Our current period of AI excitement is fueled by FOMO. New models drop weekly, techniques spread through the industry like wildfire, and no one wants to be left behind. Everyone is willing to put in the extra work to learn, configure, and onboard with the new hotness just to stay current. Users are showing a level of patience and openness rarely afforded to products.
This is a temporary subsidy. You didn’t earn it, and you can’t count on it.
Products that win the next phase won’t be the ones with the most sophisticated models or the deepest integrations, but the ones that can prove value before patience runs out.
Time to First Value Hits Different for AI
TTFV, or Time to First Value, is a standard SaaS metric that tracks how long it takes customers to reach their first meaningful outcome. It’s the “aha” moment when a product clicks. In traditional software products, it’s the earliest recognizable value the product can offer, like the first order submitted, or first workflow automated.
The best AI products, however, don’t simply provide a discrete outcome, but open up new workflows and ways of thinking. We should not think of TTFV for AI products as when the first task gets completed, but rather when the user understands the new possibilities unlocked by this product.
Here’s the tension: adopting a new mental model doesn’t come cheap. AI products can’t function without context, and providing it is expensive. These tools need the right data, examples, and configuration, because bad data can be just as bad as no data. Every file upload, every setting, every onboarding step is an investment we ask from our users before we’ve proven that this investment will pay off.
Right now, FOMO covers that gap. Users push through because they’re afraid of missing out, but soon that FOMO will fade. Products with high context tax and slow TTFV are being pushed out by competitors who figured out how to deliver the paradigm shift faster.
MCP vs. Claude Skills: A Study in TTFV
Anthropic gave us a perfect example of this scenario when they first launched MCP, Model Context Protocol, in Nov 2024. The initial reaction was…measured. People understood the value proposition of standardizing how LLMs communicated with external tools and data sources. But adoption required setup and experimentation before the payoff would come. The response from many was essentially: interesting, I can’t wait to see what people come up with.
Claude Skills landed differently.
People immediately understood agent skills because Anthropic made the brilliant decision to launch with pre-built skills to handle PDFs and Word docs. These file types are the most mundane and ubiquitous things in computing, ancient structures that even cutting edge research companies fight with daily.
You didn’t need to imagine a use case or configure anything. You gave Claude a .docx file to edit and immediately grasped the paradigm shift. TTFV was seconds because you saw the feature in action with no setup.
Same company, same capabilities, radically different time to first value. The difference wasn’t the technology, but how fast users could see themselves in the product.
What I Learned at Iris
I saw this dynamic play out firsthand at my last company, an AI startup building RFP automation that promised significant efficiency and revenue gains in proposal writing. The nature of this product comes with a very high context tax. Customers had to upload data, tag their data, create a new project, then wait for generated answers. Only after seeing actual output did they understand the value. For some prospects, this could take days.
So we started chipping away at the context tax. We had our sales and CX team upload documents for prospects before we handed them their trial login. We created predefined writing styles, and offered guided prompt engineering sessions with customers who needed further refinement. And as a result, our prospects did less, converted more, and converted faster.
These techniques were simple but extremely effective. By taking the setup work off their plates, customers could focus purely on the opportunities our product generated. We operationalized many of the initial setup steps with the CX team, but the real product lesson was clear: sophisticated infrastructure doesn’t matter if users bounce before experiencing what it can do.
We also asked ourselves the next logical question: could we get a faster TTFV by delivering a different value? Well-written responses were our key offering, but were time consuming and required full onboarding of our product. What if we made a key moment out of synthesizing a writing style prompt, or automatically tagged KB documents?
The Clock Is Ticking
The FOMO subsidy is a temporary market condition. As AI tools embed deeper into infrastructure, user expectations will normalize and people will be less forgiving of painful onboarding durations.
If your product relies on user patience that only exists because of hype, you’re building on borrowed time.
This isn’t just a UX problem. It’s an architecture problem. It means designing your product around the question: what’s the smallest slice of value that makes users believe?
Figure that out, and work backward from there.