If people praised your AI tool and then quietly stopped opening it, this article answers the question you are actually asking: was the product wrong, or was the place you put it wrong?
I built something good once and watched it die anyway.
It was a prompt tool. A clean little web app where the entrepreneurs in my world could store their prompts, run them, and stop losing good work in a graveyard of Google Docs. I was proud of it. A few hundred people signed up the first week. The messages were glowing. “This is exactly what I needed.” “Finally.” I remember sitting in my office reading those messages and feeling like I had finally built the thing.
Six weeks later I could count the weekly active users on my hands. Not my fingers. My hands.
That is the part nobody posts about. The silence after the applause. I did what every founder does: I blamed onboarding, I blamed the emails, I blamed myself for not marketing harder. Then a woman in my community said the sentence that rearranged my thinking permanently. She said, “Jonathan, I love it. I just forget it exists.”
She was not lying about loving it. She was telling me the truth about her day.
Here is the direct answer to why nobody uses the AI tool you built: your product is probably fine, and your placement is probably fatal. You asked a human being to add a new habit, a new tab, and a new login to a day that is already full. They were not rejecting your product. They were obeying their own habits. Habits win. They always win. The AI products winning right now are not destinations you drive to. They are layers that show up inside the surfaces people already have open, and if your offer requires a new tab, you are fighting a habit instead of borrowing one.
That reframe changed my business. Let me show you the whole thing, including the ugly middle.
Key Takeaways
- Most AI tools do not fail because the product is bad; they fail because they ask the user to change where they work.
- Habits are slow to build and fast to defend, which means a new tab is a real cost your customer pays every single day.
- The dominant AI launch pattern in 2026 is embedding into existing surfaces like Slack, Word, Instagram, and the design canvas rather than launching a standalone destination.
- Distribution inside an existing workflow beats a superior product sitting behind a login the user forgets.
- You do not need to own a platform to build a layer; you need to find where your customer’s work already happens and become the intelligence inside it.
The Problem
Here is the trap, and I fell all the way into it.
The old playbook was clean and it felt like ambition. Build the destination. Own the customer. Own the relationship. Market people to your thing, hard enough and long enough, until they form the habit of coming to your thing. Every piece of startup advice I absorbed for years pointed the same direction: build a place, then fill it.
That playbook has a hidden assumption baked into it, and the assumption is a lie. It assumes there is an empty slot in your customer’s day shaped exactly like your product, and all you have to do is fill it.
There is no empty slot. There has not been an empty slot in years.
Your customer wakes up and opens the same four things. Their inbox. Their calendar. Their messaging app. The one document or dashboard where their real work lives. That is the whole territory. Every one of those surfaces was fought over and won years ago, and your customer’s hands know the shortcuts without asking their brain for permission. When you ask them to open a fifth thing, you are not asking for a click. You are asking them to interrupt a groove that took months to wear.
I did not understand that when I built my prompt tool. I thought I was competing with other prompt tools. I was not. I was competing with the browser tab they already had open, and the browser tab was undefeated.
What made it worse is that the feedback lies to you. People genuinely liked my tool. They told me so, sincerely, and then they went back to their day and never thought about it again. Enthusiasm is not adoption. Enthusiasm is cheap and it costs the user nothing to give you. Adoption costs them a rewired morning.
I want to be honest with you about the difficulty here, because I have watched a lot of good entrepreneurs get hurt by this. This is not a problem you can market your way out of. You can buy attention. You cannot buy a habit. You can win the launch and still lose the year.
But here is the reframe, and it is the most freeing idea I have found in a decade of building. If the habit is the immovable object, stop trying to move it. Go stand inside it.
The Evidence
I am not asking you to take my word for this. The data on this is brutal and it is public.
People abandon new software almost immediately. Business of Apps reported 2026 benchmarks showing average day-one retention around 21.1% on Android and 23.9% on iOS, falling to roughly 2.1% on Android and 3.7% on iOS by day 30. Read that again. Somewhere north of 90% of the people who wanted your thing enough to install it are gone within a month. That is not a marketing failure. That is the base rate of asking humans to change where they work.
Habits take real time to form, and your product does not get that time. Phillippa Lally and her colleagues at University College London published the definitive study on this in the European Journal of Social Psychology in 2010. Ninety-six volunteers, one new daily behavior each, tracked for 12 weeks. The median time to reach 95% of maximum automaticity was 66 days, with a range from 18 days all the way out to 254. Sixty-six days is the median for a behavior the person actually chose and wanted. Your app has about a week of goodwill before it is forgotten. The math does not work.
