Founders are often told that timing is everything.
The advice sounds obvious, but it is not particularly useful when you are inside the company, making decisions with limited information. You rarely know whether you are early, wrong, or simply not patient enough.
When we started AI Palette in 2018, AI was still a relatively new proposition for most consumer goods companies. We believed it could fundamentally change how companies understood consumers and developed products. But convincing enterprises to trust and adopt that technology was a very different challenge from building it.
The technology could reveal valuable signals, but customers were not necessarily organised to act on them.
We were building both a product and the confidence required to use it.
Eight years later, having built, scaled and eventually sold the company, I have a different appreciation of what timing really means.
And I increasingly believe that being early is one of the most misunderstood advantages in building a startup.
Being early creates two jobs
If you enter a market early, you do not only have to build the solution. You also have to help the customer understand the problem in a new way.
That education can become an advantage, but it consumes time, capital and founder attention.
At AI Palette, we were trying to change how consumer goods companies understood consumers and identified opportunities for growth.
Traditionally, much of this work relied on market research, surveys, historical sales data and the experience of insights teams. We believed AI could help companies identify emerging consumer needs much earlier, by analysing large volumes of continuously changing data.
But believing that a technology can deliver value and convincing an enterprise to change how it operates are two very different things.
We were not simply asking customers to buy another software product.
We were asking them to trust a different way of generating insights and, eventually, making decisions.
And that was a much bigger challenge.
A customer could be impressed by what the technology demonstrated and still not be ready to adopt it.
There is a significant distance between finding something interesting, believing it is useful, allocating a budget for it, and incorporating it into everyday decision-making.
The mistake is assuming that technical readiness and commercial readiness arrive together.
They rarely do.
You are not just competing against other companies. You are competing against existing processes, established ways of thinking and, quite often, the comfort of doing nothing differently.
The three clocks of startup timing
I now think about timing as three separate clocks.
The technology clock.
Is the technology sufficiently mature to solve the problem reliably, at a cost that makes commercial sense?
The market clock.
Does the customer understand the problem, believe a better solution is possible, and have enough urgency to pay for it?
The survival clock.
Can the startup survive long enough for the first two clocks to align?
Founders tend to obsess over the first clock and underestimate the other two.
Technology is often the part we feel we can control. We can hire better engineers, improve models, add features and build a more sophisticated product.
But we cannot force an industry to adopt a new way of working at the speed we would like.
And the third clock keeps ticking regardless.
Salaries need to be paid. Products need to be developed. Investors expect progress. Customers need to be acquired.
The market might eventually become ready, but the startup has to remain alive until then.
Good timing is therefore not merely predicting the future correctly.
It is designing a company that can keep learning while the future arrives.
The best early-stage strategy is often the one that buys you another cycle of evidence without losing sight of the larger thesis.
The hidden cost of creating a market
Looking back at AI Palette, I think one of the biggest challenges was not the technology itself. It was the amount of market education required.
When a category is established, customers already understand what they are buying.
A CRM company does not need to explain why businesses should manage customer relationships. A cybersecurity company does not need to establish that cyber threats exist.
The conversation can focus on why one solution is better than another.
But when you are introducing a new category or a fundamentally different way of solving a problem, you often have to establish the value of the category before you can sell your own product.
That changes the economics of building a company.
Consider how consumer goods companies traditionally purchased market intelligence.
They were familiar with subscribing to databases such as Mintel or Euromonitor. They knew what they were buying, how to evaluate the information, and which budgets would pay for it.
But an AI platform that could identify emerging consumer trends and help generate new product ideas was something different.
Before customers could evaluate our product, they first had to understand why they needed a different approach. Then they had to justify a new investment, often without an established budget or evaluation framework.
We weren't simply competing with existing research providers. We were asking customers to rethink what consumer intelligence could look like.
And that made the sales process fundamentally different.
Product development also becomes more complicated because customer feedback reflects not only what the product lacks, but also what the market is not yet ready to adopt.
Are customers not buying because the product is inadequate? Because the problem is not important enough? Or because they have not yet developed the confidence to change?
These are very different problems, but from the outside they can look remarkably similar.
This is where founders can easily misread the signals.
If you believe strongly enough in the future you are building towards, every rejection can look like evidence that the customer simply hasn't understood it yet.
Sometimes that is true.
Sometimes the market is telling you something important about your product.
Knowing the difference is one of the hardest parts of building an early-stage company.
But there is another side to being early.
The time spent educating a market can also become an advantage. You develop a deeper understanding of customer problems, build relationships and establish trust before competitors recognise the opportunity.
If those relationships eventually translate into adoption, the years spent developing the market can become difficult for later entrants to replicate.
