98% CSAT, 6 Out of 10 at the QBR: How GadellNet Rebuilt Its MSP Service Delivery

The QBR deck looked great. CSAT at 100% for the quarter. SLAs met. Resolution times down. Then, at the end of the meeting, the client got one more question: on a scale of one to ten, how would you rate the partnership?

Seven. Sometimes six.

Kurt Beard and Max Simon kept hitting that wall at GadellNet, a St. Louis MSP with roughly 180 people and about 400 clients across 35 states. Their CSAT averaged 98%. Their clients were telling them something else. Long-time partners said GadellNet no longer felt like the small IT company they hired. They felt like a small fish in a big pond.

At AISU (AI Service Unleashed) 2026, Kurt and Max shared what they found and how they fixed it. The short version: it wasn't a people problem. It's what happens to MSP service delivery when the queue becomes the product.

Your survey grades a fifth of the work. Your coaching sees even less.

GadellNet's CSAT surveys come back about 20% of the time on a good week. Some engineers hit 30%. Some clients sit closer to 10%. That's actually better than most IT service desks, where benchmarks put survey response under 10%.

Now flip it around. For some clients paying GadellNet good money, 80 to 90% of the work done for them never got evaluated. Nobody read it. Nobody found the moments worth praising or the ones worth fixing.

Coaching ran on even thinner data. The tickets that reliably got a second look were the 1 to 2% with a negative score. A lead would call within the hour and say, "Let's talk about this ticket." That was the coaching program. Kurt calls it coaching to failures. Engineers who worked onsite or behind the scenes sometimes went a full month without a single review.

GadellNet isn't an outlier. ScalePad's 2026 MSP Trends Report found that best-in-class CSAT doesn't reliably predict low churn, partly because the people who renew contracts aren't the people answering ticket surveys. A 98% score and a 6 out of 10 partnership can both be true at the same time.

The queue trained the habit. Fix the system, not the person.

As GadellNet grew, the work changed shape. Engineers used to know Sally in accounting at the construction company down the street. Now they faced an endless queue of nameless, faceless tickets. Leadership measured them on tickets closed per week and time per ticket.

The team delivered exactly what it was measured on. Speed went up. Connection went down.

GadellNet's leadership team had been reading Will Guidara's Unreasonable Hospitality, which draws a line between service and hospitality. Service is doing the work right. Hospitality is how the person on the other end feels about it. Guidara's point is that hospitality can be taught. That means it can be built into a system.

The evidence pointed straight at the system. When GadellNet started reviewing every interaction, one of its most tenured engineers landed near the bottom on client experience. Ten years of green checks on every KPI. Near-perfect CSAT. But this engineer rarely called clients, ran 200-message chats when a three-minute call would do, and closed tickets without confirming the fix worked.

Max doesn't frame that as the engineer's failure. "We were more or less a part of building this engineer's mindset," he said. For a decade, every review said speed was the job.

That's the ticket at its worst: a unit of work built for the system, not for the person who needs help. 

Great service needs a rubric, and every score needs an example

GadellNet's fix started by writing down what great looks like. Their rubric scores every client interaction from 1 to 10 across categories like these:

• Resolution factors. Not just time to resolution, but why. A "can't print" ticket that took four days because it was really an infrastructure issue tells a different story than one that sat untouched.
• Status updates. When a ticket goes to awaiting, does the client hear "I'm waiting on Lenovo, expect three to five days"?
• Response and follow-up. Are messages left unreturned overnight? Does "I'll call you tomorrow" turn into a call tomorrow?
• Clarity and tone. Can the client understand it? GadellNet once supported a seminary where staff held doctorates in Old Testament studies. Nobody there needed a lecture on Meraki.
• Ownership and documentation. When an engineer writes "docs need updating," do the docs get updated?

Kurt's rules for building one are simple. Make the categories specific. Mix hard and soft skills, because warmth doesn't rescue a fix that never lands. As he put it, White Castle could have great hospitality, and the sliders would still be terrible. Anchor every score in a real example, because a number without an example means nothing. Then revisit the rubric every quarter so it tracks what the team is working on now.

Thread's Super Magic does the reading. A lead hands it a list of tickets pulled from the PSA, the rubric, and the output they want. Back comes a scorecard with notes tied to specific moments: the password sent in plain text, the client who shared something personal and got pushed straight into troubleshooting. Instead of hunting through the queue for random tickets, the lead starts the week with plain-language notes on every interaction.

Coaching or surveillance is a choice you make in week one

AI that reads every ticket can make your team better, or it can send them job hunting. GadellNet was clear about which one it was building.

The old model is familiar. A CSAT leaderboard on the office TV. The top name gets a reward. The bottom name gets told to do better. "That's not coaching," Kurt said. It's handing out a grade and hoping.

GadellNet built the opposite:

1. Say it out loud. The team knows 100% of client interactions are reviewed.
2. Share the rubric and the prompt. Engineers can check their own work a few days later and see themselves improving.
3. Let the team shape it. The prompt started at about 30 lines. After rounds of team feedback, it's about 500.
4. Ask for coaching language, not a verdict. The AI output starts a conversation. It never gets pasted into a "shame on you" email.
5. Coach toward growth. When an engineer struggled with Mac tickets, the answer was Mac troubleshooting documentation, not a write-up.

The research backs this up. Cornell researchers found that people monitored by AI felt less autonomy and pushed back harder than people monitored by humans. When the same analysis was framed as developmental feedback, that penalty disappeared, including the intent to quit.

That tenured engineer is the proof. They pushed back at first, and fairly: ten years of reviews had called them a star. So their lead walked through real tickets with one question. How could we have handled this differently? Over time, the engineer started catching it on their own. Call the user who always picks up. Loop in the assistant for the executive who never answers. The goal was never a better score. It was a better instinct.

What you see when you can see every ticket

It didn't happen overnight. Max is clear that a system like this takes weeks, even months, to catch on. Once it did, GadellNet could see what surveys never showed them. Who thanks clients and means it. Who asks the personal question. Who explains a network problem to a seminary professor in words that land. Who confirms the fix before closing the ticket.

They also found the great work nobody surveyed. Clients write long comments when things go wrong and rarely when things go right. Now GadellNet turns its best interactions into micro-learnings for the whole team, whether or not a survey ever came back.

The QBR changed too. Account managers now walk in with two or three moments GadellNet got wrong, like a P1 worked for two hours before anyone told the client. Naming the miss before the client does builds trust fast. In Max's words, you gain trust in drops and lose it in buckets. Since the change, [QBR partnership scores moved from X to Y].

Want to try it? Here are the starter prompts Kurt and Max shared:

• Review [engineer]'s tickets from last week. Score each for clarity, tone, and follow-up. Give me one thing to praise and one thing to coach on.
• Look across the team's tickets. What patterns show up in how we handle frustrated clients?
• Find tickets where client sentiment started negative but we still got a positive CSAT. What turned it around?
• Find tickets where client sentiment looks great but no CSAT came back.

Max's last warning is the one to tape to your monitor: don't let this become a three-week shiny object. Bring it to every one-on-one. Build the habit. Hospitality is a decision, and the best MSPs make it on every ticket.

Run GadellNet's first prompt on your own desk. Book a demo, bring last week's tickets for one technician, and we'll score them live for clarity, tone, and follow-up.