Stop Guessing How Your Customers Are Doing: 3 AI Workflows for Service Businesses

👉 Unlock our Zapier Agent templates + the prompts and the handholding

the danger zone: Customer health doesn’t break because you lack dashboards.
It breaks because you can’t hold cross-system context at scale.

When you had three clients, checking customer health was as simple as texting them and calling it white glove service.

Now you have 50 clients. Maybe 100. And that intuition you relied on? Gone with the wind.

Because the signal splinters. Your project management boards show one thing. Slack threads show another. Call transcripts are buried… somewhere.

Losing clients? Not very chill. Losing them to passive churn because you didn't notice they went radio silent three weeks ago? Suppperrr not very chill.

the solution: Alex Lieberman + Arman Herzarkhani are Managing Partners at Tenex. Earlier this year, it became exceptionally hard for them to know how their customers were doing due to rapid scaling (odd flex).

So they built an AI-powered system to make sure five-alarm fires never happen under their noses. It took them a lunch break to wire their data to Claude and Zapier.

Steal their 3-level customer health system. Each level builds on the last:

The first level is conversational. You ask questions, the AI answers—no automation, no scheduling. Just you and Claude, with your project management data wired directly into the conversation.

You get a per-client summary showing issues completed, activity level, current projects, and themes. It flags potential problems—like clients who haven't reviewed work in weeks.

Instead of scanning every account, you now know exactly who deserves attention this week. As Alex puts it: "At this point, we can reach out to our technical strategist and say, 'Hey, why is so-and-so client's work stalled?' And then we uncover the bottleneck there early."

How to build it:

Once you know the questions that matter, automate them.

For Tenex, that's a Monday-morning briefing with the top three client risks, the top three opportunities to "wow," and health scores across every account.

Think about the President of the United States. Every morning, they receive a concise brief with everything they need to know—so decisions can be made early, without having to ask.

This agent works the same way. It runs on a schedule and arrives automatically each week, tailored to your business and your actual clients.

How to build it:

Now the system takes action, without overreaching,

Picture an agency with struggling accounts:

Once your weekly brief runs, add a second Zapier Agent to generate ready-to-edit email drafts—one for each account that needs attention. But it does not send the emails (that’s way too messy for AI to handle).

How to build it:

If you don’t know these tools, you can expect to build this whole system in about 1–2 hours. If you do know these tools, call it 15 minutes. The payoff: fewer surprises and a faster, calmer response when something starts to drift.

The system works regardless of where your data lives or how it’s structured. Linear or Jira. Slack or Teams. Software delivery or consulting hours. The pattern stays the same.

➿ Snag our templates and prompts here.

Zuckerberg capped his 2025 AI spending spree by acquiring Manus, a Singapore-based AI agent startup with Chinese roots, in a deal valuing the company at just over $2 billion.

Meta described Manus as “one of the leading autonomous general-purpose agents that can independently execute complex tasks like market research, coding, and data analysis,” and moved quickly to buy out all existing investors and fold the team—roughly 100 people—into its AI org.

Manus is pretty good at automating things, outperforming ChatGPT Agent, Grok 4, and Sonnet 4.5 at finishing multi-step work without falling apart halfway through. But each of these frontier models is operating at the floor of what is widely considered AGI.

in English: Right now, general-purpose agents still can't replace workers who deliver work you'd actually invoice for.

OpenAI is hiring a Head of Preparedness (aka Chief Anxiety Officer).

It’s basically someone to think about how AI breaks things after it ships.

Sam’s concern is that its models are finding real security exploits, influencing mental health, and inching toward self-improving behavior.

translation: Its models are outrunning the guardrails, and OpenAI is willing to pay north of $500K for someone to stay professionally tense 24/7.

Someone joked:

why it matters: If OpenAI needs a full-time executive to stress-test models after launch, companies adopting AI should take the hint: AI becomes a risk surface the moment you ask it to replace judgment. For businesses, that means keeping real humans inside decision loops.

This maybe obvious to some, but don’t outsource your thinking + don’t let a model make your decisions for you. Use your head more than you use AI. In business and in life: touch grass, respectfully.

AI can't improve what it can't see.

That’s one piece of the puzzle for companies trying to adopt AI.

That insight helped Scribe raise $75 million and hit a $1.3 billion valuation.

apply it:

📧 Reply with your favorite AI use case. If it’s awesome, we’ll invite you to guest-star on our weekly show.

🎉 We’re back to full-sized newsletters in the New Year! 🎉