.jpg)
A phone, a capable AI assistant and a handful of well-chosen integrations can now give even a tiny business marketing capabilities that, five years ago, might have required an agency charging thousands of dollars a month. Not every capability, and certainly not the same judgment or accountability. But enough to change the economics of running a small company. If you have been waiting for AI to become genuinely useful in marketing, rather than merely entertaining, your moment may have arrived.
The intriguing question is what happens next. When a local contractor can publish polished video, a neighborhood retailer can interrogate its advertising data, and a solo founder can operate a content calendar from a phone, competition gets tougher. Our bet is that this could be good news for customers. Once competent advertising is no longer scarce, businesses may have to compete harder on the things advertising cannot manufacture: a better product, better service and a relationship worth returning to. Think of it as Tribes squared—a new technological chapter for the communities Seth Godin wrote about in Tribes in 2008.
In this article, you'll learn
- Why AI marketing has changed so dramatically since 2022—and what the Will Smith spaghetti videos reveal about that progress.
- How AI can help run a real Webflow or WordPress site, build landing pages and handle connected marketing tasks.
- How to approach SEO, GEO and keyword planning with Ahrefs, Search Console, real customer intent and human editorial judgment.
- How to produce better Reels and marketing videos using your own likeness or real phone footage with AI-assisted production.
- What AI adoption data actually says, how to build a consistent visual system, and where human creativity still matters.
- What an AI marketing stack costs, how advertising economics are changing and why genuine customer relationships may become the biggest competitive advantage.
Remember Will Smith eating spaghetti?
In 2023, an AI-generated clip of a man resembling Will Smith eating spaghetti became an unlikely cultural landmark. Faces shifted, hands behaved strangely and dinner looked like an encounter with an alien life form. It was funny because the machine could almost do something human—and spectacularly failed at the last few inches.
Now compare that with more convincing generative video from 2025 and 2026. Motion, image consistency and scene control have improved enormously, although failures and carefully curated demos remain part of the story. The two examples below illustrate that progress; they are not a controlled, like-for-like benchmark.
The point is not spaghetti. It is the speed with which an amusing experiment became a serious production tool. Marketing has followed a similar trajectory—although its most important breakthrough has less to do with prettier images than with connecting AI to the actual work of a business.
Five years from prompts to processes
The first wave of generative AI made it easy to draft a caption, summarize a meeting or brainstorm twenty slogans. That was useful. It also produced an ocean of interchangeable material: enthusiastic LinkedIn posts, plastic-looking illustrations, and emails that sounded as though they had been written by the same relentlessly cheerful stranger.
By 2024 and 2025, the tools were moving beyond text. Image generation improved, video became more practical, models could work with more kinds of inputs, and companies began experimenting with AI assistants that could use external tools. In 2026, the important question is no longer simply what can the model generate? It is what can it read, what can it safely do, and who approves the result?
| Period | What changed | Marketing consequence |
|---|---|---|
| 2022–2023 | Consumer generative AI and early text/image tools | Fast drafts, ideas and experiments |
| 2024 | Improving multimodal models and creative tools | Better visual assets and richer workflows |
| 2025 | Broader agent and integration ecosystems | More connected analysis and task execution |
| 2026 | More capable, still uneven connected systems | Practical marketing operations with human supervision |
Illustrative timeline of broad developments; exact feature availability varies by product and plan.
AI adoption is booming. Actual deployment? That's another story.
Here's an interesting paradox. McKinsey's 2025 global survey found that 88% of surveyed organizations used AI in at least one business function. But the U.S. Census Bureau's survey of American businesses recorded roughly 17–20% reporting AI use during the previous two weeks between December 2025 and May 2026. These are different surveys with different populations, questions and definitions—not directly comparable market shares. Together, they suggest a useful distinction: talking about AI, experimenting with AI and routinely putting it to work across a business are not the same thing.
Illustrative comparison of two different measures, not a common denominator or a time series. Sources: McKinsey, State of AI 2025; U.S. Census Bureau, Business Trends and Outlook Survey, May 2026.
Stanford's 2026 AI Index also reports that organizational adoption reached 88% in 2025, while AI agent deployment was still in the single digits across nearly all business functions. In other words, the opportunity isn't just to use AI. It's to connect it to work that matters.
From curiosity to the mainstream: organizational AI adoption
Share of surveyed organizations using AI, 2023 and 2025. Source: Stanford AI Index 2026, Economy, drawing on organizational survey data. This is not the percentage of all U.S. small businesses using AI.
