Using AI for Research and Insights

Introduction

Ten tabs, three dashboards, a wall of notes, and you staring at them all trying to piece together what your customer is thinking. This is the story of thousands of Customer Success professionals. They spend hours gathering information, but when they walk into meetings they still can’t shake off that feeling: “Am I missing something important?”

In Customer Success, information is everything. The more you know about your customers, their goals, challenges, and sentiment, the more effectively you can guide them toward success. But between the dozens of accounts you have to manage, and the scattered data you’re forced to deal with, getting clarity from all that information is a nightmare. 

AI can be incredibly helpful for streamlining your prep work. But the way most people use it for research barely scratches the surface. 

In this module, you’ll learn how to use AI to transform your research and insights process:

• How to use AI tools for account research, EBR prep, and connecting external insights with internal data.
• How to identify patterns, trends, and key metrics with AI, making your data work harder for you.
• How to analyze customer feedback, call transcripts, and survey responses to uncover the why behind the what.
• How to present your AI-driven insights in a clear, strategic, and actionable way that drives alignment across teams.

By the end, you might even change your mind about research being a chore, and start seeing it as one of your most powerful levers for driving customer success.

Chapter 1: AI for Research

Back in the day, the only way we could research was to rely on manual searching and fragmented data. You look at CRM entries, revisit old notes, and maybe even check LinkedIn for company updates. The problem with this was that it was time-consuming, you didn’t have the bandwidth to go deep, and the results you got were basically surface-level. You saw what happened, but not why it mattered. Most Customer Success Managers just end up reacting to what’s in front of them rather than anticipating what’s coming next.

AI-powered account intelligence completely transforms that equation. Instead of spending hours struggling to piece together data, AI  can gather it all together. Instead of waiting for customer signals like reduced usage or a support ticket spike, AI can help you predict them. 

On top of being a timesaver, this changes the quality of your preparation. When the AI gives you insights from months of data, flags trends before they escalate, and suggests next steps that align with the customer’s evolving goals, you have the potential to make your org and your customers very happy. 

Using AI assistants as your research engine

Good research is the secret behind better conversations. In Customer Success, research helps you ask smarter questions, tailor recommendations, and connect your product’s value to the customer’s real-world goals. Whether you’re planning a renewal strategy, preparing for a QBR, joining a customer’s kickoff call, or just catching up internally, having the right context makes you sound informed, and your actions purposeful. 

Tools like ChatGPT or Perplexity can act as your personal research analyst. This can be incredibly useful for internal update meetings, QBRs, EBRs and kick-off calls. 

Example:

• Summarize the top three trends shaping the [Industry Name] industry this year.
• What goals are top executives in [Industry Name] emphasizing in recent interviews or reports?

This kind of research gives you broader context, like market shifts, leadership updates, and competitor moves. It will help you understand where your customer is heading and what pressures they’re responding to.

Then you move to more specific research related to customer data. You can feed the AI background information such as customer notes, past meeting summaries, or onboarding goals to help it build a clear narrative of who the customer is and what they care about most. 

Example:

• Here’s a summary of a customer’s onboarding goals, meeting notes, and recent updates. Summarize their key business objectives, current challenges, and how our product supports their success so far.

Making connections in research

The two-pronged approach of general and specific research that we talked about earlier helps you connect what’s happening in your customer’s industry with how your product fits into their strategy. That bridge is where great Customer Success conversations happen. But did you know that you can even get the AI to help draw those links?

Example:

• Explain how the latest trends in [Industry Name], such as [Trend 1] and [Trend 2], might impact how our customer uses [Your Product Name]. Include potential risks, opportunities, and recommendations for our next Executive Business Review.

To help in making connections between data points, as well as deeper research, tools like Notebook LM can be quite convenient. It’s designed for situations where you’re working with lots of unstructured information like meeting notes, call transcripts, customer briefs, or internal reports. 

You can upload multiple files at once and ask NotebookLM to find the links between them.

Example:

• Identify three recurring priorities mentioned across these accounts.
• Summarize how customer goals have evolved since onboarding.
• Highlight connections between product feedback and business objectives.

Streamlining research

While LLMs can get the job done, it can be tedious to upload data every time you need to prompt. But if you’re using a modern CRM, you may be able to connect it directly with ChatGPT or Claude using Model Context Protocol (MCP). This allows AI tools to securely access and query live customer data without needing manual uploads. 

