In this episode of Mastering CS: Candid Leader Insights, Irina Cismas sits down with Nir Kalish, VP of Customer Success at Blink Ops, an AI platform that helps security teams automate their operations at scale. Nir comes from a world where AI is the product, which makes his perspective on whether AI can replace CS teams particularly compelling.
He shares how Blink Ops uses its own platform to build internal CS tools, why he believes white-glove customer success cannot be replaced by agents, what a personal moment with a customer taught him about the limits of AI, and how he thinks about the build versus buy decision in a world where anyone can now create their own tools over a weekend.
What You’ll Learn
- What Blink Ops does and how the CS and support teams are structured
- Why an AI company still invests heavily in human customer success
- What AI can and cannot do in a white-glove CS environment
- How Blink Ops uses its own platform to build internal CS tools and agents
- Why over-relying on AI is just as dangerous as ignoring it
- How Nir thinks about the build versus buy decision for CS tools
- What the ratio of CSM to customers looks like with AI assistance versus without
- Why certain roles will be disrupted by AI in the next two years while CS will not
Key Insights & Takeaways
AI cannot replace relationship-building. No agent can walk into a conversation, sense that something is off, and create a genuine human moment with a client. That is still entirely human territory.
White-glove CS is safe, long-tail SMB CS is not. If you are running a high-touch enterprise CS motion, AI will not replace your team. If you are running a volume-based SMB model, it will significantly reduce headcount.
AI is a multiplier, not a replacement. With the right tools, a CSM who used to manage 10 accounts can manage 20. That is a meaningful efficiency gain without losing the human touch.
Using AI too much makes you dependent and uncritical. AI hallucinates, invents numbers, and misses important data points. There always needs to be a human in the loop who can catch the errors.
Build versus buy is still the right question. The tools have changed but the question hasn’t. Is it cheaper and smarter to build it and maintain it yourself, or to pay for a service that maintains it for you?
AI is not a magic wand. If you trust it too much and treat it like a Harry Potter solution, you are going to crash. Treat it like a worker: give it direction, test the output, and give it feedback.
Balance is everything. Not left or right, not black or white. The right answer is almost always somewhere in the middle, and that applies to how you think about AI in your organization.
Podcast Transcript
Intro
Irina (0:05 – 0:27)
Welcome to Mastering CS Candid Leader Insights, the podcast where we dive into the world of customer success with industry leaders. I’m your host, Irina Cismas, and today I’m joined by Nir Kalish, VP of Customer Success at Blink Ops, an AI platform that helps security teams automate their operations at scale. Nir, I’m really happy to have you here.
Thanks for joining!
Nir (0:28 – 0:29)
Hey, thank you for inviting me!
Irina (0:30 – 0:43)
Let me understand your world at Blink Ops today. What does the company do? Where does CS sit in the business?
And what does the team look like?
Nir (0:43 – 3:45)
BlinkOps is a generative AI platform. We allow security organizations, and not only, because our platform is actually very open. We allow every organization that needs or wants to move from non-AI to a more AI-centric way of working to do it.
We have a platform that allows companies and departments to build agents and automations, and by that reduce load from employees and increase their productivity. If we’re talking about security specifically, having an AI platform that can help you run and build security automations means you can actually secure the business faster and increase productivity and profitability, because we can reduce the cost of doing things manually.
We have different services. We have an automation engine where you can build automations using our AI. We have more than 30,000 integrations. And we have generative AI, meaning you can build your own agents and micro-agents. One of the things that differentiates us from other agentic platforms is that our agents are very secure, because we originally built this for security organizations. You can literally tell the agent what to do and what not to do, which limits the power of the agent. It’s enterprise-grade agentic AI.
The third service is what we call AI as a service. We come to organizations, map and understand their AI needs and how to become an AI company, and we have our own solutions organization that can help companies shift to being AI-first much faster.