The AI slop era was a strategy problem
When creating something becomes cheap, the temptation is to create everything. A small company that once struggled to publish twice a month can suddenly produce two posts a day. But frequency is not a positioning strategy, and a calendar full of content is not evidence that anybody cares.
The most damaging AI output is not necessarily ugly. It is plausible, polished and forgettable. It says what every competitor says, borrows the same visual language and makes promises no customer would recognize as distinctive. That is why the future of AI marketing depends as much on taste, customer knowledge and restraint as on model quality.
Your marketing department is becoming a conversation
Imagine a small home-improvement business in New Jersey. The owner has fifteen minutes between appointments and asks: “What happened to our leads this week?” A connected assistant can help review analytics and campaign data, compare performance with a prior period, identify which landing pages lost visitors and suggest questions worth investigating. Later, the owner asks for a draft follow-up email and a new article addressing the question customers keep asking.
That workflow is possible in principle today, but the integrations matter. ChatGPT does not automatically have access to every Google Ads account, CRM or website, and not every connector can publish or change settings. Reading data, proposing an action and executing that action are three different levels of capability. Permissions, platform support and approval steps determine what actually happens.
This is the useful meaning of an AI marketing agent: not a magic employee, but a connected assistant operating inside a well-defined system. An AI automation agency or consultant can help build those connections, document the rules and make sure the workflow remains manageable when something breaks.
Why most AI marketing still sucks: the human advantage
Here is our answer to the question in the headline: much of AI marketing still sucks because it is made for people who want to skip the thinking. Someone gets an idea on the way to work, types a two-line prompt, waits two minutes and publishes whatever comes back. No customer insight. No creative direction. No attempt to decide whether the idea is good in the first place. AI slop is the predictable result of treating speed as a substitute for judgment.
The temptation is understandable: when production becomes almost effortless, the uncomfortable work of researching, questioning, rewriting and editing starts to feel optional. It isn't. The businesses that benefit most from AI will not necessarily be those generating the most content. They will be the ones putting the most thought into what they ask it to produce—and having the discipline to reject the mediocre results.
I was already writing about the human role in AI back in 2021, and I haven't changed my mind about the important part: people remain responsible for what they ask technology to do. What has changed is the usefulness of the technology itself. For a long time, “AI saves you time” often meant “AI gives you a draft and a fresh list of things to fix.” Increasingly, it can complete real steps in a real workflow. And that, rather than the latest spectacular demo, is why I'm optimistic.
A business needs to decide whom it serves, why people should choose it, what it stands for, how much it can spend and what it refuses to promise. Those decisions draw on lived experience, customer conversations and responsibility for the outcome. AI can help research alternatives and challenge assumptions. It should not be handed the authority to invent a brand's purpose.
| Human-led foundation | AI-assisted work |
|---|---|
| Positioning, customer understanding, differentiation | Research summaries, audience questions, message variants |
| Creative concept and editorial judgment | Drafts, formats, adaptations and production assistance |
| Brand identity and quality standards | Consistent application of templates and guidelines |
| Budget, ethics and final accountability | Monitoring, alerts and recommended optimizations |
| Real customer relationships | Useful reminders and better-informed service |
Our shorthand is the human as creator, AI as reproducer. It is a metaphor, not a literal claim that AI cannot produce original combinations or ideas. The distinction that matters is who supplies the intent, accepts the risk and recognizes what is worth making.
Running a website with AI in 2026: yes, an actual website
Let's begin with something decidedly unglamorous: your website. Because before AI starts writing cinematic scripts and generating photorealistic astronauts, somebody still has to make sure the contact form works.
I work with Webflow, and here's what our collaboration on the Startup Harbour website looks like in practice. A human decides what the business does, who the page is for, how the offer should be presented and what the visual hierarchy ought to be. The original design system, components, fonts, colors and page structure are built with those decisions in mind. That's the foundation. Once it's there, a connected AI assistant can help change headlines, edit copy, build a new landing page from established components, adjust sections, update metadata and prepare a page for publishing. With the right access, it can make those changes in the actual site rather than merely handing you a document full of suggestions.
For example, when we need a new article or landing page, I can describe the angle, the audience and the conversion goal. The assistant drafts the content, works within our existing Webflow structure, adds sections and adjusts SEO settings. I review the argument, design and facts. We iterate. When the result is ready, publishing is a deliberate step. That's not a hypothetical workflow; it's the way we're working on this very article.
