Updated Customer Interview Tech Stack
Nerdy post incoming 🤓 (note: I feel like AI’s ruined emojis. They use them constantly and - I have no idea if this is just me or everyone - If something smells like it was written by AI I immediately disregard and delete it. I’m almost insulted by it. I bet this changes eventually? Maybe?).
Anyway, I’ve been doing a bunch of customer interview stuff lately - we’re finishing up cohort two of this workshop, I did a pod episode on customer interviews using AI, and I’ve been running customer interviews for an idea I’m messing around with on the side (coffee related, reach out if interested/an expert/own a cafe).
The short update is - this is the best time to run customer interviews… ever.
New AI tools (AI giveth, AI taketh away) have made the whole thing way more efficient, and the emergence of AI has given people another excuse to not run interviews, which is HUGE for the people who do. It’s a way to create separation. Whoever knows their customer better wins.
So, here’s the updated stack / flow that I’d recommend to run magical customer interviews right now. There are three parts of interviews - outreach, the actual interview, and synthesis. Here are tools for each:
Outreach:
A little trick right off the bat. Absolutely don’t use AI for the part of customer outreach you really want to use it for.
Don’t use it to “customize” the first line of your message so that you can send 10,000 “personalized” emails. Don’t use it to help you write the body of the email to “personalize” it further.
Humans are exceptionally good at sniffing stuff out, and it is wildly easy to sniff out an AI message that’s been “customized.” Mostly because we get a TON of practice. Here’s a screenshot of my inbox (I’d imagine yours is similar):
An insufferable barrage of spray and pray AI slop all day every day. A freight train as relentless as the 1927 Yankees.
Life is about contrast, and this is your chance. Whatever the AI slop du jour is, as you can see above it’s just saying the person’s first name a bunch, do the opposite.
Instead of spray and pray to 10,000 people, pick the 20 you REALLY want to speak to and send them an actual personalized email. Be specific and human. Tons more on how to do that in the workshop, if you feel like learning it (running another one starting Oct 27). The main idea is to make it obvious why, out of all the people in the world, you’re choosing to speak to them.
The secret to like 85% of life is that everyone just wants to be chosen. Make it clear that you did.
And, if you want to use AI to find people, group them, build a CRM, get notes on their background so that you can write your personalized email - great. Just remember that you aren’t slipping AI writing by anyone, and no one wants to read it.
2. The Interview itself: Granola.ai
No brainer. It runs in the background so you don’t have to “invite” it to your Zoom meetings. There’s also an iPhone app for in-person meetings - put it on the table and record. And, if you’re doing an old fashioned phone call - do the call from your laptop or use whatsapp to make the call and Granola will pick it up. It’s also free for what you need it for.
Granola will create bullet-points from the call, but those are usually not all that useful. What we want is the full transcript. Which we use with…
3. The Interview Synthesis: Claude
Here’s where AI really starts to sing. Create a new project in Claude (my favorite AI tool for these, though they’re all pretty much the same).
Tell it what you’re up to, something like “I’m going to dump a bunch of transcripts for customer interviews I’ve run trying to find an idea for a startup. I want to use this as a master database - so, keep track of the interviews and who they were with, and I’ll query this database for questions about specific types of customers, problems they have, solutions they’ve tried, etc.”
Then, copy paste in your transcripts with information about the person you interviewed.
Now you have a living, breathing, database of actual customer knowledge.
The other way to use Claude for synthesis is directly after you run an interview, particularly if it was in-person.
Flip over to the “audio” version of the AI - click the little speaker - and you can have an actual conversation with the tool. Prompt it to ask you questions about the interview you just ran - tell it to be curious about specific problems and stories and processes. Then, have it add this context to the database.
This is a great way to squeeze all the juice out of an interview after you’ve finished.
4. The Prompts
You can obviously ask your new database questions like “which problem does it seem like customers were most willing to pay to have solved,” or “what commonalities do you see from the interviews,” but you can also ask for a start on generating something like an OPLP (one person landing page).
“If I was going to create an OPLP for each of these customers, with an H1 that was in the format of “You’re struggling with X specific problem and it’s keeping you from Y,” and an H2 in the format “We’ll help you Z (wild success point) without Q (biggest risk)” - what might that look like?”
Or, “based on these interviews, what newsletters might it be good for me to sponsor to try and acquire customers?”
And, as you go, you can dump in more information on your customer.
Pretty cool.
Try it out and let me know how it goes. Customer interviews are (somehow) even more useful now.