross-business record of declared loyalty that no single retailer, and no single agent platform, can see on its own.

The fight worth fighting

Being a loyalty dashboard in 2026 is a bad place to be. Being the place where a person’s relationships with the businesses they love actually live is the better bet. Agents make that problem more urgent, not less.

Agents change the interface. They don’t change the architecture.

meed was always a bet on the architecture.

Subscribe now