WordPress can support a similar workflow through its publishing APIs, plugins and automation services. But let's not pretend every WordPress site is equally easy to operate. Themes, page builders, permissions and plugins introduce differences. A Gutenberg-based site and a heavily customized Elementor installation may require very different methods. Some changes are straightforward API operations; others need browser-based editing, developer work or a careful human check.
A practical workflow for your next landing page
- Human: Define the offer, audience, promise, proof, call to action and approved brand components.
- AI: Draft page copy, suggested layout, FAQ, metadata and variants of the headline.
- Connected workflow: Create or duplicate a page in Webflow or WordPress, insert content and configure links, forms and SEO fields where supported.
- Human: Check mobile layout, tracking, legal claims, links and the actual conversion path.
- AI or human, with permission: Publish, then monitor traffic and conversion data.
The breakthrough isn't that AI can produce HTML. It could do that before. It's that the distance between asking for a change and seeing that change on your actual website is getting much shorter.
SEO, GEO and a rather uncomfortable question about writing
Here's a problem I still haven't solved for myself. There are two kinds of articles I can write. The first is the kind I actually want to write: a personal essay, an opinion, a little storytelling, maybe a provocative thought that has been bothering me for weeks. I'll enjoy writing it. Five people might read it. Three of them might be my friends.
The second is built around what people are searching for. We check Google Search Console, investigate keywords in Ahrefs, look at search demand and competing pages, study the questions customers actually ask, and build something genuinely useful around that evidence. I might be less romantically attached to the result. Google and AI search systems may be among its most attentive readers. But the article can bring somebody to the website who is actively looking for the service we sell.
So which is the better article? I don't have a universal answer. Everyone gets to decide what they want their writing to achieve. But speaking as a marketer rather than a frustrated columnist, the business result matters more than my personal satisfaction with a beautiful paragraph. Ideally we can have both. In reality, they sometimes compete for the same afternoon.
This is where AI becomes extremely practical. Connect the research process to actual data: Ahrefs for keyword opportunities, backlink analysis and competing content; Search Console for the queries your pages already appear for; GA4 for what visitors do after they arrive. Ask the assistant to identify topics where demand, your expertise and commercial intent overlap. Then let it help outline, research, draft, link and update the content. The human still needs to challenge the argument, verify the numbers and decide what is worth saying.
And GEO—generative engine optimization? Treat it as an extension of being discoverable and understandable, not a magical new science. Clear answers, credible sources, useful structure, original expertise and technically accessible pages can help people and search systems understand your work. Nobody can honestly guarantee that a particular AI assistant will recommend your company.
One practical distinction worth making: SEO keyword research and paid-search keyword planning are related, but they aren't the same job. In Ahrefs, investigate what people search for, who currently ranks, whether the page can satisfy that intent and whether you have a credible chance of competing. In Google Ads Keyword Planner, explore paid-search demand and bid estimates—but remember that suggested bids are not your actual CPC and search volume is not a lead forecast.
For example, “AI marketing agent” may sound like a promising topic, but before writing a 2,000-word article, ask three questions: Who is searching? What are they trying to accomplish? And do we sell something that solves that problem? If the answer to the third question is no, the traffic may flatter your analytics and do absolutely nothing for the business.
Build a simple content brief with the target query, search intent, business relevance, existing ranking pages, internal links, original examples and a conversion goal. Use AI to accelerate research and production, then judge the result by qualified enquiries rather than keyword density. Ahrefs API access and usage allowances depend on the subscription and integration; its official documentation explains the available keyword data and API unit costs.
A repeatable SEO and GEO workflow
- Pull actual queries and page performance from Search Console.
- Use Ahrefs to evaluate related terms, difficulty, search intent, competitors and links.
- Prioritize topics with a credible connection to your product or service—not just impressive search volume.
- Have AI prepare a sourced outline, then add human observations, examples and claims worth checking.
- Publish through your CMS with internal links, readable headings, metadata and structured information where appropriate.
- Review indexing, impressions, clicks, qualified leads and assisted conversions; update what underperforms.
One important caveat: Ahrefs data doesn't automatically flow into ChatGPT because you bought an Ahrefs subscription. You need a supported connector, API access, export or automation workflow, and the available data depends on your plan and permissions.
