An AI agent working for a creator business could spot a sudden drop in revenue at two in the morning, work out which platform is responsible and recommend a solution before anyone has woken up. It could also misunderstand the problem, make a perfectly confident recommendation and send the business in completely the wrong direction. When you’re dealing with thousands of customers, multiple platforms and transactions happening around the clock, both possibilities have very real financial consequences. Creators’ fans, subscribers and viewers are spread across the world, but the creator is still one person who cannot be online all the time. AI is already helping creators and their teams stay connected with audiences around the clock. It doesn’t replace the creator but it can have significant impact on their income.
This is why I believe the creator economy is becoming one of the most interesting testing grounds for enterprise AI. These businesses are experimenting with autonomous agents, developing their own software, managing digital identities and finding new ways to connect systems that were never designed to work together. They’re confronting many of the same challenges facing banks, retailers and major technology companies, often with smaller teams, tighter margins and considerably less patience for lengthy technology projects. The results could tell us a great deal about where enterprise AI is heading, particularly as the technology moves towards making decisions and carrying out work independently.
I’ve spent more than twenty years building technology businesses, including online communities long before anybody started talking about the creator economy. One of the biggest changes I’ve witnessed is the sophistication of the businesses supporting successful creators. Some manage substantial international operations, with employees, specialist teams, complex customer relationships and an increasing dependence on technology. Their operational challenges would be familiar to anybody running a large company, although the speed at which things change can be extraordinary. A decision made in the morning might affect thousands of customer interactions by lunchtime, and an unexpected platform change can disrupt processes that have taken months to establish.
AI offers an opportunity to make these operations considerably more efficient. Much of the work involves collecting information from different sources, interpreting what is happening and deciding what to do next. Ai is already handling much of this work today, from tagging and organising content to preparing materials for creators, assisting with written conversations on platforms that permit it, and helping teams establish performance metrics and monitor their results. An agent might identify unusual patterns in customer behaviour, investigate a sudden change in revenue or highlight an opportunity that would otherwise have taken somebody hours to discover. Giving that same agent permission to make changes introduces an entirely different level of complexity, particularly when the information available to it is incomplete. A decline in sales could be caused by changing customer preferences, a technical issue or a platform introducing new rules. Understanding the wider commercial context remains essential, and an incorrect interpretation can quickly become expensive.
The industry’s growing use of AI-generated personalities and digital identities adds another fascinating dimension. We’re approaching a point where a creator could license their voice or likeness to support customer interactions in different languages, create new forms of content or extend their commercial presence without personally producing everything. The technology is advancing quickly, while the questions surrounding ownership, consent and commercial control are proving rather more difficult. Who has the authority to create and operate a digital version of somebody? How should the revenue be shared, and what happens when the individual withdraws permission? I know this from experience. In one of our new projects, we’re developing something close to an AI avatar of a real creator, with that creator directly involved and in control of how it is developed and used. These questions are therefore very practical for us, rather than simply theoretical. The same questions will confront larger organisations as they experiment with synthetic brand representatives, AI customer service and digital versions of real people.
Another development I’ve been following closely is the growing appetite for creator businesses to take control of their own technology. For years, companies largely adapted their operations around whatever software happened to be available. If a product handled most of what they needed, they tolerated the parts that didn’t work particularly well. Increasingly, teams are using open APIs to create their own dashboards, build proprietary management systems and connect applications around their particular requirements. I’ve seen how valuable this can be when a business has developed its own way of working and needs technology that accommodates it. The question of ownership also becomes important because an operation built entirely around somebody else’s closed ecosystem can be extremely vulnerable when that provider changes its rules or commercial arrangements.
AI coding agents could accelerate this shift enormously. I’ve been working on making these tools more accessible to people without traditional engineering backgrounds, and the opportunities are particularly interesting for entrepreneurs who understand their businesses intimately but have never written software. For someone who knows exactly where an inefficient process sits, the ability to build a solution themselves could change how they approach technology altogether.
Somebody who knows exactly where an inefficient process sits can increasingly experiment with developing a solution themselves. Previously, that might have required a development team, a substantial budget and weeks of discussions before anything could be tested. There are limitations, of course. Producing functioning code is relatively straightforward compared with ensuring that software remains secure, reliable and maintainable. Nevertheless, the ability to turn operational knowledge into working technology is becoming much more widely available, and that has significant implications for the established software industry.