On the team side, under my organization there are two teams: customer success and support. Support is reactive, like any other support function. They address tickets and customer questions. The customer success managers are the ones who take ownership of the customer once they sign. They own onboarding, training, showing business value on every interaction, and growing the customer into renewals while building the path for upsells and cross-selling, working together with the account executives to achieve those opportunities. And of course the standard CS responsibilities: ongoing meetings, identifying risks, building risk mitigation plans, and following through on them.
Do You Still Need a CS Team When You Sell AI?
Irina (3:45 – 4:38)
Can I ask you a very bold and direct question? You work for an AI company that builds agents for others. How come you still have a CS team and a support team made up of people and not agents? Do you still need them? Because that’s the question everyone is asking. I was at an event with our CEO and other heads of CS and we were discussing the same thing. Do we still need a CS team, and in which cases, if AI can solve everything we want and we can have an army of agents?
Nir (4:39 – 6:58)
Let me set the expectations. If someone thinks that AI can replace people, they’re wrong. AI can replace some jobs, yes, but not all of them. AI has problems. I work with Claude, ChatGPT, and Gemini every day, and they don’t do everything perfectly. That’s one thing. Second, it depends on the situation.
In our case, we sell to enterprise. We are a white-glove CS organization. At the end of the day, a CISO doesn’t want to speak with an AI. He wants to speak with someone like me or one of my team members, where we can guide them, ask questions, and create a personal interaction.
Will AI replace some CS roles? Yes. If you’re running long-tail SMB, AI will probably cut a lot of your team members. But do you need to do it right now? I think every CS leader who chooses to replace their CS employees with AI is going to regret it very fast. Because there are things that AI cannot do. Building a relationship is something AI cannot do. It doesn’t matter how well it talks or how nice it is, it cannot create a real relationship.
Why? Because it cannot stop a meeting, turn off Zoom, and say, hey John, I feel that something is off, tell me what happened. And John says, I have a lot on my mind. And you ask, tell me more. And he says, my son has a brain tumor and most of my energy goes to him. I’m using this example because that’s something that happened to me a few years ago. There is no AI in the world that can create that kind of personal, intimate moment with a customer that down the road leads to more business. The right business. There is no AI in the world that will do that.
And we do use AI. We have agents that we build with our own platform.
How Blink Ops Uses Its Own Platform to Build Internal CS Tools
Irina (6:59 – 7:03)
Do you have those agents that you built with your own platform to solve your CS challenge?
Nir (7:04 – 9:46)
We leverage a lot of AI tools, our own and others. We use AI as a multiplier for the CSM and for support. There are things that AI handles so the human can focus on what matters.
We leverage AI to help us write emails in the tone we want, and to analyze calls in a much better way than existing tools. We built something that is now being used across the company, originally built by one of our AEs: we analyze all our calls and at any given moment, before going into a meeting, we can ask for a snapshot of the customer. It analyzes all the business calls, extracts information, and gives us a status overview of the customer.
We also use AI for data analysis. I built an agent inside Gemini, and soon I’m going to move it to Claude, that acts as a CS helper. It has 10 roles and the CSM can consult it: hey, I have a customer that wants to churn. And it will ask tough questions. I literally built the agent to speak like me. I’m very direct and I can be challenging in my questions, and the agent challenges the CSM the same way. It helps them address competitive analysis, churn requests, how to prepare for EBRs, all of those things. It’s just a multiplier.
Can you replace the CSM? Not today, and I don’t think in the next five years either. If you are a high-touch, white-glove CS organization, you cannot replace your team with agents. You can use agents to do more.
You might need fewer people because each person can manage more customers. If the usual white-glove ratio is 1 to 10 or 1 to 5 depending on complexity, with AI you can move to 1 to 20. That’s a huge saving for a company, but it’s not replacing people. It helps, not replaces.
Irina (9:47 – 10:39)
I think this is a very strong statement, and I’m glad I asked you, because you represent an AI company. If an AI company is still protecting and investing in its team, that means there’s something to it. You need to have a balance. This is what I consider the gray area. It’s not black or white. It’s in the middle. You have to embrace both: the technological trend and the reality we all live in, but also adapt it to your own reality and figure out what, from the technological advances we are living through, actually helps you do a better job.
