For most of the past two decades, brand influence moved through a fairly predictable pipeline. You built awareness through media, earned attention through content and search, and converted that attention on a website or in a store. Marketers argued endlessly about attribution, but everyone agreed on the basic shape of the journey: a human being saw something, remembered something, searched for something, and eventually bought something.
That shape is now changing. Increasingly, the entity doing the searching, comparing and shortlisting is not a person at all. It is an AI agent acting on a person's behalf. Someone tells an assistant to find them the best medical aid option for a young family, or to book a weekend in the Midlands under a certain budget, or to reorder whatever washing powder offers the best value this month. The agent goes off, gathers information, weighs options and comes back with a recommendation or, in some cases, a completed transaction.
This is what people mean by the agentic era, and it quietly rewires how brand influence actually works. The brand no longer only needs to persuade a human. It also needs to be legible, credible and retrievable to a machine that is doing the human's homework for them.
The new middle layer
Traditional marketing theory treats influence as something that flows from brand to consumer, with media, retailers and word of mouth acting as channels. In the agentic model, a new middle layer sits between brand and buyer: the agent itself, along with the models, indexes and data sources it draws on.
This middle layer behaves differently from any channel we have dealt with before. A retail buyer can be wined and dined. A journalist can be pitched. A search algorithm can be reverse engineered, at least partially, through SEO. An AI agent is stranger. It synthesises information from thousands of sources, weighs them in ways that are not fully transparent even to its makers, and produces an answer with a confidence that most consumers will not think to question.
The practical consequence is that influence now flows through two parallel routes. The first is the familiar human route: advertising, sponsorship, social content, packaging, experience. The second is the machine route: everything an agent can read, verify and cite about your brand. Reviews, spec sheets, pricing data, news coverage, forum discussions, structured product data, regulatory filings, complaint records. All of it becomes raw material for the recommendation an agent eventually makes.
Brands that have spent years polishing the first route while neglecting the second are going to find the agentic era uncomfortable.
From share of voice to share of model
Marketers have long used share of voice as a proxy for future market share. Shout louder than your category, the logic goes, and growth follows. In an agentic world, a more useful concept is something like share of model: how often, and how favourably, a brand appears in the answers that AI systems give when consumers ask category questions.
Share of model is earned differently from share of voice. Volume matters less than verifiability. An agent asked to recommend a bank account for a student is not swayed by a catchy jingle. It is swayed by fee structures it can parse, terms it can compare, complaint patterns it can detect and independent commentary it can weigh. A brand with a strong television presence but thin, inconsistent or contradictory information online may find itself invisible at exactly the moment a decision is being made.
There is a South African wrinkle here worth naming. Much of our commercial life still runs on relationships, WhatsApp groups, community recommendations and informal trade. Agents will not replace the stokvel conversation or the neighbour's advice overnight. But for the growing share of decisions that start with a query typed into an assistant, especially in banking, insurance, telecoms, travel and consumer electronics, the machine route is already carrying real weight. And because agents tend to compress a market's options into a shortlist of two or three, the penalty for being left off that shortlist is severe.
Trust gets outsourced, then concentrated
One of the more profound shifts in the agentic era is what happens to trust. When a person delegates a decision to an agent, they are also delegating the trust judgement. They may never see your website, your advertising or your carefully crafted brand story. They see the agent's summary of you.
This concentrates enormous influence in how agents assess credibility. Signals that were once background hygiene, such as accurate business listings, consistent pricing across channels, responsive complaint handling, honest product data and a clean record with regulators, become primary drivers of whether a brand gets recommended at all. The unglamorous plumbing of a business starts doing the work that campaigns used to do.
It also means that reputation problems compound faster. A pattern of unresolved Hellopeter complaints or a string of critical Reddit threads used to be a slow leak, damaging mostly among people who went looking. An agent goes looking every single time. What was once a minor reputational drag becomes a standing objection raised in every comparison your brand appears in.
The flip side is genuinely encouraging for challenger brands. An agent does not care that your competitor has been on billboards since 1994. If your product is demonstrably better value and the evidence is well documented and easy to retrieve, an agent can put you on the shortlist next to giants. Machine intermediaries flatten some of the advantages that pure spending power used to buy.
What this asks of marketers
None of this makes classic brand building obsolete. Humans still set the brief. A person who tells their agent to find them the best trail running shoes is influenceable long before that instruction is given, and a person who says find me Salomon trail shoes has already been influenced. Memory, emotional association and cultural presence still determine which brands get named in the prompt rather than discovered by the agent. In some ways, being the brand the consumer asks for by name becomes more valuable, not less, because it bypasses the machine's comparison altogether.
What changes is the second half of the job. A few priorities stand out.
First, treat your public information as a product. Pricing, specifications, availability, terms and support policies should be accurate, structured, current and consistent everywhere they appear. Contradictions between your site, your retail partners and third party listings do not just confuse consumers anymore. They lower a machine's confidence in you.
Second, invest in earned credibility that machines can read. Independent reviews, expert coverage, industry certifications, published test results and transparent responses to criticism all feed the evidence base agents draw on. This is public relations rediscovering its original purpose, which was never coverage for its own sake but the accumulation of third party proof.
Third, measure differently. Alongside brand tracking and search visibility, brands need to start auditing how AI systems actually describe and recommend them. Ask the major assistants the questions your customers ask, in the language your customers use, and study what comes back. If the answers are wrong, outdated or unflattering, that is a strategic problem, not a curiosity.
Fourth, fix the operational truths. Agents surface patterns. If your delivery times are poor or your call centre is a horror show, no amount of content strategy will hide it from a system built to find exactly that kind of signal. In the agentic era, the honest brand and the well marketed brand need to be the same brand.
Influence becomes an ecosystem property
The deepest change is philosophical. Brand influence used to be something a company projected outward, largely on its own terms. In the agentic era, influence becomes a property of the ecosystem of information surrounding a brand, most of which the brand does not control and cannot buy. It flows from the accumulated record of what a company actually does, as read and interpreted by machines whose loyalty lies with the consumer.
That should worry brands built mostly on image. It should excite brands built on substance. And for everyone in between, it sets the agenda for the next decade of marketing: earn the human's preference, and make sure the machine can prove you deserve it.
