I got cancelled in the most unexpected way. It was the worst feeling I can remember in a long time.
But before I tell you that story . . . let's talk about AI.
(And if you're using AI in your search, you'll want to pay attention.)
AI is undoubtedly the biggest tech revolution we may see in our lifetime (though social media and cell phones all came to be in my lifetime too).
I love technology. I love how it can create efficiency and allows mere mortals to be able to do more than ever imagined on their own.
Even as I drafted this piece, I had AI building me some well overdue company brand guidelines. The ability to multi-task and produce more output is unprecedented . . . but is the output better?
The way you brief it and give it context, and more importantly, hold it accountable to a high standard and the correct output, is what makes it work well.
I spent several decades working in marketing and branding. So I know what a solid brand guidelines document should include and look like.
By using AI, I'm trying to get what's in my head out into the world, so my team and the folks that work for us can use our brand guidelines, rather than get corrected for brand rules they didn't know existed.
A founder probably shouldn't be building a slide deck from scratch, so having a tool to, as my team says, put the bike in gear, is a great time saver!
However, I'd been working with AI for a few days on the brand guidelines and it still wasn't quite getting it. I had to go in and fix a lot of things, but at least it got things started for me.
Had I not had the years of experience I did with brands (and many times, building those damn guidelines from scratch), I think it would have been pretty good!
However, it was missing some of the magic (yes I said it) that good brand guidelines should convey in order to give clarity and ensure consistency. And it's my experience that tells me it's not quite right and needs adjustments before rolling out.
Without that experience, it's hard to know if the AI output on most things is right, or just telling you what you want to hear.
I recently watched a hilariously worrying account of a real estate entrepreneur sharing that ChatGPT almost blew up a $50 million dollar real estate deal. Both the buyer and the seller went to GPT to ask if the price was right, and serendipitously, it told the seller it was undervalued AND it told the buyer it was overpriced!
ON the same deal! (You can watch it here)
Dangerously, we are outsourcing expertise to a technology that is still in its infancy, and hasn't developed the discernment, critical thinking, and expertise of someone who's been on the front lines and has real experience.
Here at AcquiMatch we are bullish on AI. We are always building multiple projects simultaneously, constantly replacing or fixing processes with AI, and scaling things at a level I never thought possible.
But it works because before we built a single thing in AI, my team had processed hundreds of thousands of deals, had hundreds upon hundreds of broker and owner calls, and worked with dozens of first-time buyers getting them under LOI and closing. We have the front-line experience to know what the output should look like.
I do not sign off a project until it's at least as good, if not better, than what me and my team could do.
It's a tool to take things off someone's plate that they don't need to do, so they can focus more on the things only humans can do.
We work with AI experts who can do the development needed to give our AI the best guardrails and context. These guys are experts who know HOW to work with AI to get the right outputs.
We've been able to move entire team members to better seats as a result of our work, we've been able to get matching results on hundreds of data points on a buyer that a human wouldn't have been able to remember or retain, and we've added customization and speed we couldn't have achieved without AI either.
BUT we know what the output should look like, and we kept training and training until it was as good, and now arguably better, than what my team did. (In fact, we went from throwing out 66% of CIMs to 25% because the front-end matching is SO good!)
Worryingly, I'm seeing a lot of folks talk about using AI for their business acquisition search. New "influencers" are running webinars to teach you how to use AI to run your search, who are not AI experts or deal sourcing experts . . . new tools are popping up all the time, and plenty of searchers are "vibe coding" in an effort to outsource the slog that is searching.
Vibe coding is a software development approach where you use natural language to prompt AI assistants to generate, debug, and iterate on applications, focusing on the overall "vibe" or outcome rather than writing syntax line-by-line.
This AI revolution in search, at least in the short term, will inevitably make your search worse, not better.
Let me explain.
I need to go back to that time I was cancelled to bring this all home for you.
It was my first semester as a lecturer at University of Texas McCombs School of Business. I was teaching how to turn an idea into a real business for undergraduate students taking the Entrepreneurship Minor.
I decided to have the class centered around actually starting a business. The student groups funded their startup with $20 each (the money that would have normally gone to books) and had to choose an idea they could initiate and sell within the semester.
No Total Addressable Market (TAM) fake business pitches here . . . They were going to be graded on really going for starting a business and learning from it.
Within each group of four students, they each chose which department they'd be in charge of: operations, marketing, sales, or finance.
Towards the end of the semester, one of the assignments was to write an individual assessment from the vantage point of their department. They were to talk about how it was going, assess themselves and their co-founders, and do a SWOT analysis for their business idea in the current landscape with real issues they faced to scaling.
Things like the fact they were all students and couldn't be full-time in the business, for example, should have shown up in the operational assessments, because that would impact their delivery. Or for finance, that they were unpaid labor and would have to address pricing and potential profitability if they decided to move forward with building the business and hiring people.
In the first weeks of the semester, ChatGPT had just started gaining momentum. No stranger to tech (having been co-founder of a food tech media business), I made it clear in the syllabus and in expectations that they could use AI tools for things like research and ideas, but the final output on assignments needed to be their own writing.
As I was reading the papers, I noticed that my student from Spain had all of a sudden attained perfect grammar and vocabulary she hadn't had in previous papers.
I alerted my TA, who was in her masters on machine learning and AI, that I thought this student had used AI.
