Customer support has been a big area of focus for CloseBot since its inception in November 2022, but even more work has been done on support systems in the past few months to allow for 24/7 human support. Let’s take a look at what it takes to operate a 24/7 support system and lessons we’ve learned along the way. If you’re offering support to clients or customers, hopefully you can learn some lessons here before you reach some of the roadblocks we experienced.
Lessons Learned at Each Stage
The following are the stages of growth our support system went through, with some lessons learned at each stage. You can see what we offered for support at that time and what we did well at and what failures we experienced that helped us reach the next level in our systems.
Getting Started
1-10 support requests per week
1 support staff (just me)
40 hours weekly support
November 2022
When CloseBot first started, it was just me ( Bryce – CEO and co-founder). I did the dev work, the coding, the finances, legal, etc. I also did support. In the beginning, this just meant responding to support requests via email. Not a big deal with 1-3 support requests per week across a dozen customers.
If I had to start over, I would do the same thing. It’s important to get up and running quickly and also important for the head of product to do support. This way they can feel the user’s pain early and steer the product gradually into a strong product-market fit. Working support myself allowed me to iterate extremely quickly this is where CloseBot began to take root in its success.
Starting to Scale
10-40 support requests per week
4 support staff (me + 3 others)
60 hours weekly support
Breaking point: July 2023
Eight months after launching CloseBot, email support wasn’t cutting it. I found my time living in my email, struggling to label and respond to everyone in a timely manner. That’s when the company adopted Intercom. At this time there were an average of 20 new support requests each week. We changed over to 100% chat support (no tickets and no emails). This was nice because it allowed the full support team (now 4 people) to work together on support items. We also turned on Intercom’s AI at this time, which allowed us to deflect about 50% of all incoming support messages by having the AI answer questions based on its training from our help documents.
If I had to start over, I would have implemented both a chat and ticketing based support system at this time. This is something we eventually did, but doing it sooner would have helped manage user expectations around response times for more complex dev issues in work. Without tickets, and without adequate auto-messages about our non-24/7 coverage, we had issues with users starting to become upset over long wait times.
Continued Growth
40-100 support requests per week
6 support staff (me + 5 others)
80 hours weekly support
Breaking point: November 2023
This was a difficult time for support. CloseBot was gaining massive traction and, therefore, we were getting more and more support inquiries. I tried to solve the scaling problem by doing more of the same thing… we had AI deflecting, we added more support staff members, we covered more hours. We experienced problems with shift changes leaving some users being “left on read” without a reply for long periods of time… the support staff members would sometimes have internet outages and we would be unaware of their absence… the support team lacked guidance because I was overrun with coding work.
At this point, we appointed our senior support member to manager. He provided the leadership the other support staff needed and worked on implementing automations within Intercom to auto-notify and auto-route based on shift changes. He was also able to monitor performance and raise any concerns about support quality or dev issues that he noticed. This is something that should have been implemented much sooner. With a support team of 3 or more, it’s important to have someone who isn’t the CEO leading the support team. It’s also important that you have automations in place early to route based on staff availability. As you scale, this cannot be done manually.
Bust
900 support requests per week
13 support staff
168 hours weekly support
Breaking Point: June 2025
We continued with the systems mentioned for quite some time. Was it the best? No. Did we continue to improve our systems with automations and routing logic? Yes. Improvement on systems in addition to our improvements to our help docs (and thus improvements to the Support AI), helped us scale alright.
What Happened
But then came the V2 launch and the ZappyChat acquisition. This introduced many new factors, some of which we had attempted to prepare for.
✅ Spike in New Users – we spent a month hiring and training 5 new support team members in preparation for the launch.
❌ Higher User Expectations – when V1 was first introduced, it was the first AI platform many had ever seen… now AI platforms are common-place and users expect more! This is understandable, but the bugs V2 introduced left some users feeling lost.
❌ Higher Product Complexity – the newest version of CloseBot is amazing… but there are more features to learn as well. This raises more questions.
❌ User Interface Troubleshooting Issues – V1 had message logs that users depended on. When V2 first came out, it didn’t have these at all. This left both our Level 1 support team and our users confused about why agents said things.
How it was Fixed
We heard our users loud and clear. They were screaming for stability and support… and the CloseBot team jumped into action.
Level 2 Support Team – We established a Level 2 support team comprised of more advanced individuals. Level 1 can now elevate here when they are feeling lost. The Level 2 team helps to solve these more complex issues, and then trains the Level 1 team on what the resolution was.
Ticketing System – We implemented a ticketing system for higher-complexity dev items. This way, Level 1 support could focus on quick wins with back-and-forth chat, without an inbox crowded with complex issues. Level 2 support can elevate here if they find that something is likely to be a dev issue and not user-error.
Hard Stop on New Features – We immediately put a stop on all new work on the dev side. All focus went immediately into stabilization. This is where we’ve been putting our focus still 4 weeks later.
10 Hours Office Hours – We now offer almost 10 hours of office hours calls every week in addition to 24/7 human support. We initially put this in place in an attempt to get quicker feedback loops with our users, but have left it in place as we’ve seen many prefer this group-zoom setting.
User Studies – We’ve conducted user studies to see what our users are struggling with, and have started to roll out with User Interface updates that will help to make things more intuitive (easier to self-resolve issues).
Documentation and AI Support – Our previous Level 1 manager has moved into the role of optimizing the AI Intercom support and internal automations. It’s his primary goal to make sure automations can solve as many user inquiries as possible without needing a human.
Focus on Training – The Level 2 team now has regular training times with the product manager and the Level 1 team has regular trainings with the level 2 team. The Level 1 trainings are recorded and kept in a library that they can reference, even if they are working hours where they cannot attend these trainings live.
Summary
It takes a lot to run a company… but it takes more than most can imagine to offer 24/7 support for a scaling company. At CloseBot, our users are most important to us. We’ve spent considerable time listening to user feedback in order to improve our support systems to the point they’re at today and we will continue to improve for all of our users.
