10 questions to ask conversational AI for finance that your dashboard can’t answer

Conversational AI for finance usually means a bank chatbot, which is the assistant in a banking app that answers balance inquiries, flags fraud, and helps customers pay bills. A second kind gets less attention. It sits on top of your own business data, and owners and finance teams use it to ask about revenue, margins, costs, and time.

It's easy to run into problems with that second kind of conversational AI if you treat it like a talking dashboard. Ask for last month's revenue, and you'll get the number your dashboard already shows, but just displayed as a full sentence. Sure, it's correct, but it doesn't actually help you decide anything. You need to understand a comparison, the reason behind the numbers, a cutoff for what's worth reading, and the decision you're making based on the information.

A tool that only repeats totals saves you a few clicks. A good question with an AI that actually knows your business can help you make a pricing decision, create a staffing plan, or inform a renewal conversation.

Owners, finance leads, and service managers who already live in their dashboards will get the most from this article:

• What banks learned from their chatbots
• A step-by-step guide to ask the best dashboard questions
• Ten questions you need to ask about dashboards
• Where human approval fits into this AI workflow

What conversational AI for finance means right now

Conversational AI is software you talk to in plain language instead of clicking through menus, screens, and applying filters to sort data. Natural language processing figures out the user intent behind your words, and in newer tools, generative AI crafts a sensible reply. Artificial intelligence and machine learning have been part of financial services for many years, mostly out of sight and used in fraud detection and fraud prevention models screening transactions in real time, and credit models scoring applications. The conversational interface is the newer part.

Conversational AI in finance shows up in two places: customer engagement tools that face a company's clients, and internal assistants and agents that work on the company's own data.

Conversational banking and the customer experience

Conversational banking is the customer-facing kind of AI. Financial institutions put a finance chatbot or virtual assistant in their apps to handle routine queries like balance inquiries and requests to pay bills, along with real-time notifications such as fraud alerts. The more capable ones connect to core banking systems and CRMs, use transaction history to offer personalized insights, answer in multiple languages, and stick to approved scripts so their answers are as accurate as possible.

For leading banks, their goals in using AI are to create a better customer experience and find operational efficiencies. A virtual assistant can provide instant answers at any hour of the day and handle high volumes of routine questions without adding staff. That keeps support costs down and gives support teams more time for more complex cases, as long as there's a clean handoff to human agents when a conversation gets complicated.

Voice agents can also help trim operational costs on the phone line, and sentiment analysis can flag a frustrated customer. Investment firms build the same kind of artificial intelligence into investment tools that summarize market data and market trends for professional investors.

Understanding and communicating account facts are easy. Personalized financial advice, or help with anything more complicated around business finances, is much harder and needs better AI tools.

An AI assistant for your own numbers

The internal kind of AI is an AI assistant connected to your business systems: billing and payments, accounting, your PSA (professional services automation) tool, a CRM, or inventory. It answers questions specifically about your business and help you make more informed decisions. Which clients grew? Where did the team's time go? Which costs rose faster than prices?

That's the focus here, and it's where conversational AI does the most for decision-making, because the answers come from your clients, agreements, and costs rather than generic advice. For the basics, see our guide to what an AI finance assistant is.

What the finance sector learned about customer expectations

Across the financial industry, banks and card issuers have run customer-facing conversational AI for years, so their results are a useful preview. In May 2026, J.D. Power released four satisfaction studies covering 18,028 bank and credit card customers. Virtual assistants, it found, "handle simple, transactional tasks effectively" but "struggle to support customers through more complex scenarios."

Just 28% of bank and credit card app customers use them, and usage skews toward younger users, especially Gen Z and Millennials. Their customer satisfaction scores run 18 points higher than non-users' on a 1,000-point scale. Still, Jennifer White, managing director of financial services intelligence at J.D. Power, said the assistants "consistently fall short in resolving complex customer service and fraud-related issues."

These findings suggest customer expectations for AI assistants and Financial Intelligence have moved past balance checks, and the assistants at many financial institutions haven't caught up. They're good at what customers could already do in a few clicks themselves, but not so good at solving the business problems business owners really need help with.

