8:42 AM - Inbox triage
I open Threshold at 8:42 on a Tuesday morning. Coffee in hand. The dashboard loads and I see three new signups overnight. Two are from the cold email campaign I ran last week - enterprise consultants in different verticals - and one is a referral from Carol Reyes at Reyes Family Practice, who converted to a paid plan three weeks ago.
Carol's note says: "Threshold saved me hours already. Just sent a friend who has the same problem." That lands different than a sale notification. It makes the whole thing feel real.
I switch to Gmail. Fourteen unread emails. Most are automated: Slack alerts from Threshold when new questionnaires complete, a Stripe revenue report from yesterday (seven hundred and thirty-two dollars - three new subscriptions and a renewal), a calendar reminder for today's 2 PM call with Rajesh Kumar, and three customer replies that need actual thought.
The first one is from Marcus Webb, a marketing consultant who signed up five days ago. He's asking whether Threshold can skip the "project scope" section for certain clients. He doesn't want every prospect going through the full intake if they're just asking for a quick audit. Fair question. I flag it mentally and write back: "Absolutely. In your next setup call, I'll show you the conditional logic - you can set it up so first-time prospects get a shorter path if they select 'quick audit' on the first screen. Takes five minutes to configure."
Simple. I've done this twice now. Each customer thinks they're the first to ask, and each time it takes nothing to solve.
I check the Slack notification log from overnight. The system flagged one questionnaire as incomplete - someone started but dropped at question eight of twenty-two. The name is David Lee, a recruiter who signed up yesterday from a LinkedIn outreach. I make a note to reach out to him around noon, see if he had a technical issue or just got busy.
10:15 AM - A flagged conflict
I open the Threshold admin panel to review outputs from new intakes. Three questionnaires completed last night. Two are clean - both prospects are well-qualified, clear scope, realistic budgets. The third one is flagged by the system as "Confidence: Medium."
It's from Sarah Chen, a nonprofit executive director who hired Threshold to gather intake info for a proposal on grant writing support. The AI asked her about her nonprofit's annual budget in the second question. She answered: "Around three million."
Then later, at question seventeen, when the AI asked about her scope, she said: "One volunteer handles grants part-time. We need help building a system so it scales."
The flag is correct. The system sees the mismatch: a three-million-dollar organization with one volunteer on grants suggests either she's undervaluing the scope, or she doesn't realize what's actually involved. The AI can't resolve that. Neither could I, not without talking to her.
I pull up the full transcript. Reading through the questions and her answers, I can see where Sarah got thoughtful and where she got vague. She's clearly smart but tired. The nonprofit space does that. I send her a message via email: "Hi Sarah, I reviewed your intake brief and want to clarify one thing before we schedule a call. You mentioned the grant writing is currently one part-time role. If we're building a scalable system, that's probably a 120-hour project minimum plus training. Does that timeline feel realistic for your team?"
I'm asking directly because the AI output is too ambiguous to act on. This is the real work. The AI asks good questions. I decide what questions matter.
12:30 PM - Lunch and the metrics check
I grab lunch at my desk and open the Stripe dashboard. Three new subscriptions landed since yesterday. Two are the cold email signups at $99 a month. One is an upgrade from a customer who was on the $49 starter plan and moved to $149 for the "priority support" tier. That's Marcus Webb. He bought more than he signed up for.
Week-to-date revenue is sitting at four thousand eighty dollars. That's on pace for about seventeen thousand by month end. Not a bad Tuesday, but I'm a long way from the mid-range ARR benchmark of a hundred eight thousand. I'm in month two. This is fine. It's early.
I check the pipeline in Slack. The bot posts a daily summary every morning at 7 AM: how many prospects are in each stage of the intake-to-call funnel. Twelve prospects have completed intake but haven't scheduled a call yet. Another five completed intake yesterday and got on my calendar already. The conversion from "completed intake" to "booked call" is running at about 60 percent, which I think is good but I don't have a baseline yet.
I message my co-founder: "Three new subs today. One upgrade to $149. Sarah Chen intake flagged for scope mismatch - I'm asking her a clarification question." He replies with a thumbs-up emoji.
2:08 PM - The conversation that matters
I get on a call with Rajesh Kumar, a management consultant who signed up eleven days ago and completed intake last week. He's been quiet since, so I want to know what he's thinking.
"The brief was detailed enough that I didn't even need to send a follow-up email," he says. "I got on a call with that prospect, and I'd already answered all the questions I normally ask. It felt like they'd already been through the intake, not like this was the first conversation."
That's exactly what I'd hoped to hear, but hearing it from a real customer still hits. He's not trying to sell me on anything. He's just saying it worked.
We spend twenty minutes talking about how he's using it: one intake per new prospect, one outcome brief to read before the call, then he does the call. He's not thinking about the AI anymore. He's just using it. That's when you know the product is working.
He says: "I'm going to move up to the $149 plan next month. The Slack notifications are saving me so much context switching."
Two upgrades in one day. That wasn't planned.
3:45 PM - A customer churned
I get an email from Angela Torres, a freelance UX researcher who was on the plan for twenty-eight days. She's canceling. Her note: "This is great, but I realized I don't actually get many referrals, so the intake questionnaire doesn't save me time like I thought it would."
Fair enough. She was a bad fit. The cold email line I lead with - "eliminate 3-5 back-and-forth emails per new client" - assumes you're getting new clients regularly enough to care. Angela gets maybe two or three projects a year from word-of-mouth. Threshold wasn't going to solve a problem that doesn't exist at her volume.
I don't try to convince her to stay. I reply: "I totally get it. If you get busier with referrals, we're here." No hard feelings. This is the work too: knowing when someone shouldn't be a customer.
4:55 PM - Pipeline review and one more approval
I check on David Lee, the recruiter who dropped out of intake yesterday. He never replied to the email I sent at lunch. I send a Slack message to my notification bot and search for his record. The system shows he's still in the system, just inactive. I'll follow up tomorrow. Sometimes people just get busy.
I review the two clean intake briefs from this morning one more time. Both are going into calls tomorrow. One prospect is a small law firm looking for intake automation for client onboarding. The other is an HR consultant building a practice around executive coaching. Both are real opportunities. Both have signal that they're seriously evaluating solutions, not just kicking tires.
I approve both briefs and they generate the auto-reminder emails to the prospects with the call links. Calendly integration fires and the meetings are confirmed.
6:15 PM - Closing the laptop
I close the Threshold tab and sit for a minute. It's 6:15 PM on a Tuesday.
Today I approved three questionnaire outputs, fixed one scope ambiguity before it became a problem call, sent four direct customer emails, reviewed the metrics, lost one customer who wasn't a fit, upgraded two customers, and took one call that made me believe the product is actually solving the right problem.
The AI is doing maybe seventy percent of the work. It's writing the intake flows, asking the right questions in the right order, flagging the uncertain cases. But it's not running the operation. I am. I'm deciding which flags matter, which customer questions need a real answer, which people should get called and when.
This is not passive income. This is not a robot that hired itself.
But it's also not the nightmare of back-and-forth emails I remember from before. I'm not reading twenty separate message chains trying to reconstruct what each prospect actually needs. The AI did that work. I just review the output.
That feels like the right trade.
I open Slack. I've got sixteen new notifications. I delete most of them. Three are worth a reply for tomorrow.
I shut the laptop.