Ask Your Business Anything: 101 Use Cases for Conversational Finance

We asked ourselves a simple question when building Well: what would you ask your business if it could talk back?

That question led us to document 101 specific use cases for conversational finance. Not hypothetical. Real queries from real business operators, categorized by function, ranked by frequency, and built into our product.

This post walks through the most compelling ones, how the AI decides to respond, and why breadth of use cases matters more than depth in any single one.

The question that changes everything

Most finance tools start with a dashboard. You log in, you see charts someone else decided were important, and you try to find the number you actually need.

Conversational finance flips this. You start with your question. The system figures out how to answer it.

This sounds simple until you realize what it requires: the AI needs access to your banking data, invoicing records, CRM contacts, accounting entries, and the relationships between all of them. It needs to understand that "the Paris project" means a specific client, that "last quarter" means Q4 2025, and that "margins" in your business means something different than a textbook definition.

We've been building this infrastructure at Well for months. The 101 use cases are the result of watching what business owners actually ask when the friction of asking disappears.

10 use cases that convert skeptics

From 101 documented queries, these 10 consistently change minds. Not because they're the most sophisticated, but because every business owner immediately recognizes the pain they solve.

1. "Show me overdue invoices." The most requested query by volume. Business owners check this daily across invoicing tools, bank accounts, and mental notes. One sentence replaces a 15-minute routine.

2. "Why did margins drop last month?" This is the query that makes CFOs lean forward. It requires the AI to compare revenue against costs, identify which cost categories increased or which revenue lines declined, and surface the specific drivers. Not a chart. An explanation.

3. "Monthly revenue by client for the last 6 months." A grid. Rows are clients, columns are months, cells are revenue. The AI renders this as a table because that's the right format. Nobody wants a paragraph describing revenue trends when a grid communicates it in seconds.

4. "Which clients haven't paid in over 30 days?" Combines invoicing data with payment records. Surfaces the answer as a list with amounts and aging, sorted by risk. This query alone has saved operators hours of cross-referencing per week.

5. "What's my cash position right now?" Pulls live bank balances, subtracts pending outflows, adds expected inflows from outstanding invoices weighted by payment probability. The answer is a number with context, not a dashboard.

6. "Compare Q4 to Q3." Triggers a side-by-side comparison across revenue, expenses, margins, and client activity. The AI picks a table format with change indicators, because that's what makes the comparison scannable.

7. "Who are my top 5 clients by lifetime value?" Requires aggregation across invoicing history, payment reliability, and contract duration. The output is a ranked list with supporting data points. Business owners use this for relationship prioritization weekly.

8. "Show me recurring expenses over €500." Identifies patterns in transaction data, flags subscriptions and regular vendor payments, and presents them sorted by amount. Turns a finance team's afternoon project into a five-second conversation.

9. "What will my cash flow look like in 90 days?" Projection. Uses historical patterns, known receivables, scheduled payables, and seasonal adjustments. The AI generates a chart here because trend visualization is the right format for forward-looking data.

10. "Flag anything unusual this month." The open-ended query. The AI scans for anomalies: payments that deviate from patterns, new vendors, missing expected invoices, margin shifts. This is where conversational finance stops being a query tool and starts being a monitoring system.

These aren't features. They're sentences. And the gap between typing a sentence and getting an answer is where the entire product lives.

How the AI picks its answer format

Not every question deserves the same response. "What's my cash position?" needs a number. "Monthly revenue by client" needs a grid. "Cash flow projection" needs a chart.

We built three rendering paths into Well's conversational layer:

Text for explanations, summaries, and answers that require narrative. When the AI explains why margins dropped, it writes a few paragraphs connecting causes to effects. A chart would hide the reasoning. Text surfaces it.

Grid for structured comparisons and lists. Client revenue tables, expense breakdowns, aging reports. When the data has rows and columns, the AI renders rows and columns. Forcing this into a paragraph would be unreadable.

Chart for trends, projections, and time-series data. Revenue over 12 months, cash flow forecasts, seasonal patterns. When the shape of data matters more than individual values, the AI draws it.

The AI selects the format based on the query structure and data type. It's not a user setting or a toggle. The system decides, and it's right most of the time. When it's wrong, you can tell it: "Show that as a table instead." It adjusts and remembers your preference.

We also built a pre-built graph library: common visualizations for the most frequent queries, optimized for readability and speed. When you ask for monthly revenue, you don't wait for the AI to generate a chart from scratch. It pulls from a library of tested templates, populates them with your data, and renders instantly.

This matters because speed is trust. If the AI takes 8 seconds to draw a chart, you'll go back to your spreadsheet. If it takes 1 second, you'll ask another question.

Why breadth beats depth

A conversational finance tool that handles 5 use cases brilliantly is a novelty. One that handles 101 is infrastructure.

Here's why breadth matters more than depth for activation:

Discovery happens through variety. A business owner asks one question, gets a useful answer, and immediately thinks of three more. If the system handles all three, they're hooked. If it fails on the second one, they close the tab. The 101 use cases aren't a marketing number. They're a retention strategy.

Different roles ask different questions. The CEO asks about cash position. The accountant asks about reconciliation status. The sales lead asks about client payment history. A narrow product serves one persona. Breadth serves the entire team, and team-wide adoption is what drives enterprise contracts.

Use case diversity signals total addressable market. Investors look at this closely. A product that answers 5 types of questions is a feature inside someone else's platform. A product that answers 101 is a platform itself. The range of queries a system handles correlates directly with the number of businesses it can serve and the depth of value it provides to each one.

Edge cases build trust. When the AI handles a question you didn't expect it to understand, your confidence in the system jumps. Our 101 use cases include niche queries: "What's the tax implication of this expense in France?" and "Show me which vendors have price increases this year." These aren't high-frequency. But when they work, they create the moment where a user goes from "this is interesting" to "I need this."

We don't pretend all 101 use cases work perfectly. Some are better than others. The AI struggles with ambiguous time references, company-specific jargon it hasn't learned yet, and multi-step calculations that require assumptions. We're transparent about these limitations because trust matters more than demo polish.

The full picture

The 101 use cases span 8 categories: cash management, receivables, payables, revenue analysis, expense tracking, client intelligence, forecasting, and compliance. Each category has between 8 and 18 specific queries, from simple lookups to complex projections.

We publish the full list because we think it changes how people think about finance software. When you see 101 natural-language questions that a system can answer, you stop thinking about features and start thinking about conversations. That's the shift.

The traditional finance stack gives you tools. You learn them, configure them, navigate them, and eventually build habits around their limitations. Conversational finance gives you answers. You ask what you need to know, in the words you'd naturally use, and the system figures out the rest.

Not all of it works perfectly today. AI gets time zones wrong sometimes. It can misinterpret ambiguous queries. It occasionally picks the wrong visualization format. These are real problems, and we're fixing them in production every week.

But the core bet is right: business owners have questions, not workflows. Build for the questions, and the workflows become invisible.

What would you ask your business if it could talk back?