There is a question underneath every conversation about artificial intelligence: what happens to the people who are trying to build something with it?
Not a demo. Not a clever image. A real thing: a website, an app, a store, a research archive, a media library, a newsletter, a customer experience, and the systems that keep all of it moving.
That is the kind of question Philly Tours makes practical. The platform is still unfinished—the iOS and Android apps have more work ahead—but it already shows what a founder can attempt when one person can work with AI across software, writing, research, media, marketing, and operations.
The future is not that AI removes the need for people. The nearer future is that one person can finally work across more of the distance between an idea and a working institution.
The next year: AI becomes a working partner
The next twelve months will make AI feel less like a chatbot and more like a junior team member.
It will inspect a repository, explain an unfamiliar file, draft a change, run a test, find a broken link, revise a page, prepare a campaign, and help turn a rough instruction into a sequence of work. It will move between code, images, audio, spreadsheets, research notes, and customer-facing copy with less friction than today.
That does not mean every output will be correct. It means the cost of trying will keep falling. Stanford’s 2025 AI Index reported that the cost of using a model at roughly GPT-3.5 capability fell more than 280-fold between November 2022 and October 2024. Cheaper capability changes who gets to experiment.
For a small platform, the immediate opportunity is not to pretend to be a giant technology company. It is to remove the bottlenecks that keep a good idea trapped in one person’s head:
- A founder can describe a feature and get a first implementation.
- A researcher can turn notes into a structured archive and a public story.
- A media maker can create several formats from one strong piece of source material.
- A small business can test messages and landing pages without waiting weeks for a separate department.
The human work becomes more deliberate: decide what matters, check what is true, protect people’s information, and make sure the result still sounds like the organization that is putting its name on it.
The next five years: a small team can operate like a larger one
Five years from now, the most important AI systems for small businesses may not be visible to customers as “AI.” They will be the quiet machinery behind the experience.
A founder might have a research agent that gathers sources and keeps citations attached. A production agent could turn a tour outline into an audio script, transcript, translation, caption file, social cut, and accessibility review. A software agent could maintain the app, monitor errors, prepare releases, and explain what changed. A marketing agent could propose campaigns, but a human would still decide which promises are honest enough to make in public.
This is where the difference between automation and authorship matters. Automation can multiply a decision. It cannot decide whether the decision deserves to be multiplied.
For Philly Tours, a useful five-year future would include a research companion that helps people follow a family line across counties and generations without inventing relationships. It could connect a neighborhood, a church, a migration route, a newspaper, a land record, and a family memory—but it would need to show its sources and mark uncertainty. In genealogy, a confident error is not a small bug. It can overwrite a family’s understanding of itself.
The labor market will feel this transition unevenly. The International Monetary Fund has warned that generative AI will affect a large share of jobs, with some tasks automated and others made more productive. The people best positioned for the transition will not only be the people who know how to prompt a model. They will be the people who understand a field deeply enough to recognize a bad answer.
The next ten years: the platform becomes an operating layer
Ten years is too far away for confident product predictions, but one direction is visible: AI will become an operating layer inside many organizations.
Instead of opening five separate tools, a founder may describe the objective: preserve this archive, launch this route, explain this history, reach these customers, or improve this service. The system will propose a plan, create the working materials, execute approved steps, and report what happened.
That future could widen access to institution-building. A local historian, a neighborhood organization, a family archivist, or a small tour company may be able to produce work that once required a staff of engineers, designers, editors, producers, and marketers.
But abundance will create a new problem: too much plausible material and not enough trustworthy judgment. When every organization can publish, the value of provenance will rise. People will want to know where a claim came from, who reviewed it, whether a photograph was altered, how a family record was interpreted, and what remains unknown.
The organizations that last will build trust into the product itself. They will preserve source links, keep an audit trail, separate memory from documentation, ask permission before using personal material, and leave room for a human to say, “We do not know yet.”
The NIST AI Risk Management Framework and its generative-AI profile point toward this kind of discipline: identify risks, measure them, document decisions, and keep accountability attached to the people and institutions using the system.
What AI cannot supply by itself
AI can help build the container. It cannot supply the reason the container deserves to exist.
It cannot inherit a grandmother’s warning about a county courthouse. It cannot know why a changed surname still feels like an injury generations later. It cannot replace the relationship between a guide and a visitor, or the trust between a family and the person handling its records. It can help organize those realities, but it should not flatten them into content.
That is why the future of AI is also a future of boundaries. The more powerful the system becomes, the more important it is to know what it should not decide:
- It should not invent a family connection because the names look similar.
- It should not turn a painful history into a frictionless marketing slogan.
- It should not treat private memories as free raw material.
- It should not publish a historical claim merely because the sentence sounds finished.
- It should not confuse speed with understanding.
The founder’s advantage
The founder’s advantage in the AI era will not be having access to a machine that everyone else lacks. Access is spreading too quickly for that to remain a durable advantage.
The advantage will be knowing what to ask, what to protect, what to refuse, and what kind of world the work is trying to make.
A platform like Philly Tours can use AI to extend a small team’s reach: more routes, more formats, better accessibility, faster iteration, stronger research tools, and a deeper connection between public history and lived experience. But the platform’s meaning still comes from the people who choose the stories, name the harm, preserve the evidence, and invite others to walk with care.
In one year, AI will help more people make things. In five years, it will help small teams operate at a larger scale. In ten years, it may become part of the basic infrastructure through which organizations think and work.
The question is not whether humans will still matter. The question is whether we will use the extra capacity to produce more noise—or to do more careful, generous, accountable work.
That choice is still ours.