Nir (10:39 – 10:50)
And by the way, also, when we sell our service, a lot of AI companies come and sell the dream. You won’t need people. You will be able to fire everybody, and the AI…
Irina (10:50 – 11:01)
Everybody thinks it’s like, you know what, now I have Claude. I can fire half of my team, or if not, I can be me. One company, and I can do them all, because I have Claude.
Nir (11:01 – 12:15)
When we sell, we don’t sell it that way, because we really believe in the right approach. We come to CISOs and executives and tell them: you already have pressure where you cannot hire more people. That pressure exists across all companies. You need to do more with less, increase team productivity, and secure your organization more. With AI, securing the organization becomes an even bigger challenge, because bad actors are also using AI to do more. It takes time to build security automations, and you need to enable the business, which today runs fast. We can help you do that.
But it’s not going to make you want to eliminate all the people in your organization and leave only one person. That’s not what we want you to do. We want you to take the team you have and allow them to do more, so that in every single hour they work for you, they can do double or triple the work, and then they go home. You still gain from that.
Because here’s the thing: everybody thinks AI is a magic trick. But let’s be honest. If you start using AI more, your AI costs are going to increase.
Irina (12:15 – 12:29)
I know, I know. And nobody speaks about that. So I’m just expecting the moment when they will realize how much the AI costs if it’s not built and used carefully all along the way.
Nir (12:30 – 13:20)
I’ll be even more honest with you. I’ve worked with all the AI tools since they started, since ChatGPT launched, I was there. Using AI too much also has a negative impact. It makes us dependent and it makes us stop questioning what we see.
There are enough stories from the last six months of companies that moved everything to AI, took decisions based on it, only to discover six months later that the AI was wrong. I have cases from my own personal life where I said, wait, you’re lying to me. I asked it, why are you inventing numbers? And it said, oh, you caught me. Great.
Irina (13:20 – 13:23)
Really?
Nir (13:23 – 16:05)
This is why we cannot just shift everything to AI. There needs to be someone who walks on two legs, has a brain, and is human enough to decide what is a good AI outcome and what actually needs to be better.
There are jobs that AI can replace. If I need to predict the future, we are probably two years away from seeing one person in marketing. Maybe zero BDRs or SDRs because AI will do that for them. Engineering is probably going to have a big impact. You can now build an entire system yourself. I literally build apps on weekends using AI with no help from anyone, and the last time I wrote code was about 15 years ago. So there are going to be departments and organizations that AI will hit badly, and people there need to understand it and start thinking about what comes next.
But in customer success, sales, and solution engineering, I don’t see it happening in the next five or six years. In order to fully replace CSMs and salespeople, we would need to reach a place where companies speak with bots, bots speak with other bots, and humans feel fully comfortable trusting AI. We are not there yet. There is a reason we still don’t have fully autonomous cars. They are driving, but not across the board. And I still ask myself how many people are willing to board a plane with no pilot.
When we get to that place, roles that require human interaction will start to be replaced. I don’t think that’s happening in the next five years. AI is going to have an impact, some positive and some negative, on society, and we will need to address it. Not in a sci-fi way, but in a very practical way: what do you do with a lot of people who can no longer work in their field? How do you take care of them? We need to take care of society, and that’s a complexity that AI also creates.
The Build Versus Buy Question in a World Where Anyone Can Build
Irina (16:06 – 17:26)
I have another question. You said that you no longer need tools because you can build what you need with AI. Funny enough, yesterday evening I was talking with a head of CS who told me that in her company, the philosophy is build before you buy. She said, Irina, I have these conversations with my manager where I come and say, look, I found this awesome tool, I can adopt it, I can use it. And the response is always, great, but how can you build it with Claude?
So a head of CS who was supposed to be doing CS is now expected to build all the tools she needs, because that’s the trend. Where do you stand on this? You also mentioned building things, and funny enough, she also said during her weekends. And what I can tell you is that from the moment Claude entered my life and I started building things for myself, I work twice as much as I did before AI.