She exclaimed, OH MY GOD, THAT'S IT! I THOUGHT A BUNCH OF THEM HAD BEEN CHEATING!
Turns out about 10 students had papers that all sounded incredibly similar to each other, including near identical SWOT analyses, despite this being an individual paper.
Not to mention, they were bad analyses. They were generic and didn't take into account anything individual or unique to the team or their department.
I was upset. I felt like I had been lied to. And I had to decide how to move forward.
Though I was a first-time lecturer, I was a serial entrepreneur with the scars to show it. Here was a moment to show my character, and to teach the value of owning your shit to my students. Something that every entrepreneur needs to know.
I told the students we were aware a number of them had used AI on their paper, and reminded them it wasn't allowed for papers, as was outlined in the syllabus.
They had two choices.
1. Own it, come forward and tell us, and get the opportunity to redo the paper for a maximum grade of C. (I also gave them extra credit options to identify how to use AI in both business and the classroom in a productive way so they could make up the points lost.)
2. Don't come forward, and risk us identifying them and putting them forward as a cheater to the university. Risk getting kicked out at worst; at best they'd fail my class.
Sure enough, all of the students we'd identified came forward. (Including an additional very panicked, honest student who'd used Grammarly, which we were OK with and wasn't part of the offending group.)
Some of the offenders had just not read the syllabus or paid attention to that part of our opening lectures and didn't realize it wasn't ok.
Fun Fact: The last student to come forward (after a reminder from my TA) said both his laptop and phone had fallen off a balcony and that's why he hadn't messaged. The Dog Ate My Homework trick is alive and well.
The lessons I was hoping to impress on the offending students were:
1. Understand the contracts you agree to, read them fully (i.e. the syllabus).
2. Know that these tools do not work well if you don't yet have the experience. Get your real reps in so you can know whether the output is on track, as well as learning the right ways to use these tools.
3. Owning when you mess up isn't the end of the world, as long as you're honest and try to make it better.
Even if we hadn't caught all of this, their grades on those papers would have been bad anyway.
The analyses were weak and not circumstantial to them, their business, or the reality of the local market. No matter the department, all of the analysis looked exactly the same for the group. And more worryingly, they weren't correct or helpful if they wanted to move forward with the business.
The entire course was set up to learn by doing. Forcing yourself to analyze your own business through a lens that was foreign to you is HARD, but it starts developing a skillset you won't have access to without going through the process of learning (and in entrepreneurship, it's usually developed by learning the hard way).
You cannot, and should not, outsource critical thinking to AI in business (or in your search!)
Take the 2-minute readiness assessment and find out.
Most of the students were gracious about the whole situation.
A few were pissed off.
So pissed off, in fact, that they took to a teacher rating platform and created 30+ accounts to leave me 1 star reviews talking about how bad of a lecturer I was (I had less than 30 students, for the record).
It was, hands down, one of the most horrific moments of my life, reading review after review talking about me being a fake, knowing nothing about business, etc. They threw every punch they could think of and they didn't hold back.
Rationally I knew most of it was untrue, and some angry students venting. But there was some stuff in there I really couldn't avoid listening to. Mostly about how I gave feedback.
To work so hard to give them an educational experience I would have killed for myself, and be repaid with being slated on the internet, sucked.
But ultimately I stuck to my values and felt good about that. The university even used my way of handling it to show other educators, who were also getting a footing for how to deal with a new technology in the classroom that could deny students the opportunity to learn real discernment and wisdom.
And . . . I learned something too.
Their negative reviews, along with the official end of semester feedback forms, showed me I needed to learn more constructive ways of giving feedback.
And so, the next semester I started using AI to help me take my real feedback and position it in a way that landed. I used my real feedback as the input, told AI to use the WISE feedback framework, and then reviewed and learned the results so that I could write that way without AI in the future.
As a result I learned how to get through better to my students. My feedback was important, but how I packaged it made the difference of whether it was heard or not.
Pay attention to how I used AI to learn. The input into the AI was my feedback and original thought, the output was its suggestions to soften the blow, and the outcome was me learning through this repetition.
Despite the AI situation, in that semester five of the eight groups actually made sales and a few were even profitable! In my second semester, all of the groups not only made sales, but were all profitable! They EARNED money to learn.
They learned how to quickly validate ideas, and at the end of it, with all the knowledge they had, they got to make go/no go decisions. Without raising a single penny of investment or making a fake pitch.
They now have a repeatable framework and skillset they can use in the future when they have business ideas to quickly validate and get to market.
Despite the cancellation, it was still worth it to me.
Back to searching and AI . . .
That lack of real knowledge to interpret results was, and still is, my concern with using AI.
Especially on something as high stakes as doing a search for you, on something you're going to put millions of dollars on the line for. Something that could bankrupt you and take your home. Something that, if you close down, could cost your employees their security too.
If you don't yet have the eye for what the output should be, or know how to properly brief these things, be wary. Yes, they might give the illusion of saving time, but will they find a good match? How will you know?
The way you learn is by getting reps. You can go faster and be more efficient in how you get those reps, but make sure you don't outsource the high stakes thinking to something if you don't know how to interpret the results!
Cancel me all you want, but there's too much on the line here in business-buying land.
(And no, I didn't use AI to write this one. Every word is mine.)
This is for education only and isn't financial, legal, or tax advice. Talk to your own lender and advisors about your specific deal.