Switching surfaces has a measurable cognitive price. Gloria Mark’s research at UC Irvine, published in the CHI proceedings on interrupted work, produced the number everyone quotes: roughly 23 minutes and 15 seconds to return to the original task after an interruption. Her more recent data is even more sobering, showing that average time on a single screen before switching has collapsed from about 2.5 minutes in 2004 to around 47 seconds. Every new tab you require is a tax, and your customer’s brain is already overdrawn.
Distribution beats product. It is not close. Peter Thiel put it plainly in Zero to One: a company with an average product and great distribution routinely beats a company with a great product and no distribution. Slack is the case study nobody wanted to be. Slack won the affection. Teams won the default. Microsoft did not out-build Slack; Microsoft put Teams inside the bundle that was already installed on the corporate desktop. By mid-2025 Teams was reporting roughly 360 million monthly active users while Slack’s growth had flattened, and in April 2026 Salesforce and Slack took Microsoft to London’s High Court over exactly that bundling. You can be right and lose.
And look at what launched this month. The pattern is not subtle anymore. On Product Hunt right now, Framer AI Agents put agents directly onto the design canvas where designers already live. Yasmine Works bills itself as an AI coworker that lives in your Slack. Meta launched Muse Image on July 7 with no standalone app at all, dropping it straight into the Meta AI app, Instagram Stories, and WhatsApp. OpenAI’s GPT-5.6, launched July 9, became the preferred model inside Microsoft 365 Copilot across Word, Excel, PowerPoint, and Chat. Anthropic made Claude selectable inside Microsoft’s own Copilot Chat and pushed Cowork onto web, mobile, and into Microsoft 365 with real write access to email, calendar, and files.
Nobody built a destination. Everybody built a layer.
One honest caveat, because I will not oversell this to you. Microsoft reported more than 450 million commercial paid Microsoft 365 seats in its FY26 Q2 results, and paid Copilot seats have been reported in the neighborhood of 20 million. Even perfect distribution does not hand you automatic adoption. Being where the work happens buys you the shot. It does not take the shot for you.
The Solution
Here is what changed for me.
I stopped asking, “Where should my product live?” and started asking, “Where is this person already sitting when they have this problem?”
That is the entire shift. It sounds small. It is not. It reorders everything downstream of it.
When I applied that question to my dead prompt tool, the answer was humiliating and obvious. My people were not sitting in a prompt library when they needed a prompt. They were sitting inside their AI chat window at 6:40 in the morning with a blank box and a deadline. That was the moment. That was the surface. My beautiful app was one tab away from the moment, and one tab away is the same as a hundred miles away.
So I stopped building the destination. I put the Perfect Prompt Framework where the blank box was. Templates that lived inside the tool they already had open. Prompts that traveled with them instead of waiting for them. No new login. No new habit. No “remember to come back.”
Usage did not double. Usage became a different category of thing, because I was no longer asking anyone to remember me. The work reminded them. That is the difference between a destination and a layer: a destination requires memory, a layer requires nothing.
Now here is the part I need you to hear, because I think most people misread this idea and then make a worse mistake.
Becoming a layer is not the same as becoming small. Layers are not lesser. The intelligence inside the workflow is frequently worth more than the container around it, because the container is a commodity and the judgment is not. Meta did not diminish Muse Image by putting it inside WhatsApp. They aimed it at the exact second a human being wants an image, which is the second they are already typing to someone. That is not humility. That is a heat-seeking missile.
And this is not just a software company problem. This is your problem if you sell coaching, consulting, services, courses, anything. Does your offer require your client to log into your portal, open your platform, watch your dashboard, remember your thing? Every one of those is a new tab wearing a costume. The best consultants I know do not summon clients to their world. They show up inside the client’s Monday, in the client’s Slack, in the client’s document, in the client’s actual mess. Same expertise. Radically different placement. Wildly different results.
I fought this for a while, if I am honest. Owning the destination feels like status. Owning the layer feels like renting. But I would rather be indispensable in a room my customer is already standing in than be the landlord of a beautiful room nobody enters.
Practical Steps
Here is how I would run this if I were sitting next to you today.
Find the moment, not the market. Write one sentence: “My customer has this problem at [specific time] while doing [specific task] inside [specific tool].” If you cannot fill in those three blanks with real specificity, you do not have a placement strategy yet, you have a guess. Go ask five customers to narrate their morning to you.
