
“Find me a business bank account. Keep the fees down. Make sure it works with my accounting software.”
Imagine giving that brief to an AI agent rather than opening ten browser tabs. It compares the options, explains the trade-offs and, where permitted, helps you take the next step. The bank’s homepage might never get a visit. Its beautifully produced launch film might never get a view.
So who is the advertising for?
For brands investing in video production, this is a more useful question than which AI model can generate the most convincing footage. AI is changing how content is made. It may also change who encounters it, when they encounter it and what persuades them.
Our view: the future of video advertising has two connected jobs. Make the product easy for an agent to assess, and give a person a reason to want, understand or trust it. The balance will vary by purchase. There is no evidence that every category is heading towards the same agent-led journey.
Parts of this shift are already visible. In its analysis of browsing by 900 US adults, Pew Research Center found that traditional result links received clicks on 8% of Google visits with an AI summary, compared with 15% without one.
That is a snapshot of March 2025 behaviour, with search results collected the following month. It does not establish causation, measure video advertising effectiveness or predict UK buying behaviour. It does show why marketers should question the assumption that discovery always leads to a website visit.
Assistance is also moving closer to transactions. Google announced the US rollout of Universal Cart across Search and Gemini in October 2026, helping shoppers organise products from different retailers and complete purchases. That is an emerging capability in specific markets, not proof that consumers have handed over all their decisions.
The distinction matters. Asking an LLM for a shortlist, approving an agent’s recommendation and delegating an entire purchase are different behaviours. A video strategy should be clear about which one it is responding to.
There is a credible argument that some purchases will need less advertising.
If I have already specified the exact product, maximum price and delivery date, an agent may only need to check availability and transact. A persuasive film adds little to that task. The same logic could apply to routine software renewals or tightly specified business services.
For a bank or SaaS provider, a stylish advert is unlikely to help an agent reconcile unclear fees, missing integrations or conflicting product information. If the buying brief is entirely functional, accurate facts may do more work than a cinematic brand promise.
That could put pressure on generic explainer videos that simply turn a feature list into moving pictures. If an assistant can answer the question more quickly, brands will need a stronger reason to ask for someone’s attention.
We should take that possibility seriously. We should also avoid turning a plausible scenario into a prediction that video advertising is finished.
“Find the cheapest account” and “find an account I would trust with my business” are different briefs.
People bring expectations, memories, preferences and concerns to a purchase before an agent compares anything. Advertising can help shape those preferences. A brand someone already recognises may be named in the brief, investigated more closely or questioned less quickly. Those are hypotheses to test, not guaranteed effects of an agent-led buying journey.
There is also a difference between checking eligibility and making a commitment. An agent may compare subscription prices. A business owner may still want to see the software solving their problem. A finance team may want to hear from a credible customer before changing a system that everyone relies on.
Banking is not an emotion-free category simply because the product has no fabric, fragrance or finish. Money involves confidence, risk and identity. The creative challenge is to make those concerns understandable without substituting reassurance for evidence.
For fintech brands, the challenge is not to make a bank card look more desirable. It is to make the service understandable, distinguishable and worth trusting when an assistant may handle the first comparison.
That suggests two connected tracks for fintech marketing. Keep the facts useful for comparison: who the product serves, what it costs, how it works and where its limits lie. Then use creative advertising to make the proposition matter to a person. A business owner wants to know what changes in their working day, not just which features appear in a table.
Video can show a real payment workflow, explain an unfamiliar service or bring a customer's experience to life. Brand campaigns can give a functional proposition a memorable idea. The point is not to film a shinier piece of plastic; it is to make the benefit and the reasons to believe it clearer.
Advertising inside AI assistants is already a live question. OpenAI began its ChatGPT ads test in February 2026 and announced a UK launch in August. It also introduced a beta self-serve Ads Manager and cost-per-click buying in May. This is an emerging media channel, rather than a platform that has only just launched this week.
OpenAI says paid placements are labelled and separate from ChatGPT's answers. Buying an ad is not buying an organic recommendation. Nor should a video production brief assume that every platform offers the same creative formats.
Claude takes a different position. Anthropic says Claude remains ad-free, while supporting user-initiated ways to discover and interact with businesses. There is no equivalent Claude advertising platform to recommend here. Being useful when an assistant researches a category is a different task from buying media inside it.