Even so, these approaches work best when combined with tools that give you quick visibility into what’s been happening with your customers. You might want to try out specialized tools, like the Velaris Headlines feature in Velaris. This can be useful, since it can instantly summarize everything that’s happened with an account in the past quarter from the data the CS platform centralizes. 

One great use case for research using AI is kick-off call prep. Every kick-of call is your chance to position yourself as a true business partner to your customers. By using AI, you can condense hours of research into minutes and equip yourself with information that proves you know the ins and outs of your customer’s industry and challenges. Here’s a template you can follow for kick-off call prep:

Kick-off Call Prep Template (Deep Research Mode)

Act as a Customer Success Manager preparing for a kick-off call with [Customer Name].
Summarize:

1. The company’s top three strategic goals based on public reports and news.
2. Key industry challenges relevant to them.
3. Bios and recent public comments from executives attending the meeting.

Using this template, the outputs you’ll get will make it clear how useful AI is in compressing hours of research into actionable insight. Next time you have a kick-off call come up, try this approach to prepping with AI; you’ll notice how much sharper your conversations become when you walk in with context.

Chapter 2: Quantitative Data Analysis

Data is at the heart of Customer Success. Understanding customer behavior, feature adoption, and engagement trends is what separates reactive account management from proactive, strategic action. And since data is crucial for influencing strategic decisions, being the person who can uncover that data positions you as a trusted voice and grows your influence within an organization.

If you want to capture the attention of executives and become a go-to resource for other teams like Product, Marketing, and Finance, bringing actionable insights to the table through AI data analysis is a no-brainer. It reminds them that you can do more than manage accounts, and that you can drive data-backed decisions across the organization when you combine AI with your unique position in CS.   

But since the time to sift through dashboards and spreadsheets is a luxury most CS professionals can’t afford, there’s a risk of missing patterns that could inform renewals, expansions, or risk mitigation. And so we turn to AI. 

Back then, you had to be a technical wizard capable of mastering SQL queries, building complex Excel formulas, or spending hours cleaning raw data. But the great thing about AI is that you no longer need these deep technical skills to analyze data and generate actionable insights.

With the right framework, even non-technical CSMs can use AI to analyze complex datasets efficiently. One highly effective approach for quantitative data analysis is the DIG framework: Description, Introspection, Goal-setting. 

Step 1: Description

Start by having the AI describe the dataset. Ask it to list columns, show sample values, and flag inconsistencies or missing data. This step ensures you understand the format, scope, and potential limitations of the dataset before diving into analysis.

Pro Tips:

• Pull multiple samples for each column to detect inconsistencies or outliers.
• Ask the AI to run a data quality check, highlighting missing values, unusual formats, or potential errors.

Example:‍

Analyze the attached customer usage dataset. List all columns, show five sample entries per column, and flag any missing or inconsistent values that could affect analysis.

Step 2: Introspection

Next, have the AI brainstorm questions that the dataset could answer. Ask the AI to justify why each suggested question is valuable. This step helps you discover insights you might not have considered and confirms that the AI has interpreted the data correctly.

Pro Tips:

• Include follow-up prompts to confirm which columns or metrics are required to answer each question.
  
• Identify gaps where additional data may be needed to answer important business questions.

Example:‍

Based on the attached dataset, suggest 10 questions we could answer about customer health and feature adoption, and explain why each is valuable. For the first three questions, specify which columns are needed and whether the current data is sufficient.

Step 3: Goal-setting

Finally, define what you want to achieve with the analysis. A clear goal allows AI to prioritize the relevant columns, metrics, and insights, ensuring outputs are actionable rather than technically correct but practically useless.

Pro Tips:

• Frame the goal in terms of customer impact (e.g., identifying underutilized features, detecting at-risk accounts).
• Use follow-up prompts to create step-by-step analysis plans, including cleaning, ranking, and visualizing metrics.
• Ask the AI to anticipate questions stakeholders might have about the analysis.

Example Prompt:‍

Our goal is to identify which features are driving adoption among high-touch accounts this quarter. Using the attached product usage and engagement dataset, highlight the top features by adoption rate and suggest opportunities for targeted engagement campaigns.

Common Data Prompt Examples

1. Feature Adoption by Account Tier

Analyze the attached dataset of feature usage for the past quarter. For each account tier, calculate the number of accounts that have used each feature at least once, the average number of uses per feature, and the adoption rate per tier. Present the results in a table and highlight the top 3 features by adoption for each tier.