The same gap shows up when conversational AI works on a business's own data. The easy questions, like last month's revenue or your biggest client, have answers on a dashboard which are just regurgitated in chat with most AI assistants. The hard questions, like why margin dropped or which clients cost more to serve, are where a conversation with AI assistants still needs to prove themselves. They're also where a tool is most likely to give a confident answer that's missing context.

Just because your AI assistant has a conversational tone doesn't make its vague question precise.

Simply repeating dashboard totals doesn't help

Dashboards are good at monitoring metrics, and someone decided which numbers should matter to you, defined them once, and put them where you can check them in ten seconds. Keep using them for that. The problem is when a metric gets treated as an explanation without context, comparison, or any other insight.

A total metric compresses everything that produced it. Say recurring revenue rose 3% last quarter. That could be steady growth, or a price increase that covered up a lost client. An average project margin of 32% could be eight healthy projects and two that lost money. Product sales up 12% could be one big order at a deep discount.

The dashboard can't tell you which of those happened, or which detail matters for the decision in front of you. Asking a finance AI chatbot to read the same total back is really fast, but that's about it. Real-time data in a chat window is still the same number. The value is in the follow-up conversation and action that you really need from AI.

From a dashboard question to a decision question

Getting more out of conversational AI doesn't mean you need all new questions, but rather upgrading the ones you already ask. Here's one ordinary question:

Start with the dashboard version: "What was our recurring revenue last month?"

Add a baseline. "How did that compare with the same month last year?" A number only means something next to another number. Use last year for a seasonal business, a three-month average when monthly figures bounce around, or your budget if you keep one. Historical data turns a reading into a trend, and a budget tells you whether the trend is good enough.

Ask for the drivers. "Which clients account for most of the change?" This breaks the total into the pieces that moved: clients here, projects in project work, and product lines in product sales.

Set a threshold. "Only show clients whose recurring revenue moved by more than $500 a month." Most of your client list didn't change much, and you don't need to read about it.

Name the decision. "I'm deciding which agreements to review before they renew." This tells the assistant what matters, so it can include details like renewal dates. It also tells you when to stop asking.

Altogether: "Compared with the same month last year, which clients account for most of the change in recurring revenue? Only show the ones that moved more than $500 a month. I'm deciding which agreements to review before they renew."

In short, you don't have to type it all in one go. A good conversational interface remembers context, so you can ask the first part of your question and then narrow it down from there.

10 questions that go beyond dashboard totals

The questions below apply the same upgrades across recurring revenue, project work, and product sales, grouped by the area of the business they cover. Each starts from a dashboard number, then adds the data it needs and a follow-up check.

None of them depend on a particular conversational AI product. Whether a given AI assistant or finance AI chatbot can answer one comes down to which systems it can read.

Pricing and renewals

Long-term relationships with clients have the ability to slowly erode, so ensure you review your client agreement while the relationship scales and the work grows.

Margins

Staffing and capacity

Risk and spending

The concentration question is risk management as much as reporting. Concentration creeps up slowly, and can transfer bargaining power to your largest clients, especially those with renewals coming up.

How to tell if AI answers are actionable

Accurate answers depend on the data underneath them. In a May 2026 poll of about 220 finance professionals at Wolters Kluwer's CCH Tagetik Global inTouch 2026 conference, trusted data was the top prerequisite for adopting autonomous agents, cited by 47%, and 58% said their finance technology is still manual or siloed. "Finance leaders need faster scenario response, stronger data foundations, and AI they can trust," said Atul Dubey, executive vice president and general manager of CCH Tagetik.

Trust in conversational AI gets built one answer at a time, so be sure to ask multiple follow-up questions until you're convinced the answers are good, especially before you act. Here are some options:

"Show me the records behind that." Ask for the invoices, time entries, or transactions that produced the information or numbers. Generative AI will always write confident-sounding sentences around any number, right or wrong, so the sentence proves nothing if the AI tool can't show its work or its source.

"How is that calculated?" Many disagreements between reports come down to definitions. Is revenue invoiced or recognized? Is margin averaged by month or weighted across the period? In Alti, the financial intelligence inside Alternative Payments, dashboard margins and ratios are weighted across the whole date range, and MRR shows the latest month's level. Two tools can both be right and still disagree.

"What's missing?" Ask what was left out: clients with no time entries, products without a category, or a month whose books aren't closed. Then check one record manually in the source system.