The latest research suggests this change is already affecting purchasing decisions. McKinsey’s August 2026 State of AI survey found that 32 per cent of respondents said their organisations had decided against purchasing at least one software product or feature because they could build the functionality internally using AI coding tools. That’s a striking development for an industry accustomed to selling businesses increasingly elaborate software subscriptions. Building more software internally won’t eliminate the need for reliable underlying technology. Businesses will still need secure data storage, robust transaction processing, permissions and systems that integrate properly. They may simply become less dependent on individual vendors to determine how those capabilities are packaged and used. This is where I want to take our technology next. My goal is to build a platform where creators and their teams can ‘vibe-code’ whatever tools they need, rather than depending on software that doesn’t fit the way they work and frequently promises more than it delivers. I suspect the same appetite for greater control over technology exists in almost every other industry.
For larger enterprises, these developments arrive at an interesting moment. The same McKinsey research found that 40 per cent of respondents at companies generating more than $1 billion in annual revenue were scaling AI agents, up from 27 per cent the previous year. However, only 6 per cent of respondents qualified as AI high performers, reporting both significant value and an EBIT contribution of at least 5 per cent from AI. There is clearly considerable work to do in translating technological progress into meaningful financial results. I wonder whether the experience of smaller digital businesses might encourage enterprises to examine the processes they are trying to improve more closely. There is a tendency in established organisations to introduce new technology around existing structures, even when the structures themselves could benefit from changing.
The economics are something creator businesses have little choice but to confront. When an operation processes large volumes of customer activity, even relatively small technology costs can become significant. The commercial case for using an expensive AI model to complete a routine task may disappear if conventional software can achieve the same result more reliably. McKinsey found that around one in five respondents had already constrained their use of AI because of operating costs, including the cost of processing tokens. As the technology becomes more capable, I expect businesses to become more selective about where they deploy it. Sophisticated models will have an important role in interpreting complex information and dealing with unfamiliar situations, while simpler systems should continue handling tasks that require consistency and predictable outcomes.
I’m particularly interested in how the relationship between people and autonomous software will develop. Creator businesses frequently operate across time zones, with teams managing customers and commercial activity while senior decision-makers are asleep. AI agents could become valuable colleagues in this environment, monitoring operations and helping people respond more quickly, but the permissions they receive will need careful thought. There is a considerable difference between allowing software to identify an issue, recommend a change and implement that change without approval. The more valuable the information and the greater the commercial consequences, the more important it becomes to understand what the agent is doing and retain the ability to intervene. Companies will also need to think carefully about what happens when different agents interact, particularly if they are working across systems controlled by separate providers.
Good human judgment will remain valuable throughout this evolution. Experienced operators often recognise changes in customer behaviour or commercial conditions before those changes are obvious in the data. They understand relationships, context and the consequences of decisions that might appear perfectly sensible to somebody looking at a dashboard. I expect the most successful applications of AI to draw on that knowledge, allowing smaller teams to operate more effectively while giving experienced people better information and more time to focus on important decisions. The opportunities are substantial, although businesses will need to develop new skills in evaluating software, interpreting its recommendations and recognising when something doesn’t look right.
At the same time, people are increasingly willing to follow AI recommendations. For certain tasks, the technology can process more information, more consistently, than a person could reasonably manage alone. Even so, I wouldn’t rely on a single model. I believe there’s more value in orchestrating several different large language models that can challenge and check one another. A single model can sound completely convinced of its own answer, even when that answer is wrong. Using several models doesn’t eliminate that risk, but it offers another opportunity to question a recommendation before acting on it.
There are obvious differences between a creator business and a multinational bank or manufacturer. Larger enterprises have regulatory obligations, legacy infrastructure and commercial risks that demand much more extensive safeguards, and some of the experiments taking place in the creator economy would be entirely inappropriate in those environments. Still, the underlying issues surrounding interoperability, trust, commercial returns and control over technology are becoming increasingly similar. Creator businesses are addressing them in an environment where the consequences of a decision can become apparent almost immediately, and where the people responsible for the business generally have a very direct interest in whether the technology works.
The creator economy has spent years being underestimated as a technology market. Given the changes now taking place across its operations, I wouldn’t be surprised if some of the ideas that eventually reshape enterprise AI emerge from businesses that most corporate technology executives have barely considered.