2. Churn Risk Analysis

Our goal is to identify high churn-risk accounts. Using the attached subscription dataset, identify accounts that have decreased their usage of at least two key features by more than 30% in the past month. For each account, provide the account name, subscription tier, features with decreased usage, and percentage change. Flag accounts that meet these criteria as 'high churn risk'.

3. Feature Usage Trends Over Time

Our goal is to identify declining features. Analyze the attached dataset of weekly feature usage for the last 12 weeks. For each feature, calculate the total weekly usage, the percentage change week-over-week, and identify features with a decline of more than 15% for two consecutive weeks. Present a summary table showing feature name, weekly totals, percentage changes, and trends flagged as 'declining' or 'stable'.

4. Account Health Metrics Summary

Using the attached dataset, calculate the following for each account: total logins in the past month, number of features used, and number of completed onboarding tasks. Provide a summary table with account name, total logins, feature count, and onboarding completion percentage. Highlight accounts in the bottom 10% for any metric.

5. Revenue Impact by Feature Usage

Our goal is to determine which features are most strongly correlated with revenue. Using the attached dataset with account subscription value and feature usage, calculate the correlation between the number of times each feature was used in the past quarter and the total revenue for that account. Provide a table showing feature name, correlation coefficient, and a brief summary of which features are most strongly associated with higher revenue.

Chapter 3: AI for Qualitative Data Analysis 

While numbers can tell you what is happening, qualitative data is ideal for understanding the “why”. But in Customer Success, that “why” is usually hidden inside open-text feedback, survey responses, call transcripts, and customer emails. And that’s where the problem is: there’s just too much of it.

It’s more or less impossible to consistently produce insights by reading through hundreds of lines of customer feedback or call transcripts, especially with the limited bandwidth CSMs have. 

Although AI was once mainly used for crunching numbers and analyzing quantitative data, it’s now become remarkably good at understanding qualitative information too. It can process, categorize, and interpret qualitative data, giving you insights much more efficiently than if you were to attempt this manually. 

Using AI for Voice of Customer (VoC) analysis

Voice of Customer data is worth its weight in gold, but only if you can organize and interpret it. Luckily, AI is quite good at doing tasks like automatically tagging feedback by topic, detecting sentiment, and even suggesting product improvements based on what customers are saying most often.

If you provide data like NPS or CSAT survey comments, support conversations, community posts, call summaries, and transcripts, it can analyze and spot sentiment patterns across different sources.

Example:

• Review these 200 survey responses and support tickets, and list the top five recurring customer concerns.
  
• Using the attached data sources, identify which features receive the most positive sentiment and which ones generate frustration.
  
• Summarize the main themes in this customer feedback from the last quarter. Group them by product area and highlight any patterns in negative sentiment.

Not only is this helpful for you in CS, you can pass your insights to Product and Marketing teams and help them prioritize what really matters.

If your CS platform integrates qualitative data from multiple channels, like Velaris does, AI can go a step further. Velaris’ Trending Topics feature helps uncover patterns in unstructured data, like recurring themes in customer feedback, support tickets, and survey responses. By spotting what topics are gaining traction or causing concern, you can identify emerging risks early and prioritize actions that fit their evolving needs.

Turning insights into action

A common misconception is that qualitative analysis is solely about understanding your customers better to build better relationships. While that’s part of it, qualitative insights are also key in shaping what you do next. 

Once you have your AI-generated insights, make them part of your regular workflows:

• Share a monthly feedback digest with your product team.
• Capture the actual words customers use to describe their wins, challenges, and outcomes. Feed those phrases into marketing campaigns, product copy, or renewal decks for more authentic, resonant communication.
• Update your health scoring model with AI-identified sentiment trends.
• Use qualitative patterns to update your onboarding or renewal playbooks. For example, if feedback shows frustration around setup complexity, build personalized check-ins into your onboarding sequence.

Exercise: Build This Month’s Voice of Customer (VoC) Report

It’s time to put your AI-assisted qualitative analysis skills to the test. In this exercise, you’ll build a Voice of Customer (VoC) Report using AI to transform raw feedback into structured insights.

Step 1: Identify key themes

Provide the AI with a dataset of customer feedback (survey comments, NPS responses, or support transcripts). Ask your AI tool to analyze the dataset and extract the top 5 positive and top 5 negative themes.

Example prompt:‍

Analyze this customer feedback dataset and list the top 5 positive and top 5 negative themes. Include sentiment scores for each theme and summarize how frequently each one appears.

