Pollup – NRR from Signals

More and more I’m seeing people refer to any AI tool as “agents”.

I understand the need for marketers to simplify things… especially when it comes to AI and its ever-changing complexities.

But for real-life application in business, it’s essential we understand the different levels of autonomy that tools give – as they determine productivity, risks, and compliance needs. And the reality is: Very few commonplace tools are actually AI Agents. At best, they’re AI Assistants.

Approaching this topic from the lens of Customer Success in B2B for my audience, here’s a breakdown of different AI tools, what they actually do, and how CSMs can utilise them in their roles:

1. Assistant = digital task helper

An assistant helps with specific tasks. It’s reactive: you give input, it responds.

Examples:

Drafting email replies
Clustering CSAT feedback
Analyzing KPI trends
Generating video meeting transcripts and summaries

Typical Tools/Functions:

ChatGPT
Copilot (Microsoft), AI Companion (Zoom, Salesforce)
AI features in CRM systems

2. Automations = predefined workflows

Automations execute routine processes based on clear rules. There is no interpretation, just predefined logic.

There are two main types of automation: user-friendly, low-code tools (like Zapier or Make) and integrated or custom-built automations within platforms or frameworks, which can range from basic rules to complex, AI-powered logic.

Typical Use Cases:

Basic Automations:

Reminders after inactivity
Escalations for negative feedback
Follow-up tasks post-onboarding
Automatic update of lead status in CRM
Confirmation email after booking

AI supported Automations - Helps for example personalizing customer communications:

All automated emails fully personalized (not only the name of the person but the whole content) thanks to a LLM generation module
Segment of one approach thanks to a Context engine module
Sending the right resource to the right customer in the right moment thanks to a Knowledge Base module
Interactive Q&A (instead of static FAQs)

Data Integration Map for Email Personalization

Advanced AI supported Automations:

Support Chatbot

3. AI Agent = digital team member

An agent receives a goal (i.e. reactivating inactive users) and decides how to reach it. It can make decisions, take actions, combine tools, and adapt steps dynamically.

Note: "Agent" is a technical concept. Most AI assistants are not true agents – they just appear smart.

AI agents go beyond assistants and automations by acting independently toward a defined goal. While assistants wait for prompts and automations follow predefined rules, AI agents make decisions, trigger actions, and adapt based on real-time data. In the context of AI integration, agents represent a shift from supporting tasks to actively managing them – making them powerful teammates in a hybrid setup.

Typical Use Cases:

AI agents can handle a large volume of customer inquiries, weigh up the type of problem, level of urgency, client history, and take the best course of action using reasoning.
AI agents can continually analyze customer and user behaviour, and using logic, send the most relevant personalized check-ins, nudges, surveys, or follow-up messages in the moment.

Utilising AI agents in Customer Success is like having another team member who can process data quickly, make decisions based on their training, and act automatically without pre-set prompts. So really, the possibilities are endless!

❗Are there plug-and-play agent tools for Customer Success?

Not really, and this is not the right approach. Start mapping your processes and integrating AI step by step.

The step of introducing AI agents requires more than just technology. It demands context. Unlike simple automations or assistants, AI agents need to understand the environment they operate in. They must access structured data, follow clearly defined processes, and rely on explicit logic to make the right decisions independently.

Without this foundation, agents risk making irrelevant or even harmful choices. That’s why many AI projects fail – not because the model is bad, but because the system around it lacks clarity.

AI agents don’t just learn from data; they learn within systems. So no structure = no intelligence.

➡️ Real plug-and-play agents for Customer Success are rare. Today, most are experimental or developer-driven. (So don’t believe everything you read in a LinkedIn ad 😉)

4. How do I choose which process to automate first?

I wrote an automation starter guide a few weeks ago.

When talking with users about this, I realized the most difficult point for them was where to start!

So I’ve now created this decision map.This table helps evaluate which processes are suitable for automation. By assessing factors like repetitiveness, time effort, data structure, and flexibility needs, you can make informed decisions about where automation creates the most value.

Which tool to choose?

Another difficult part is which tool to choose.

Though most teams have a tech-savvy, process aware person who can help start this journey – it’s worth checking with the team if someone feels like using one of the no/low code tools existing in this realm. These can give you more leadership over your process – even if you work with a tech partner later. It’s better to own these processes and understand the basics from the get-go!

This comparison helps you choose the right automation tool based on your technical needs, data privacy requirements, and budget. It contrasts 3 popular platforms – Zapier, Make, and n8n – across key criteria like flexibility, workflow complexity, hosting options, and ease of use.

Okay, I’ll leave it there for this edition. That was a lot of info… Feeling overwhelmed? Do not hesitate to contact me, or provide feedback on what would actually help you! I’d love to hear.