Any conversational AI tool for finance should also support security and regulatory compliance. Where is your data stored, and do the vendor's data residency controls meet your regulatory requirements? Does it have a current SOC 2 Type II report? If you handle personal data of people in the EU or UK, will the vendor sign a GDPR data processing agreement?

AI agents, approvals, and the human touch

Everything so far is about asking. AI agents pair generative AI with permission to take actions, like updating a record, creating an invoice, or sending an email. Some routine tasks can safely run on their own, but anything that touches a client or moves money needs approval controls.

Good controls include limits that stop an agent from doing more than the person using it is allowed to do, a person's approval before any action that touches a client or moves money, and a record of who approved what.

In Alti, permissions follow your dashboard access level, so "Alti can only do what you are already allowed to do." Before acting, Alti shows a confirmation card with a plain-language summary of the exact action and a countdown. If time runs out, nothing happens. Each confirmed action is logged with who approved it and what was done.

A client's margin may have dropped because you agreed to a discount during a hard year. The discount is in your system; the reason for it isn't. Alti is a helper, not a replacement for your judgment, and that's a good rule for any AI tool. For more on which tasks can run alone and which need a person, read our article on AI agents in finance.

Financial literacy and AI training for your team

Conversational AI means more people in the business can ask a financial question, not just the finance team. It doesn't teach anyone what a good answer looks like. That's a financial literacy gap, and closing it deserves as much attention as training people on the tool.

Start with vocabulary. If a service manager can't explain the difference between gross and net margin, their follow-ups will go sideways. Create a one-page list of your ten most-used terms, with your definitions.

Then give each role a few starting questions. Owners might begin with client margins and concentration. Whoever runs financial operations might take expense growth and report definitions. Service and project managers might start with where time went and which projects are over budget.

Finally, review questions, not just answers. Once a week, look at one question someone asked, how they checked it, and what they decided. Keep a shared list of the questions that led to decisions.

How Alti turns questions into reports and dashboards

Alti brings this kind of back-and-forth to the data in your Alternative Payments account. You type a question the way you'd ask a teammate, refine the answer, and keep the questions that are worth asking over again.

Ask, refine, and save the question

Alti works from live account data, combining customers, invoices, payments, and emails, and it can't see any other company's information. Answers come back as tables, 100 rows at a time, with a note on how many rows remain. It remembers your conversation, so you can follow up with specific questions like, "now sort just the ones over $5,000" without starting over.

When a question is worth asking every week, save it as a Routine Report. It runs daily, weekly, or monthly and emails the result to you and the teammates you add. Routine Reports only report; they never take actions.

Describe the dashboard you want from Alti

You can build custom dashboards by asking. Describe the view you want, and Alti proposes KPIs before it builds anything. Changing one is just another request: "Remove utilization from my Ops board."

Any Alternative Payments customer can build dashboards from their payments data. With Analytics, you also get KPIs from your PSA and accounting system: ConnectWise Manage, HaloPSA, or Autotask, with QuickBooks Online. Analytics adds views such as Client Trends (average MRR, MRR change, profit margin, and labor cost impact) and Utilization. Only Administrators and Sub-Admins can access Analytics.

Over time, the questions you keep asking become the dashboards you check. If your dashboards tell you what happened and you'd like to ask why, book a demo of Alti to see how it answers questions from live account data and turns the useful ones into reports and dashboards.

Frequently asked questions

How do financial institutions use conversational AI?

Mostly for customer service, an area often called conversational banking. Banks, card issuers, and other financial services firms use a chatbot or conversational banking agent for balance inquiries, transaction history, fraud alerts, and payments, handing harder cases to human agents. J.D. Power's 2026 research found these assistants handle simple tasks well and struggle with complex ones.

Is a finance chatbot the same as an AI assistant?

Not always. In financial services, "finance chatbot" usually means a customer-facing tool that answers questions about an account, though some people use it for internal assistants too. An AI assistant connected to your business systems answers questions about the business from live records. AI agents go a step further by taking actions, which is where approval controls come in.

Can a finance AI chatbot replace my dashboards?

It shouldn't. Dashboards are the fastest way for finance teams to watch numbers they've agreed on, and conversational AI is better for the follow-up questions a total raises. Use both, and turn the questions you ask most often into saved reports or dashboard cards.