Step 2: Surface the ‘Insight of the Month’

Look for one powerful finding that captures a shift in customer sentiment or an unexpected trend. This should be something that sparks discussion and drives action.

Example prompt:

From the identified themes, choose one high-impact insight that stands out due to frequency or emotional intensity. Summarize it in two sentences and include one or two direct customer quotes that illustrate it.

Step 3: Turn insights into recommendations

Translate those insights into practical next steps. Think of them as mini action plans for your internal teams.

Example prompt:

Based on these findings, suggest one actionable recommendation for the product team and one for the marketing team. Each recommendation should include rationale from the customer insights.

Once completed, you should have a report that is ready to be shared with your team. The more qualitative insights are integrated into your workflows, the more connected your team is to the customer voice. And AI is the most efficient way to discover these insights at scale.

In the next chapter, we’ll talk about how to communicate and distribute your findings, so that you aren’t the only person who benefits from AI-generated insights. Your whole company can get value from these insights, as long as you deliver them right. 

Chapter 4: Communicating Your Findings 

So you’ve done your research, spotted the patterns, and discovered useful insights through data analysis. Now comes the real test: can you get people to care? 

How you present your findings can make the difference between it being filed away or shaping next quarter’s strategy. Let’s find out how you can make the latter more likely. 

From analysis to narrative

Flooding your team with data is never a good idea. We need them to make sense of it, and we need them to recognize its impact. So what we need is a story that helps them see what’s happening and understand why it matters.

Instead of manually writing out this narrative, you can prompt the AI to compile your findings in the desired format.

Example:

Turn these customer insights into a short summary I can present in a team meeting. Focus on what customers value most, what’s frustrating them, and how we can respond.

Tailoring insights for different audiences

Different teams need different kinds of context. A good communicator knows how to translate customer insights into each team’s language:

• For product teams: Focus on themes around usability, feature requests, or adoption friction.
  Example:
  ‍“Seventy percent of enterprise customers mention confusion during setup. Can we simplify onboarding steps?”

• For marketing: Emphasize customer language, wins, and emotional triggers that can inform messaging.
  Example:
  ‍“Customers consistently use the phrase ‘time to value’ when describing success. Let’s feature that in upcoming campaigns.”

• For leadership: Connect qualitative findings to strategic outcomes like retention, growth, and satisfaction trends.
  Example:
  ‍“We’re seeing positive sentiment spikes from accounts that received early training, correlating with 15% higher renewal likelihood.”

Using visuals to tell your story

Qualitative insights don’t have to be walls of text. Visuals make your message easier to absorb and remember.
If you want to make visual aids manually, you can use tools like Canva, Figma, or Notion charts to represent patterns clearly. But using AI to generate the visuals can save you a lot of time. 

Try visual approaches like:

• Theme maps: Cluster related feedback or customer quotes under categories like “Ease of Use” or “Support Responsiveness.”

Example:

From this list of customer feedback, group comments into key themes such as product usability, onboarding experience, and support responsiveness. Create a visual map showing how often each theme appears and how they connect.

• Sentiment heatmaps: Show how customer sentiment changes over time or across segments.
  Example:
  Analyze this dataset of customer survey responses. Create a heatmap showing sentiment intensity (positive, neutral, negative) by month and customer segment.

• Quote boards: Highlight direct customer quotes that capture emotional context. One powerful sentence can hit harder than a chart.
  Example:
  From this collection of feedback comments, select 10 short, impactful customer quotes that represent different sentiments and themes. Format them visually like a quote board I can share in a presentation.

The power of synthesis

Presenting insights alone is useful, but the more you can connect them, the more value they hold. Instead of presenting feedback in isolation, try showing how insights relate to broader company goals.

For instance, if multiple customers mention slow implementation, don’t just flag it as a complaint. Tie it to its business impact:

“Delays in onboarding are contributing to lower expansion opportunities. Streamlining this step could directly improve NRR.”

AI can help you craft these connections by summarizing impact-driven narratives:

Summarize how these customer frustrations could affect retention and how addressing them might improve satisfaction scores.

Embedding insight-sharing into your routine

Insight-sharing shouldn’t be a once in a blue moon sort of thing. Every team should try to make it part of their rhythm. Here are a few simple ways to make it stick:

• Weekly insight highlights: Share one short customer quote or theme in your team’s Slack or Notion workspace.
  
  
• Monthly “voice of the customer” sessions: Present qualitative insights alongside quantitative dashboards to give a fuller picture.
  
