> For the complete documentation index, see [llms.txt](https://help.bizmachine.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.bizmachine.com/en/ai-prioritization/ai-prioritization-tips.md).

# How to write a use case for AI Prioritization

Tips on how to describe your product and your ideal customer so that the AI agent evaluates companies more accurately.

*Updated: 22 October 2026*

The use case is the only input you give the AI agent. It tells the agent what you sell and to whom. The agent finds everything else in the BizMachine data by itself. This is why the quality of your use case decides how useful the results are.

When you run AI Prioritization, you choose a saved use case under **Select Use Case**. You describe the use case itself in the **Value Proposition** field.

### Be specific, and don't be afraid to write more

The most common mistake is a use case that is too general. "We sell software for companies" gives the agent nothing it can use to tell companies apart, so they all get a similar value. Instead, write the way you would explain your product to a new colleague in sales: which problem it solves, which type of company it is for, what a customer must have for the product to make sense, and where you usually do not succeed.

Length is not a problem. The agent can work with a use case of three sentences or of three paragraphs. The more specific conditions you give it, the better it can tell the companies apart.

### The 3 parts of a good use case

#### 1. What you sell and which problem it solves

Describe your product or service. Focus on what makes it different and which problem it solves for your customers.

> *"We offer software for customer support across several channels. It lets you handle customer cases from one place: by phone, web chat and social media."*

#### 2. Who your ideal customer is

State the company size, industry, region and any other conditions your ideal customer should meet.

> *"A mid-sized Czech company in an energy-intensive industry, with several production sites across the country."*

#### 3. Who is not a good candidate

List the industries or types of companies where you know you will not succeed. The agent then gives a lower value to companies where it makes no sense to try.

> *"We don't want the public sector or foreign-owned companies."*

![Run AI Prioritization dialog with a saved use case selected and its menu open with Edit Use Case and Remove Use Case](https://247275268-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYkl7SJ14Nk0vzhVx7upl%2Fuploads%2Fgit-blob-3355c80c1fad880da233d413b4cec308287f30ee%2Fai-prioritization-tips-editace-nabidky.png?alt=media)

### Let an AI assistant write your use case

You don't have to write the use case yourself. If you have connected BizMachine to Claude, ChatGPT, Copilot or another AI assistant through the MCP connector, the assistant can write it for you. It works in your BizMachine workspace, sees your segments, labels and existing use cases, and can save the new use case straight to Prospector.

For example, try:

> *"Write a use case for AI Prioritization in BizMachine. We sell \[what you sell]. Our customers are \[what kind of customers they are]. Use what you know about our company and suggest three parts: 1. what we sell and which problem it solves, 2. who our ideal customer is, 3. who we do not want. Show me the use case before you save it."*

The assistant shows you a draft, you adjust it, and the assistant saves it. A use case saved by the assistant has the visibility **Only me**. If you want to share it with your whole team, switch the **Visibility** in Prospector to the option for everyone in your organization.

You then run the evaluation in Prospector on a segment or a label, in the same way as for a use case you wrote yourself. You can also ask the assistant to summarize the results: "From my segment, pick the 20 companies with the highest value and give the main reason for each one."

To connect the MCP connector, see the [MCP connector](https://help.bizmachine.com/en/mcp-connector/mcp-connector.md) section.

### Example use cases

#### Sensors for heat monitoring

> *"I offer sensors that monitor heat loss in production halls. They are most useful for companies where a stable operating temperature matters. A mid-sized Czech company in an energy-intensive industry with several production sites. I don't want the public sector or foreign-owned companies."*

#### Customer support software

> *"We supply software for customer support across several channels. It works best in companies that have hundreds to thousands of customer cases a month and need routing between several support agents. Typically B2C or an open B2B market. Their products are complex, for example software, electronics, financial or insurance services."*

#### Sales intelligence platform

> *"We offer a sales intelligence platform for B2B sales teams that work in a large, changing market with thousands of potential customers. Typically companies that do not work in fixed supply chains, but need to split the market into segments and set priorities."*

### Step-by-step example: shift planning software

{% hint style="info" %}
This is an illustrative example. The company, the segment and the values are fictional. The steps follow the way AI Prioritization works in Prospector.
{% endhint %}

Imagine a company that sells software for planning shifts in manufacturing. Its segment has 1,200 manufacturing companies in the Czech Republic.

