For about sixty years we have drawn the customer journey as a funnel. Wide at the top, narrow at the bottom, with the customer falling out of the picture the moment they paid. It was a useful lie. It gave marketing teams a shape to organise budgets around, and it gave finance a way to argue about cost per acquisition.
That shape no longer describes what is actually happening.
What is happening now looks more like an infinity loop: awareness, discovery, consideration, purchase, post purchase, re-engagement, and then back into awareness again, with no clean end point. The change is not cosmetic. It is being driven by the fact that an AI agent can now carry a customer across every one of those stages without the customer ever leaving the conversation they started in.
Brand leaders who keep managing the funnel while their customers live in the loop are going to spend the next three years wondering why their attribution models stopped making sense.
Where the loop is actually breaking your current model
Discovery has moved off the search engine. Consumers are increasingly starting product journeys inside AI platforms rather than typing a query into a search bar and working through ten blue links. This is the quiet crisis for most marketing organisations, because a decade of SEO investment was built on the assumption that discovery happens on a results page you can optimise for. When an assistant synthesises an answer instead of listing sources, the question changes from "how do we rank?" to "how do we get cited?" Those are not the same discipline.
Lead generation has stopped being a handoff. AI agents now qualify and follow up on leads from ads and messaging automatically. The old choreography, where a form fill went into a CRM, sat overnight, and then got phoned by a junior salesperson at an inconvenient hour, is being compressed into a single continuous conversation. The lag between interest and response, which used to be measured in days, is now measured in seconds. Any brand still running on the old timing is losing customers it never knows it had.
Sales has collapsed into the thread. In-thread browsing, comparison and checkout mean the customer never bounces to a website, never abandons a cart in the way we understood cart abandonment, and never sees the carefully art-directed product detail page your team spent a quarter building. The agent is the storefront. Your product data is the shop window.
Support has become execution, not deflection. For twenty years support strategy was really cost-avoidance strategy dressed up in the language of care. Agents that resolve inquiries end to end, including troubleshooting and actually executing the fix, break that framing. Support stops being a cost centre that protects margin and becomes a retention engine that generates it.
Re-engagement runs on behaviour, not calendars. Behavioural triggers driving repeat purchase and lifetime value is not new as an idea. What is new is that the trigger, the message, the offer and the transaction can all be handled by the same agent, at the moment the behaviour occurs, without a campaign calendar or a batch send.
What this actually asks of brand leaders
Here is the uncomfortable part. Almost every capability described above sits in a different department, reports to a different executive, and is measured on a different number.
Discovery sits with marketing. Lead gen sits with demand gen or growth. Sales sits with commercial. Support sits with operations or customer experience. Re-engagement sits with CRM, which in most South African organisations is a two-person team quietly running email in a corner.
The loop does not respect any of those boundaries. A customer in an agentic conversation moves from discovery to purchase to support in ninety seconds, and they experience all of it as one interaction with one brand. If those five teams have five different tones of voice, five different sets of product facts and five different escalation rules, the customer does not see five teams. They see one brand that appears to be lying to itself.
So the first job is structural, not technological. Somebody senior needs to own the loop end to end. In practice this usually means a single accountable executive for the customer conversation, with the authority to arbitrate between teams that have historically been graded on competing metrics.
Your brand is now a machine-readable asset
This is the shift most brand teams have not absorbed yet.
For the past thirty years brand guardianship has meant controlling what humans see: the logo, the palette, the tone of voice, the photography treatment. Those things still matter. But an increasing share of your brand encounters are now mediated by a model that never sees your typeface.
What the model sees is your product data, your specification tables, your returns policy, your pricing logic, your documentation, your reviews, and whatever third parties have written about you. That corpus is your brand now, at least at the discovery and consideration stages. If it is thin, inconsistent or out of date, you will be summarised badly, and you will not get a right of reply.
Practically, this means the unglamorous work has become the strategic work. Clean product feeds. Accurate specifications. Documentation written to be parsed as well as read. A returns policy expressed in plain language rather than legalese. Structured data on every page. These used to be tickets in somebody's backlog. They are now brand infrastructure.
Trust is the whole game
When an agent can execute on your behalf, the brand promise stops being a claim and becomes a mechanism. If your agent says the item will arrive on Thursday and it arrives the following Tuesday, you have not made a marketing error. You have broken a commitment that a machine made in your name.
This has a governance dimension that most boards have not yet thought through. What is your agent allowed to promise? What can it discount? What can it refund without a human in the loop? Where does it stop and escalate? Who is accountable when it gets it wrong?
In the South African context there is a hard legal layer on top of this. POPIA obligations do not soften because the processing was done by an agent, and the Consumer Protection Act does not care whether the misleading statement came from a person or a model. If your agent is qualifying leads, it is processing personal information, and somebody needs to be able to explain the lawful basis for that. Get the compliance and legal teams into the design conversation early, not after the pilot has gone live.
Build for how South Africans actually transact
There is a temptation to import the global playbook wholesale. Resist it a little.
Local commerce runs heavily through WhatsApp, and it runs on customers who are conscious of data costs and network reliability. An agentic experience designed for a fibre connection and an unlimited data plan will fail quietly in a large part of this market. The loop still applies, but the surface it runs on is messaging rather than a slick web widget, and the design constraint is lightness rather than richness.
There is also a language reality. An agent that handles English fluently and isiZulu, Sesotho or Afrikaans poorly is not a neutral tool. It is a filter that decides which customers get a good experience.
Where to start
If you are leading a brand and this feels large, it is. But it is not a three-year programme before anything useful happens.
Start by mapping one real customer journey against the loop and marking honestly where a handoff currently drops the thread. Most organisations find two or three points where the customer is asked to start again. Fix those first, because they are also the points where an agent adds the most obvious value.
Then audit your product data as though a model were going to be quizzed on it, because one will be. Then decide, explicitly and in writing, what your agent may and may not commit to on your behalf.
The funnel let us treat the customer as something that arrives, converts and leaves. The loop asks us to treat them as someone we are in continuous conversation with. That is a harder standard. It is also, if we are honest about it, a more accurate description of what a brand has always been.