  
• Playbook updates: Incorporate new findings into your onboarding, renewal, or advocacy playbooks so every CSM benefits from what’s been learned.

You can also take this a step further with agentic workflows; these are AI systems that not only analyze data, but also act on it. Imagine an AI that automatically detects rising negative sentiment or shifts in engagement and prepares a brief for your next meeting. In the next module, we’ll explore how to build these kinds of automated, intelligent workflows to make your research and insights process hands-free.

Automating Communications 

Once you’ve gathered and analyzed your insights, the next challenge is making sure the right people actually see them. A report that doesn’t go anywhere besides sitting on your table may as well not exist at all. But manually sending them over to stakeholders every single time is a time waster. That’s why you need to give AI-automated communications a try. 

You can use AI tools to automatically generate and share updates or summaries with your team through email, Slack, or other internal tools. Instead of manually compiling weekly or monthly reports, set up workflows where the AI pulls data, interprets it, and crafts clear summaries ready to send.

Here are some ways to practically carry this out:

Automate weekly summaries

• Zapier or Make: Schedule “Every Friday 9:00.”
  Steps: pull health scores and renewals from your CS tool or data warehouse. Then send the data to an AI step (an LLM like ChatGPT)  to summarize. Finally, post to Slack channels for CS, Product, and Leadership.

Example:
Summarize the top five customer trends from the past week using these inputs: health scores, support tickets, feedback comments. Output a Slack ready update with a title line, five bullets, and one next action. Keep it under 1200 characters.

• Slack Workflow Builder: Use a Scheduled trigger, add a “Webhook” or “Form response” step as input, then a “Send a message” step. Paste the AI summary output into a formatted block.

Use trigger-based updates

• Zapier/Make: Use triggers like “new negative NPS,” “sentiment drops below X,” or “feature request count rises above Y.” Then call an AI step with a short prompt and the raw event payload. Finally, post an alert to the right Slack channel or send an email.
  
  
• Velaris automations: Use Trending Topics or health signals as the trigger, then send a concise AI written alert to Slack or email. Keep it to 3 bullets and one next step.

Communicating insights not only helps your team’s cohesiveness, but makes your org take notice of Customer Success as a strategic function. So make use of AI to effectively create and share compelling stories out of raw data. 

Conclusion

You now have the skill to use AI for research and insights! And make no mistake, it's a versatile skill that can completely transform how you prepare, analyze, and communicate in Customer Success.

Here’s a quick recap of what you’ve learned in this module:

• AI for Research: How to use AI tools for account research, market analysis, and connecting external insights with internal data for richer customer context.
  
• AI for Quantitative Data Analysis: How to uncover trends, patterns, and performance metrics with AI to make data-driven decisions.
  
• AI for Qualitative Data Analysis: How to analyze unstructured feedback, survey responses, and conversations to understand the why behind customer behavior.
  
• Communicating and Distributing Findings: How to turn insights into action by automating reports, visualizing feedback, and sharing updates across your team using AI.

Final Exercise: The AI-Powered EBR Challenge

Before we wrap up, why don’t you try putting everything you learned into practice?
In this exercise, you’ll prepare an AI-assisted Executive Business Review (EBR) brief for one of your accounts using the research and analysis skills covered in this module. What you want is an EBR brief that clearly explains your customer’s context, performance, and next opportunities. 

Your task:

1. Start with external research:
   Use ChatGPT with browsing or Perplexity to summarize your customer’s industry landscape.
   Example:
   Summarize the top three market shifts affecting the [Industry Name] industry in 2025.

2. Add internal insights:
   Upload or paste recent customer data into your AI tool.
   Example:
   Analyze this data and highlight top risks, wins, and opportunities for [Customer Name].

3. Connect both types of research:
   Ask the AI to combine these insights into a narrative.
   Example:
   Combine these external and internal insights into a 1-page EBR summary for [Customer Name], including three discussion points and one recommended next step.

4. Visualize or communicate your findings:
   Use AI to turn your summary into a short email or slide for your team or leadership.
   Example:
   Turn this summary into a short, professional update I can share with leadership. Write it as an email that includes a clear subject line, a 3-sentence summary of the customer’s current state, and 3 concise bullet points highlighting wins, risks, and next steps. Keep it under 200 words and make it sound confident and insight-driven.

You can now research like an analyst, think like a strategist, and communicate like a leader. In the next module, we’ll take this even further by building agentic workflows, which are AI systems that can automate your entire research, analysis, and reporting cycle from start to finish.