#### First version of the use case

> *"We sell software for shift planning. We are looking for manufacturing companies."*

The first run is on five companies. All of them get a value between 55 and 65. The reasons are general: "manufacturing company, may have shift work". The agent has nothing to tell the companies apart, because the use case says nothing that is not true for the whole segment.

#### Improved use case

> *"We sell software for shift planning in manufacturing with several shifts. It handles overtime, shift swaps between employees and planning for agency workers. Ideal customer: a manufacturing company in the Czech Republic with 150 to 1,500 employees, several sites or non-stop operation, currently advertising positions for production operators or foremen. We don't want: companies with fewer than 150 employees, single-shift operations, state-owned companies, or companies that plan shifts in a large HR system such as SAP."*

The second run is on the same five companies. After you change a use case, the values do not recalculate by themselves, so you need to run the evaluation again. Now the values are spread out: 84, 71, 62, 38, 21. For the company with 84, the reasons say: three sites, 12 job ads for production operators and 2 for shift foremen in the last six months, and "three-shift operation" in the text of the ads. For the company with 21: 90 employees, one site, no job ads for production positions.

#### The whole segment

The use case works, so the company runs the evaluation on all 1,200 companies. When it is finished, the company has a value for every company in the segment. It selects the top 150 companies with the value filter and splits them among its sales reps.

#### What to take from the example

* With a very short and general use case, the values come out similar and the reasons stay general. The agent is not making an error. It is missing the conditions it needs to tell the companies apart.
* Conditions that can be checked in the data help most: size, number of sites, current job ads for specific positions.
* The "who we do not want" part is as important as the description of your ideal customer. Without it, the agent also gives a high value to companies where you know you will not succeed.

### Test on a small sample first

Start with a small sample: under **Options**, choose **First Results**, or **Custom Count** with a number up to 10. Read the reasons for each one and check whether the agent takes into account what matters to you.

If the results do not match your expectations yet:

1. **Make the product description more specific.** Add specific details that set your product apart from its general category.
2. **Make the customer description more precise.** Add the conditions a company must meet (size, industry, technology).
3. **Extend the exclusions.** Add the industries or types of companies where you know you will not succeed.

When you are happy with the results on the small sample, run the prioritization on the whole segment.

{% hint style="info" %}
You can come back to your use case at any time and change it: open the menu next to the use case and choose **Edit Use Case**. A changed use case does not update the values of companies that were already evaluated. You need to run the prioritization again.
{% endhint %}

### Frequently asked questions

#### How often is the value updated?

The value does not update automatically. Run AI Prioritization again when your use case changes, or when you want to work with more recent data. Run it again especially if your prioritization depends on data that changes quickly (job positions, technologies, signals, management changes).

#### Why do all companies have a similar value?

Your use case is probably too general. When the agent does not get specific conditions, it cannot tell the companies apart. Make the product description more specific, make the profile of your ideal customer more precise, and add exclusions.

#### How is AI Prioritization different from filtering?

Filtering selects the segment, and AI Prioritization tells you where to start in it. Filters decide which companies belong in your target segment at all (industry, region, size, financials). AI Prioritization then evaluates each company in that segment by how well it matches a specific use case, and gives it a value and the reasons for it. You run AI Prioritization on a saved segment, on a label or on individual companies, not on the whole database. On a smaller, more precisely defined segment, it is easier to improve your use case.

#### What are the limits?

Each user has their own monthly credit limit. You can see your current credit status directly in the dialog you use to start a prioritization. Your sales representative can tell you how to get a higher limit.

#### Can my colleagues see the AI Prioritization results?

That depends on the visibility of the use case. The default is **Only me**: only you can see the use case and its results. If you want to share the use case with your whole team, switch the **Visibility** to the option for everyone in your organization when you create or edit the use case.

#### Can I export the results?

Yes. Click **Export** above the results table and choose **Export AI prioritization**. For what the file contains, see [AI Prioritization: overview of features and options](https://help.bizmachine.com/en/ai-prioritization/ai-prioritization-overview.md#export-to-excel).

### Related articles

* [AI Prioritization: overview of features and options](https://help.bizmachine.com/en/ai-prioritization/ai-prioritization-overview.md)
* [How credits and limits work in Prospector](https://help.bizmachine.com/en/lists-and-export/credits-and-limits.md)
