Hello, this is Matsubara, CRM Evangelist.

"We want to use AI in sales, but we don't know where to start."

There is actually a clear answer to this problem that many small and medium-sized business owners face.

If you're already using a CRM, you're perfectly suited to integrating AI.

Why? Because only CRM possesses the most essential condition for AI to generate highly accurate predictions: "both quantitative and qualitative data accumulated over time."

To perform AI analysis, quantitative and qualitative data accumulated over time are used for pattern matching. CRM systems possess both of these necessary data (customer data and activity data), making them highly compatible with AI.

This article explains why CRM and AI are a good match from a data structure perspective, and provides a clear explanation of how customer behavior prediction works, based on EMOROCO's algorithm design.

Three conditions necessary for AI to make accurate predictions

First, let's clarify what is necessary for AI to produce highly accurate predictions.

Condition ① Quantitative data – “Factual figures”

Purchase amount, number of store visits, number of business negotiations, closing rate—these are all measurable data. AI uses quantitative data to learn patterns such as "what is the probability that this customer will make a purchase" and "what will the sales be this month?"

However, quantitative data alone has its limitations. While we know that "Customer A visits the store three times a month," the numbers don't reveal "why they visit or what might cause them to stop coming."

Condition ② Qualitative data—"Context and emotion"

Customer conversation notes, staff observation records, customer values, emotional changes—these are "contextual data" that cannot be quantified.

Having qualitative data adds context, such as "The person in charge recorded that Person A was comparing it to competing products. It says that quality is more important than price," which dramatically improves the accuracy of the AI's predictions.

However, qualitative data is in "text" format and cannot be directly used for numerical calculations. As will be explained later, the key to success lies in converting qualitative data into numerical data and encoding it.

Condition ③ Time-series characteristics – “Patterns of change”

Only when we have accumulated data in a time series—data from three months ago, last month, this week—can we see "patterns of change."

"This customer's visit frequency has decreased over the past three months, and the notes from the staff member have also become shorter." — This "context of change" is the basis for the AI ​​to determine that the customer has a "high risk of churn."

CRM is the only system that satisfies all three of these conditions.

Why are systems other than CRM "incompatible" with AI?

When compared, the unique features of CRM become apparent.

Accounting Systems (SoR) and AI: Abundant quantitative data. However, only "when and how much" is recorded, with no qualitative data whatsoever. Time series data exists, but the customer's "why" remains unknown. AI cannot predict "the reason why sales have declined."

MA tools (part of SoE) and AI: Behavioral logs (clicks, opens, visits) are quantitatively abundant. However, qualitative data such as "why they clicked" and "what emotions they felt when they viewed it" are lacking. SoE behavioral data alone can only learn "superficial behavioral patterns."

BI (Business Intelligence) and AI: While they can perform analysis, they are tools that "aggregate and display past data" and are not designed to accumulate continuous relationship data with customers. Time-series relationship data does not exist in BI.

CRM and AI: Customer attributes (quantitative), purchase history (quantitative and time-series), contact records (qualitative and time-series), staff notes (qualitative), and emotional changes (qualitative and time-series)—everything is included. It is the only system that has all three elements necessary for AI prediction: quantitative, qualitative, and time-series data.

The Mechanism of CRM x AI: How Data Transforms into "Predictions"

Based on the EMOROCO algorithm design (slide 34), we will explain the process of transforming data into predictions in four steps.

Step 1: Combine quantitative and qualitative data.

CRM stores two types of data in chronological order.

The problem is that quantitative data (numerical values) and qualitative data (text) cannot be directly input into the AI. EMOROCO's algorithm uses a technique that combines qualitative data by encoding time-series quantitative data .

In this way, time-series quantitative and qualitative data are converted into a "sequence of symbols (pattern sequence)" and integrated into a format that can be processed by AI.

Step 2: Match in the pattern database.

The converted sequence is then compared against the EMOROCO pattern database.

Algorithms such as **xForest (Enhanced Random Forest)** and NN (Neural Network)** improve the accuracy of this pattern matching. These are not simple "if A then B" rule-based decisions, but rather machine learning techniques that derive the most probable prediction from countless combinations of patterns.

Step 3: Predict the "next action" using a Markov chain.

Once a pattern is identified, the next step is to predict "what this customer will do next." EMOROCO utilizes a technique called Markov chains .

A Markov chain is a method for calculating the probability of transitioning from the current state to the next state.

The AI, which has learned the chain of "visit → follow-up → purchase" from past customer behavior patterns, can predict and inform you of the perfect timing to follow up to lead to a purchase.

Step 4: Optimize accuracy for your company using "supervised learning".