Name the four surfaces. Ask your best customers what apps are open right now. Not what they should use. What is open. You will get the same four answers over and over: email, calendar, a chat tool, a doc or a CRM. That short list is your real distribution map, and it was drawn before you got here.
Audit your offer for hidden tabs. Go through your delivery, every step, and mark anywhere the customer must go somewhere new. Portals, dashboards, logins, “just click this link.” Count them. Every one is a tax you charge your customer daily.
Kill or relocate the top two. Take the two heaviest tabs and either delete them or move that function into a surface your customer already has open. Email works. A shared doc works. Their Slack works. This is unglamorous and it will feel like a downgrade. Do it anyway.
Test the layer before you build the layer. You do not need an integration to test placement. Deliver it manually inside their surface for two weeks. Paste it into their doc yourself. Send it in their channel. If usage climbs when you carry it to them, placement was your problem, and now you know what to automate.
Measure return, not applause. Stop tracking signups and compliments. Track whether the same human comes back in week four without you reminding them. That is the only number that has ever told me the truth, and it is the number my prompt tool was screaming at me while I was reading nice messages.
Pick your platform risk with open eyes. Building on someone else’s surface means they can change the rules. Real risk. Manage it by owning the relationship and the judgment, not the container, and by living on more than one surface. The alternative is owning a room with nobody in it, and that is not lower risk, it is just quieter failure.
Frequently Asked Questions
Does this mean I should never build my own app or platform?
No. It means earn the destination, do not assume it. Destinations work when the customer’s work genuinely has no home yet, or after you are already indispensable somewhere else. Start as a layer inside their existing workflow, prove people come back on their own, then graduate to owning the surface if the pull is real.
How do I embed into a tool I do not own or have no API for?
Start manually. Deliver your value inside their email, their shared doc, their Slack channel, by hand. Placement is a behavior question before it is an engineering question. If manual delivery inside their surface lifts usage, you have validated the layer and earned the right to build the integration.
Isn’t building on someone else’s platform dangerous if they cut me off?
Yes, and it is still usually the better bet. Reduce the risk by owning the customer relationship and your unique judgment rather than the plumbing, and by existing on two or three surfaces instead of one. A platform can remove your feature. It cannot remove the trust you built with the human using it.
I sell coaching and services, not software. Does any of this apply?
Completely. Every portal, dashboard, and login you make clients open is a new tab in disguise. The consultants who retain clients longest deliver inside the client’s existing workflow instead of summoning them to a separate world. Same expertise, different placement, dramatically better adoption.
If distribution beats product, does my product quality even matter?
It matters enormously; it just cannot rescue bad placement. Distribution gets you into the room where the work happens. Quality is what keeps you there once you arrive. Microsoft’s own Copilot numbers show that being everywhere does not automatically make people use you. Placement buys the shot, quality makes it.
The Close
I still think about that message. “I love it. I just forget it exists.”
For a while it stung, because I heard it as a verdict on my work. It was not. It was the kindest, most useful sentence anyone has said to me about building things. She was not telling me my product was bad. She was telling me I had built it in the wrong place, and she was too polite to say it that way.
So let me say the thing I wish someone had said to me while I sat there refreshing a dashboard that would not move.
You are not failing. You are misplaced. There is a real difference, and the difference is everything, because misplaced is fixable this week and failing is a story you tell yourself for a year. The thing you built might be excellent. The judgment you have is probably rare. You just parked it one tab away from the moment it matters, and one tab is a canyon.
Go stand in the moment. Find the exact second your person needs you, and be sitting there when they arrive. Do not make them drive to you. Do not make them remember you. Do not ask a busy human to build a 66-day habit around your login screen when you could be the intelligence inside the thing already open on their screen.
The winners this year are not the ones who built the prettiest destination. They are the ones who stopped asking for attention and started showing up where attention already was.
If this landed, come find me. Follow me, connect with me, and tell me what you are building and where you are trying to put it, because I want to hear it and I have been wrong enough times to be useful. Bring your dead app. I have one too. We can talk about what it was trying to teach us.
Stop building places people have to go. Start being the thing that is already there.
About the author
Jonathan Mast is the founder of White Beard Strategies, where he helps more than 100,000 entrepreneurs use AI to do real work instead of impressive demos. He is the creator of the Perfect Prompt Framework, a speaker, and a stubborn believer that the best AI strategy is the one your team actually uses on a Tuesday. He learned that lesson the expensive way, by building a prompt tool everyone praised and nobody opened, which is why he now measures adoption in week-four returns instead of launch-day applause.