Nir (17:27 – 17:28)
Just to make it clear.
Irina (17:28 – 17:39)
He’s telling me, oh, you use AI and okay, your job is going to be at 50%. Hell no, I’m using AI and I’m working at 200%.
Why AI Is a Multiplier, Not a Magic Wand
Nir (17:40 – 25:59)
I want to make it clear. When I say I’m working on weekends and building tools, I’m building things for my personal life, not related to work. I actually built one app that replaced three apps in my personal life.
I don’t believe in working 200% or 150%. There is work and there is life, and work-life balance is important. Sometimes there are crises, but 99% of the time you don’t need to work those hours. When I’m talking about weekends, I’m building personal stuff, not work-related stuff.
Every department has always built tools. In the past, you’d open a request for engineering to give you a developer. It took much more resources. Today, every person in your company can build tools. The question stays the same as it always was: do I build it or do I buy it? Today it’s become much easier, faster, and cheaper to build it yourself. The question is still: does the off-the-shelf tool give me everything I need, or does it only give me 50%? Is it cheaper for me to build it, maintain it, and have it exactly the way I want?
Let me give you a personal example. I had three apps on my phone that I used in my morning and evening routine. They were good, I used them for more than 17 months, but they didn’t give me everything I wanted and they cost money. So I spent a weekend designing my own app. I used Gemini as my product manager, CTO, and UX expert, and designed all three apps into one. Then I went to Claude to write the code, and used ChatGPT as the QA tester. After two days, I had one app with everything I needed and I could delete the three others.
I think down the road we will see CS and sales organizations building their own tools that do exactly what they need without paying $120 per seat. The question will always be the same: do you know what you want to build, and is it cheaper to build and maintain it versus buying a service?
I also built a travel agent inside Claude, probably one of the best travel agents out there, because I gave it all the rules I’ve ever thought about from my experience, including going through travel vlogs to extract information about trips and checking a lot of small details. Every time I need to plan a trip now, it takes me minutes. It finds cheaper flights and hotels based on my standards in the background. Could I use a travel agent? Of course. But soon, everyone will be able to build something like this themselves.
Where I draw the line is something like training an AI language model. If someone builds an AI model dedicated to customer success or sales, it will be incredibly powerful. But I’m not going to build it myself because training such a model would cost five to ten million dollars. If I ever need it, I’ll go to a service that already has it and pay $100,000 a year.
That’s where every leader and CEO needs to think. It’s not about building everything in-house. I’ve been around long enough, through the dot-com bubble, 2008, big data, machine learning, to know that the extremes never work. Big data was supposed to solve everything. It didn’t. We’re still drowning in data. Machine learning was supposed to make your life easy. Not if you don’t set it up correctly.
Same goes with AI. It’s a great multiplier, but if you trust it too much and treat it like a magic wand, you’re going to crash, and that crash is going to be painful.
I treat AI like another member of my team. I follow through on its work, I test it, I give it feedback. The only difference is it can work 24 hours. The downside is that it sometimes doesn’t ask questions. A real employee would ask, hey Nir, what did you mean by ABC? AI, with all the respect, is smart and yet stupid at the same time. Last night it gave me a great idea and showed me angles I hadn’t even thought about. At the same time, when I gave it data analysis to do, it missed some important points.
It’s always the balance. AI won’t replace all your employees. Is it easier to build a one-person company today? Yes, much easier than before. Do you need to hire developers in the first year? Probably not. But will it be perfect? No.
And remember, it’s not only about building. You also need to maintain it. Sometimes maintaining what you built is much more expensive over time than just paying for a service that maintains it for you.
Irina (26:02 – 26:38)
This was an unexpected conversation. And thank you for your honesty and for everything that you shared during the last half an hour. It’s probably, honestly, one of the best episodes that I’ve done and one of the most authentic and to the point.
So thank you for joining me. And thanks for tuning in. Until next time, stay curious, keep learning and mastering customer success.
Thank you.