Typical AI "doesn't have training data, so it needs to be nurtured by the user (accumulating training data and selecting algorithms)"—this is very much like "recruiting a new graduate." It takes 6 to 12 months to train one from scratch.

EMOROCO employs a "supervised learning" model where employees have already completed training on customer service. This means they are ready to be immediately effective as "mid-career hires (4th year or later)."

Even more important is **retraining**.

Based on data obtained from daily activities, the system undergoes retraining, and the more it is used, the more its calculated values ​​are optimized for your company, improving accuracy over time.

The more you use it, the more accurate it becomes—this is the essence of "AI optimized for your company."

A revolution in customer grouping: From "preconception-based classification" to "automated detection."

The most dramatic example of how well AI and CRM work together is in "customer grouping."

Traditional customer classification is done after a person decides, "Let's classify them using these criteria."

EMOROCO's AI automatically calculates which axes to use and how to group customers to highlight their individual characteristics .

The revolutionary value of CRM x AI lies in moving beyond the traditional approach where humans forcibly categorize mysterious groups of people into existing categories, to AI independently discovering "important groups with unknown names."

Prepare for AI with EMOROCO CRM Lite

While the higher-end EMOROCO product integrates AI functionality, the first thing small and medium-sized enterprises (SMEs) should focus on is "accumulating data that AI can learn from." EMOROCO CRM Lite functions as the foundation for this AI readiness.

Three principles for designing high-quality data for AI preparation.

Principle 1: Quantitative data should be stored in a "time series" format.

These are not just isolated purchase records, but rather meaningful as part of a chronological pattern.

Principle 2: Qualitative data should be recorded in a structured manner.

Free-form notes are difficult for AI to process. Recording information using selection or tagging formats brings qualitative data closer to an "encodeable format."

Qualitative data recorded using selection options, checkboxes, and selection tags will be used to improve the accuracy of "encoding → pattern matching" in future AI analysis.

Principle ③: Preserve the context of "why it happened that way."

AI can learn "what happened" from numbers, but it cannot learn "why it happened" without contextual data.

"AI that can be used immediately" vs. "AI that can be nurtured"—realistic choices for small and medium-sized enterprises.

Small and medium-sized enterprises considering the use of AI have two options.

① AI that can be used immediately (supervised and pre-trained models)

This AI has "pre-trained patterns" specific to each industry and sector. You can start with a certain level of predictive accuracy without having to accumulate your own data from scratch. EMOROCO's top-of-the-line products employ this approach.

② Nurturing AI (self-optimization using proprietary data)

This approach involves accumulating your company's data while the AI ​​learns your customer patterns, resulting in improved prediction accuracy the more you use it. Accumulating quantitative and qualitative data over time with EMOROCO CRM Lite is the preparatory step for this "nurturing AI."

A realistic first step for small and medium-sized enterprises is to **start accumulating high-quality data today that can be used to inform AI decisions**.

When "emotional temperature," "reasons for lost deal (multiple choice)," and "next action deadline" are recorded in the CRM, this accumulated data becomes training data for the AI. A single line of memo you enter today will become the material that determines the accuracy of the customer behavior prediction one year from now.

CRM x AI Utilization Roadmap: 3 Phases for SMEs

Summary – Three reasons why CRM and AI are a good match

Reason ①: CRM is the only system that possesses quantitative, qualitative, and time-series data. Purchase data (quantitative) × employee notes and emotional changes (qualitative) × time-series changes — only CRM satisfies all three conditions for AI prediction. ERP, MA, and BI do not possess all of these.

Reason ②: Qualitative data from CRM can be "encoded → pattern matching." Structured qualitative data such as emotional temperature and ICX change signs can be converted into a "sequence of symbols" that can be processed by algorithms such as xForest and NN. This becomes the material for behavioral prediction using Markov chains.

Reason ③: The more you use it, the more the AI ​​retrains itself to optimize for your company, continuously learning patterns of your customers and improving its prediction accuracy. As an "AI that grows," the data accumulated in the CRM itself becomes a competitive advantage.

EMOROCO CRM Lite is a CRM that prepares you for AI utilization starting from ¥1,500 per user per month. Start today by recording the "emotional temperature" and "reasons for lost deal (multiple choice)" of just one customer. That single data point will become the initial learning material for future AI predictions.
EMOROCO CRM Lite Product Page

Related article: [What is ICX (Implicit Customer Experience)? – How to design unarticulated customer experiences with CRM 4.0 and EMOROCO CRM Lite]

Related article: [Series Part 2: Why Only CRM Can Handle SoI—Evolution from CRM 1.0 to 4.0 and "Deepening Insights"]

Related article: [Is customer data an "asset" or a "liability"? The difference between CRM that simply accumulates data and CRM that actually utilizes it]