AI graphics: the human is the creative director, the machine is the craftsman
Let's talk about the endless parade of shiny AI-generated images. I suspect most people are asking the wrong question. They want to know which image model is the best. My answer? Pick a strong model from the leading group, test whether it handles your particular work, and stop obsessing over the leaderboard.
The model isn't your brand. Your brand is the set of decisions you made before opening the image generator. The logo, typography, palette, composition, visual references, subject matter, proportions, sense of humor and—perhaps most importantly—what you refuse to publish.
We recently worked on a series of before-and-after concepts for Dateonic, a data consultancy. The creative idea was ours: an unruly favela contrasted with orderly Viennese streets to explain data architecture; water eroding stone contrasted with a dam to explain data flow; a chaotic intersection versus a carefully designed interchange to explain data integration. Those metaphors, and the brand's green palette, came first. AI helped produce variations within that direction.
It's a similar story with Startup Harbour. Once the visual language and templates exist, AI can help replicate the system across article covers, campaign assets and formats. This isn't about asking a machine to invent a different personality for your business every Tuesday. It's about creating the backbone once and using AI to produce consistently within it.
The practical graphic production loop
- A human creates the concept, logo, brand palette, typography rules and one approved master composition.
- AI generates variations for specific messages, placements and aspect ratios.
- The human reviews legibility, consistency, rights, accuracy and whether the metaphor makes any sense.
- AI adapts the approved version into social, display, blog and landing-page formats.
- Approved files go into the asset library, ready for the next campaign.
You can use ChatGPT's image tools, Midjourney or another capable model. Results and controls differ, especially for exact typography and layout. The creative direction matters more than the novelty of the tool. Human: creative director. AI: very fast production studio. A surprisingly productive arrangement.
The new video studio fits in your pocket
Video is where this change becomes visible. There are now two compelling approaches for a small business owner who does not have a production crew. Most importanlty...no green screen. Just record in your home and make it the White House in a few clicks :)
1. Your likeness, an AI-produced scene
With tools that support authorized likenesses or reference footage, a creator can place their own image or performance into AI-assisted scenes and templates. That can make imaginative advertising more accessible. The result still needs review for uncanny details, authenticity, platform disclosure rules and rights to the likeness, music and source material.
2. Your real footage, AI-assisted post-production
This is often the stronger option. Record a genuine explanation on your phone, at home or in the shop. Use software to help trim pauses, clean audio, add captions, reframe for vertical video, produce supporting graphics and generate illustrative cutaways. A good short-form video still needs a hook, a point of view and a reason to watch. AI can help it look more polished; it cannot guarantee it will be interesting.
Midjourney is useful for exploring visual directions and creating still imagery. Runway and other video platforms offer different generation and editing workflows. The best choice depends on the rights, controls, consistency and costs required—not on whichever demo looks most cinematic.
One conversation to manage the whole marketing operation
Now for the part that excites me most. Not the graphics. Not the videos. The possibility of asking a normal question about your business and getting an answer based on your actual data, without opening twelve browser tabs and pretending to enjoy dashboards.
Imagine connecting GA4 for traffic and conversions, Google Search Console for organic visibility, Ahrefs for keyword and competitor research, Google Ads and Meta for campaign performance, your CRM for lead quality, and Webflow or WordPress for publishing. A data connector such as Windsor.ai can help consolidate marketing data for reporting and analysis. Workflow platforms such as Make, Zapier or n8n can help move information between services or trigger supported actions. The assistant becomes a conversational layer over a carefully designed system.
Monday morning. Coffee. Phone. “Which campaigns brought us qualified leads last week, not just cheap clicks?” The assistant can compare spend with tracked conversions and CRM outcomes if those systems are correctly connected and the definitions match. “What should we write next?” It can combine existing Search Console performance with keyword research imported from Ahrefs. “Prepare a landing page for our new service.” It can draft it and, where the publishing integration supports it, create a draft in the CMS. “Increase the budget.” That's where permissions, limits and an explicit approval step should kick in.
None of this happens just by typing “you are my CMO” into a blank chat. Someone has to establish the tracking, authorize the integrations, define what counts as a lead, and decide which actions can be automated. In other words, the exciting part is not the prompt. It's the system behind the prompt.
| Question from the owner | Information the system needs | Useful outcome |
|---|---|---|
| Which marketing channel actually brought customers? | GA4 + ad platforms + CRM | Lead and customer performance, not vanity metrics |
| What should we publish next? | Search Console + Ahrefs + business priorities | Topic shortlist based on intent and opportunity |
| Why did our campaign get expensive? | Google Ads / Meta + conversion history | Diagnosis and proposed changes for approval |
| Can we launch a landing page this afternoon? | Brand system + CMS access + tracking setup | Draft, review, QA and controlled publication |
| Are we becoming more visible in AI search? | Search visibility data + defined monitoring method | Directional evidence, not invented attribution |
What to connect first
- Make sure GA4, Search Console and conversion tracking are correctly configured.