The challenge is that you rarely know in advance whether the investment in market education will pay off.
When the market suddenly caught up
When ChatGPT arrived in late 2022, something interesting happened.
AI suddenly became something people could experience for themselves.
For years, enterprise AI companies had been explaining what machine learning and artificial intelligence could potentially do. Much of that conversation involved demonstrations, presentations and promises about future capabilities.
Generative AI made the possibilities tangible to a much broader audience.
People no longer needed to imagine what interacting with AI might feel like. They could try it.
That was a major shift in the market clock.
But awareness is not the same as adoption.
An executive using ChatGPT personally does not automatically mean their organisation is ready to deploy AI across business-critical workflows.
Enterprises still have to address data security, reliability, integration, accountability and organisational change.
In other words, the market clock accelerated, but it did not suddenly synchronise with the technology clock.
And another important thing changed.
The cost of building AI software began to fall dramatically.
Capabilities that previously required substantial engineering teams, specialised machine learning expertise and significant infrastructure can increasingly be accessed through APIs and foundation models.
A small team can now build a convincing AI application in weeks, sometimes days.
This is an extraordinary development for entrepreneurs.
But it creates a new problem.
When technology becomes easier to access, technology itself becomes a weaker source of differentiation.
In 2018, building was hard. Today, distribution is hard.
I think this is one of the most important changes in the AI startup landscape.
In 2018, being able to build sophisticated AI capabilities was itself a meaningful advantage.
There were fewer experienced AI engineers, less accessible infrastructure and fewer off-the-shelf tools.
Building the technology required significant investment.
Today, many of those barriers have fallen.
That does not mean building reliable enterprise software is easy. It certainly isn't.
But creating an impressive demonstration is no longer particularly difficult.
And when hundreds of startups can build similar products using the same underlying models, the question becomes less about who can build the technology and more about who can get it adopted.
Distribution and customer access matter more than ever.
But distribution in enterprise AI is not simply having access to potential customers.
It is knowing who has the problem, who has the authority to solve it, where the budget sits, and what needs to happen for a new technology to become part of an organisation's everyday workflow.
A founder with deep industry relationships and a clear understanding of how customers make purchasing decisions may now have a greater advantage than a technically stronger team starting without that access.
This is a significant shift from the AI landscape we entered in 2018.
I also think this changes how founders should think about defensibility.
A technical head start can disappear surprisingly quickly when foundation models improve.
But deep customer relationships, proprietary workflows, accumulated domain knowledge and strong distribution are harder to replicate.
In some ways, the advantage is shifting from the company that knows how to build AI to the company that knows where and how to apply it.
And perhaps more importantly, how to get customers to use it.
What I would do differently today
When we started AI Palette, we believed AI would fundamentally change how consumer goods companies understood markets and developed products.
I believe that even more strongly today.
But if I were starting a company now, I would approach the opportunity differently.
I would spend less time asking whether the technology could solve the problem and more time understanding whether the customer was ready to change.
Is the problem painful enough?
Who owns it?
Who controls the budget?
What happens if the customer does nothing?
And perhaps most importantly, do I have a credible way to reach these customers repeatedly and economically?
I would also be more cautious about building technology simply because it has become possible to build it.
The speed at which AI applications can now be developed creates a temptation to start with the solution and look for the problem afterwards.
But a working product is not evidence of a viable business.
And a rapidly growing market does not guarantee that every company entering it will succeed.
In fact, when the barriers to entry fall, the importance of choosing the right problem and having the right distribution strategy only increases.
This does not mean technology no longer matters.
It means technology needs to be part of a larger advantage, rather than the entire advantage.
Being right is not enough
One of the privileges of building AI Palette for eight years was seeing how much the world changed during that period.
Ideas that once required lengthy explanations have become part of everyday business conversations.
AI has moved from being an emerging technology that enterprises were cautiously exploring to something almost every leadership team is trying to understand and adopt.
But the journey also changed how I think about entrepreneurship.
Being early gives you time to learn, build relationships and develop capabilities before others recognise the opportunity.
But it also means spending resources while the market catches up.
Being late means entering a more crowded market, but one where customers may already understand the problem and have budgets to solve it.
Neither is automatically better.
The question is whether the advantages of entering early outweigh the cost of waiting, and whether you have a way to capture value when the market finally becomes ready.
Eight years ago, I thought one of the biggest advantages a startup could have was seeing an important opportunity before everyone else.
I still believe there is value in being early.
But I have come to appreciate that identifying where the world is heading is only part of the challenge.
The harder part is figuring out how to build a business while the world gets there.
And if I were starting another AI company today, that is the question I would spend much more time answering before writing the first line of code.
Loading comments…