For fintech advertisers, availability is only the first check. OpenAI's ad policies and category-and-country eligibility guidance restrict financial services; financial software is treated differently from products such as lending or investing. Check the current rules and the specific offering before allocating budget. A ChatGPT launch in the UK does not mean every UK fintech can advertise there.
Our recommendation is to prepare a focused experiment where eligible: one real customer question, one substantiated proposition and a landing page that answers it. Measure qualified enquiries or suitable sign-ups, rather than celebrating cheap clicks. Keep building brand preference through video in channels where the audience actually watches, and make the supporting facts accessible when assistants research the product.
The opportunity is to learn how conversational discovery changes the customer journey. It is not a reason to move a whole media budget on the strength of a new platform announcement.
A spreadsheet can tell you the composition of a coat. It cannot fully convey how the fabric moves, how it falls on different bodies or what wearing it might feel like. A beauty product’s ingredients do not show its texture or finish. An artwork's dimensions and materials cannot replace seeing its composition, surface and presence in a space.
Video can demonstrate those qualities in motion. It can also establish a world around a product: who it is for, where it belongs and what makes it desirable. For these purchases, an assistant may narrow the options while a person still wants to see them.
But the divide is not simply aesthetic products versus functional products. Fashion also involves price, sizing and returns. SaaS involves usability, workflow and confidence in the people behind it. The more useful distinction is between what can be specified, what needs to be demonstrated and what someone wants to feel.
A single purchase can contain all three.
A distinctive campaign can introduce a brand and make its point of view memorable before someone starts a buying task. This is where character, story, sound and recognisable brand assets matter. Producing more interchangeable footage will not solve a lack of creative direction.
Use film to demonstrate movement, interaction, process or experience. For a software brand, that might be a real workflow from problem to outcome. For a beauty brand, it might be an honest application sequence under consistent lighting. The test is whether seeing it changes someone’s understanding.
Customer stories, product demonstrations and clear explanations can help a buyer evaluate a promise. They need enough specificity to be useful: who is speaking, what happened, what is being shown and which limitations apply. A confident voiceover is not a substitute for substantiation.
There are already experiments that combine these worlds. In September 2026, Google invited eligible US retailers into a Business Agent for YouTube ads beta, letting viewers ask product questions within the viewing context. The important point is the design direction: conversational assistance can sit alongside video. It need not replace it.
AI-generated video can make ambitious ideas more achievable and help teams produce variations. It does not decide which idea deserves to exist.
Our Stratiphy launch campaign is a useful example. A fintech product gave us a functional subject, but the creative idea used contrasting characters and a surfing scenario to make the importance of timing tangible. AI was the production approach; the story gave the film its purpose. That project is not evidence that the same campaign would succeed in a future agent-led market.
Our Sage mentor films show another role for video: helping people understand a journey. AI-generated characters and localisation made it possible to carry a consistent idea across languages and regions. This is different from buying media to generate awareness, but it matters to the broader question of where video remains useful.
For brands planning AI video production, the brief should start with the audience’s unresolved question. Then decide whether the right execution is filmed, generated, animated or a combination. If the product’s actual appearance or behaviour is the evidence, preserve that truth. Do not generate a more flattering demonstration and pass it off as the product.
Before commissioning another batch of assets, ask:
These practices make content easier to understand and reuse. They do not guarantee that an LLM will retrieve, cite or recommend a brand. No production company can credibly promise that a particular video format will secure an AI recommendation.
The UI question follows naturally, but deserves its own article. If an agent handles more routine tasks, people may use fewer screens. Yet they still need ways to set permissions, inspect a recommendation, understand a charge, challenge a mistake or take control.
Our hypothesis is that some interfaces will shift from helping people perform every step to helping them supervise delegated work. That creates questions for product design, trust and brand communication. It does not make the interface disappear, and it does not remove the need to explain a product clearly.
If more buying decisions pass through AI, brands cannot rely on a homepage visit to tell their story. They will need clear information that can travel through systems, and creative work that gives people a reason to care wherever they encounter it.
Some video will become less useful. Some will become more closely connected to product questions and decisions. We do not yet know the balance.
What we can do now is stop treating video production as a race to fill more channels. Make the film that changes someone’s understanding. Build the evidence around it. Give it a recognisable idea worth remembering.
Planning a campaign for a changing customer journey? Talk to Sidekick about creative strategy, video advertising and AI-powered production.
Perspective researched on 17 November 2026. Platform announcements describe capabilities and rollouts, not universal adoption. Predictions and recommendations above are Sidekick’s editorial view.