- Connect advertising data and confirm spend, campaign and conversion fields match.
- Add Ahrefs through a supported API, connector or regular export if the plan allows it.
- Bring in CRM outcomes so “lead” doesn't automatically mean “good lead.”
- Connect the CMS with least-privilege access and keep publishing or budget changes behind approvals.
- Test the workflow on a real question and verify the answer against the original source before relying on it.
A final practical note: Windsor.ai is a data integration and reporting layer, not a universal remote control for every ad account. Likewise, the existence of an API does not mean every AI assistant can use it out of the box. Read access, write access and reliable execution are separate capabilities. The point isn't to sell a fantasy of effortless autonomy. It's to build something useful enough that you genuinely get some of your time back.
The actual marketing stack: tools, integrations and prices
There is no single universal “AI marketing plugin.” A workable system is assembled from an assistant, data sources, connectors, publishing tools and approval workflows. Some parts are free; others charge subscriptions, usage fees or both. The figures below are illustrative planning references, not a live price quotation. Always check the linked vendor pricing pages before purchasing.
| Tool | Role | Cost model / reference |
|---|---|---|
| ChatGPT | Assistant, analysis and connected workflows | Plan-dependent; official pricing |
| Google Analytics 4 | Traffic and conversion measurement | Standard version free; details |
| Google Search Console | Search queries, indexing, SEO diagnostics | Free; details |
| Google Ads | Paid search and other campaign channels | Media spend plus optional management; details |
| Meta Business Suite | Facebook and Instagram management | Core suite free; ads charged separately; details |
| Windsor.ai | Marketing data connectors and reporting | Tiered subscription; official pricing |
| Make / Zapier / n8n | Workflow automation between services | Free or paid tiers, usage-dependent; Make, Zapier, n8n |
| Webflow / WordPress | Website and publishing | Webflow site plans or WordPress hosting/plugins; Webflow |
| Midjourney | Image concepts and assets | Paid subscription tiers; plans |
| Runway | Video generation and editing | Plan and credit-based limits; pricing |
| CRM | Lead capture and follow-up | Vendor and seat-dependent; integrations vary |
A free analytics account is not the same as a free, functioning AI integration. Data mapping, permissions, authentication, maintenance and quality checks all take work. Before buying an automation subscription, confirm whether the connection can only read data or can also write changes back to the source system.
What does it cost to build?
The honest answer is that the software bill is only part of the cost. A solo owner might combine free analytics, an existing website and one paid assistant. A more connected setup adds an integration platform, automation usage and creative tools. A mature operation may also need custom development, governance and ongoing support.
| Scenario | Typical stack | What drives the bill |
|---|---|---|
| Solo business | AI assistant + free Google tools + existing website | Assistant subscription and hosting |
| Small team | Above + connectors + automation + design/video tool | Seats, tasks, data volume, creative credits |
| Advanced system | Above + CRM + custom workflows + support | Implementation, monitoring, maintenance and people |
Compare costs in three separate buckets: software subscriptions, implementation and ongoing human work, and advertising spend. A low monthly tool bill does not mean a complete marketing operation is free.
The economics of attention have not disappeared
AI can lower the cost of preparing a landing page, an image or a first draft. It does not make attention free. A click on a competitive search term still has a price; acquiring a qualified lead still depends on intent, geography, competition and conversion. Lower production costs may even encourage more businesses to advertise, increasing the pressure on scarce attention.
In our earlier Google Keyword Planner research, the phrase “AI automation agency” showed approximately 390 monthly searches and an 88% year-over-year increase in the selected report. “AI marketing agent” showed about 170. These are rounded, location-dependent estimates from a prior planning exercise, not universal US demand figures or guaranteed traffic. More importantly, the searches express different needs: someone looking for an automation agency may want sales operations, finance or CRM rather than marketing.
| Keyword in prior planning report | Approx. searches / month | Year-on-year change |
|---|---|---|
| AI automation agency | 390 | +88% |
| AI automation consultant | 50 | +200% |
Source: Startup Harbour's previously collected Google Keyword Planner results. Geographic targeting, report date and currency settings should be confirmed before using these figures as market-wide evidence.
