We Already Tried AI and It Did Not Work: How to Handle the Objection
- This is the highest-quality objection you will hear. They spent money on this category already, which means budget exists and the pain is real.
- Do not defend AI or contrast yourself with what failed. Ask three questions about what broke. Diagnosis beats reassurance every time.
- Almost every failed AI project fails for one of three reasons: no owner, no verification, or no real problem. Name which one it was.
- Then de-risk the next step down to something so small that failing again would cost them almost nothing. A paid pilot with one measurable outcome, not a full engagement.
What This Objection Actually Means
When someone says they already tried AI and it did not work, they are telling you three things at once. They had a problem real enough to spend money on. They believed a promise. And the promise did not survive contact with their business. The third part is the one shaping the conversation you are now in.
That makes this a completely different objection from price or timing. Price means they are not convinced of the value. Timing usually means no. This one means they were convinced once, acted on it, and got hurt. The belief was there. What is missing now is trust, and trust is not rebuilt by making a better promise.
The Three Questions to Ask First
Ask before you say anything about your own work. Every sentence you spend explaining why you are different, before you know what broke, sounds exactly like the pitch they already bought once.
- What exactly did you try, and what was it supposed to do for you? You are looking for the job it was hired for, in their words.
- Who inside your team owned it day to day? This one question identifies the most common cause of failure by itself.
- What was the moment you stopped trusting it? Not when it was switched off, but the moment somebody decided it could not be relied on.
The third question is the one that produces the useful answer. People remember the specific incident: the invoice that went to the wrong client, the summary that invented a meeting, the automation that ran twice. That incident is the actual objection. Everything else is a summary of it.
The Three Reasons These Projects Fail
Once you know what happened, name the cause. Not as an accusation, as a diagnosis. Nearly every failed AI project I have been asked to look at fell into one of three buckets, and clients find it a relief to hear their situation described by someone who has seen it before.
| The cause | What it looked like from inside | What has to change |
|---|---|---|
| Nobody owned it | It launched, everyone was busy, and within a month it was nobody's job to check it | One named person owns it, with a recurring check that takes minutes and cannot be skipped |
| Nothing verified the output | It produced confident results, one of them was wrong in public, and trust never recovered | The output gets checked against something objective before anyone relies on it |
| It solved a problem they did not have | It worked exactly as demonstrated and nobody's week got easier | Start from a task someone actually does every week, not from what the technology can do |
Why the previous attempt failed, and what it tells you about the next one
Notice that none of the three is a technology problem, which is why buying a better tool did not fix it and why your version has to be sold differently. If you pitch a better model, you are answering a question they did not ask.
What Not to Say
- That was not real AI. This is condescending and it says their judgment was the problem.
- The technology has moved on a lot since then. Possibly true, and it is exactly what the last vendor said.
- What tool did they use? Asked too early, this sounds like you are looking for someone to blame instead of something to understand.
- We do things differently. Every vendor says this. It carries no information and costs you credibility.
- Any defence of AI in general. You are not there to represent a category. You are there to solve one problem for one business.
De-Risk the Next Step Until It Is Almost Free to Say Yes
After a failure, the size of your proposal is the objection. A full engagement asks them to make the same size bet that already went badly, and no amount of reassurance makes that comfortable. So make the next step small enough that failing again would barely register.
The shape that works is a short paid pilot on one task, with one measurable outcome agreed in advance, and a defined stopping point. Paid matters, because free reintroduces exactly the dynamic that failed before: a project nobody owns because nobody paid for it.
- One task, chosen by them, that somebody on their team does every week.
- One number that decides whether it worked, agreed before you start and measured the same way afterwards.
- A fixed end date, with a clear decision at the end rather than an automatic rollover.
- A named owner on their side who checks the output, because the last one failed without one.
The sentence that closes this is straightforward. You have already paid once for something that did not work, so I do not want you to take a big swing on my word either. Pick the one task that annoys you most, we will do that and only that, and you will know in three weeks whether it is worth continuing.
Why This Is the Best Lead You Will Get This Month
A burned client has already done the two hardest parts of the sale for you. They proved there is a budget by spending it, and they proved the problem is real by trying to fix it. What is left is proving that you are not the last vendor, which is entirely within your control.
They are also, in my experience, better clients afterwards. Someone who has watched an AI project fail asks sharper questions, insists on verification, and assigns a real owner, because they have felt what happens without those things. That is a client who makes your work more likely to succeed rather than less.
What to Do With This
Next time you hear it, resist the reflex to reassure. Ask what they tried, who owned it, and when they stopped trusting it. Name which of the three causes it was. Then propose something so small that saying yes is close to costless. That sequence has turned more dead conversations into paid pilots for me than any script.
Inside the Claude Code Profit Room, members bring the actual objection they got, in the actual words the client used, and we work out which of the three failures is underneath it and what the smallest paid next step looks like. If you have a conversation stalled on this right now, bring it in and we will take it apart.
Short, practical drops on offers, outreach, pricing, and closing clients with Claude Code. No spam, unsubscribe anytime.
Frequently asked
How do I respond in the moment when a client says they already tried AI?
Agree with them first, then ask what they tried and what it was supposed to do. Agreeing costs you nothing and immediately separates you from the vendor who sold them the last thing. Defending AI as a category is the one response guaranteed to fail.
Should I ask which tool or vendor they used before?
Not early. It sounds like you are hunting for someone to blame. Ask what it was supposed to do and when they stopped trusting it. If the specific tool matters, it will come up on its own once they are talking freely.
Is a free pilot a good way to overcome this objection?
No. Free recreates the exact condition that caused the original failure, which is a project nobody owns because nobody paid for it. Make the pilot small and paid, so it gets a real owner and a real decision at the end.
What if the previous project failed because of something the client did?
Say it plainly and without blame, usually as no owner or no verification. Clients respect a diagnosis that includes them far more than a pitch that pretends the failure was purely technical, and it sets up the conditions your own project needs.
How small should the first project be after a failed attempt?
Small enough that failing again would be an inconvenience rather than a loss. One task, one number that decides success, a fixed end date, and a named owner. If they have to think hard about the risk, it is still too big.
Last reviewed August 14, 2026.
Co-founder of the Claude Code Profit Room. Went from shipping software to closing paying clients, and now teaches builders the selling half of the equation.
More from David Iya →