Keyword demand is a useful starting point, not a forecast of sales. Search volumes depend on geography, period and keyword settings; competition, click prices and lead quality require separate evidence. The sensible experiment is to track qualified leads and customers—not to celebrate a spreadsheet full of searches.
When everyone can market, what happens to competition?
If more businesses can build respectable websites, produce credible video and understand their campaign data, the baseline rises. That does not mean every local company becomes a world-class marketer. It means fewer can rely on their competitors being unable to do the basics.
We see a hopeful possibility. As polished communication becomes easier to produce, customers may pay closer attention to whether a company answers the phone, keeps its promises, fixes mistakes and delivers a service worth recommending. Advertising can introduce a business. It cannot make a disappointed customer loyal.
Tribes squared: relationships after the content explosion
In Tribes, Seth Godin argued that people organize around shared interests, ideas and leaders rather than merely receiving mass messages. AI makes publishing and personalized communication cheaper. It does not create belonging on demand.
That is the paradox at the heart of the new marketing era. The more content machines can produce, the more distinctive a genuine human relationship may become. A neighborhood shop that remembers its regulars, a service provider who explains an awkward decision honestly, or a founder who shares the work behind the product has something no generic content calendar can reproduce: earned trust.
Call it Tribes²: technology increases the reach of a community, while human attention determines whether that community deserves to exist. It is our interpretation of where the market could go, not a claim that Godin coined the term.
The future of marketing might be more human than we think
We began with an AI-generated spaghetti dinner because it captured a moment when technology could imitate the surface of human behavior but not convincingly hold it together. Today the surface is far better. The harder challenge is giving the system a purpose.
The opportunity for small businesses is not to remove people from marketing. It is to remove some of the technical and production barriers between people and their ideas. A founder can make better-informed decisions, publish more confidently and spend more time talking to customers—provided the business builds the right foundations and keeps a person accountable.
Your business. Your strategy. Your AI marketing system. At Startup Harbour's AI Marketing Agent, we help businesses design those foundations, connect the tools and turn a chat interface into a practical marketing workflow without surrendering the human touch.
What's next?
For now, AI is a remarkably loyal researcher. Ask it to investigate a market, compare competitors or organize a mountain of information and it can be genuinely useful. Ask it what all that information means for your particular business, and things get a little more interesting. It can already produce decent positioning matrices. But connecting the strategic dots—finding the one insight that changes the whole plan—is still much less reliable.
I suspect that will change. At some point, I can imagine an AI system studying your customers, competitors, margins, ambitions and resources, then proposing a marketing strategy that actually fits your business. Not another list of twelve generic tactics. A strategy. One that knows why you should not do certain things.
But there's another upgrade I'd like even sooner: critical thinking and the ability to argue with us. AI is often far too eager to agree with its operator. We present a questionable idea, and it responds as if we've just reinvented marketing. Wonderful for the ego. Less wonderful for the business. A genuinely useful marketing partner should occasionally tell me, “No. That's a terrible idea, and here's why.” Preferably before I spend the advertising budget.
Those capabilities look achievable to me. Emotional understanding is a trickier question. AI can already recognize emotional cues and generate words or images that move people. But recognizing a feeling, reproducing its language and living that feeling are not the same thing. I don't believe machines will replace the human experience behind the stories that matter most to us. And as long as people buy with their hearts as well as their heads, marketers still have a very good reason to get out of bed in the morning. Emotion is not a decorative layer of marketing. It's the center of it.
Personally? I'm especially excited about what happens next in video and animation. I want the tools to become good enough that a small business with a great story can tell it beautifully without needing a Hollywood budget.
And I have one more, slightly less ambitious wish for the future.
That people stop generating a random picture on a free ChatGPT plan and calling it design. Or, worse, marketing. ;)
Sources and further reading
Verified data and integration references (October 2026):Stanford AI Index 2026 — Economy; McKinsey — The State of AI 2025; McKinsey — State of AI 2026; U.S. Census Bureau — AI Use by U.S. Businesses; Ahrefs — API Guide.
US Census Bureau — AI in business · NBER — research on AI and productivity · OpenAI — product updates · Midjourney documentation · Windsor.ai · Google Ads Help. The final published version will replace broad research links with the exact studies supporting each quantified claim.
