How to choose an AI consulting firm in St. Louis: 10 questions to ask

Choose the firm that gives specific, written answers about what it would automate first, what the AI can access, what happens when it’s wrong, how the result is measured and what you keep if you stop. Ask every firm the same ten questions below and compare the answers side by side. Vague answers are a signal.

Start with the work, not the tools

Before you call anyone, write down the work you want gone. Not “we should use AI,” but “two people spend Fridays retyping purchase orders” or “we check every overdue account by hand before a reminder goes out.” Add rough numbers if you have them: how often it happens and how long it takes. A clear problem makes every conversation more useful and makes the differences between firms easier to see.

Know which kind of firm you’re talking to

It helps to know where a firm is coming from. You may talk to:

• Strategy firms that help leadership decide where AI fits and what to do first.
• IT providers that keep your systems running and secure, and may add AI tools to that service.
• Agencies that build AI features such as chatbots, content tools or lead-generation systems.
• Automation-focused firms that work on the repetitive steps people do between business systems.

Each can be the right choice for a different problem. If yours is the work between email, spreadsheets and the ERP, what an AI-native consulting firm does differently explains one approach. The ten questions work for all of them.

What should “local” mean in practice?

You don’t need a firm in St. Louis. But a local firm can help with the part of automation that isn’t software: seeing how your company really runs. If a firm says it’s local, here’s what that should mean:

• It meets you at your plant, warehouse or office. Ask whether it will sit with the people who do the work and look at the real inbox, spreadsheet and ERP screens. That’s where you hear about the exceptions nobody wrote down.
• It covers both sides of the river. The metro spans Missouri and Illinois, and the Bureau of Labor Statistics reports it as St. Louis, MO-IL. If you have sites in both states, ask whether the firm will come to each.
• It works your hours. A firm in your time zone can join the Monday order rush or the month-end close without an early call.
• It comes back after launch. Ask who will come in person during the first weeks after an automation is switched on, while your team is still getting used to it.

It should also mean knowing the region’s work. The BLS counts jobs in the St. Louis metro by sector, and these are four of them for August 2026 (preliminary figures):

Ask each firm what it knows about the work in your part of that mix. For a manufacturer, that might be purchase orders and supplier confirmations; for a distributor, freight bills and price updates; for a professional services firm, new client setup.

Local doesn’t replace good answers. A distant firm with clear answers to the ten questions is a better choice than a nearby one with vague ones.

The 10 questions

1. Which workflow would you automate first, and why that one?

A good firm asks about volume, rules and risk before it answers, then names one workflow and explains the choice. Be wary of a plan that starts with everything at once.

2. What exactly will the AI have access to, and can we revoke it?

You should hear a specific list of systems and permissions and a simple way to cut access off. The automation should have its own account or role, not a person’s login.

3. What happens when the AI is wrong?

It will be, sometimes. The answer should cover whether the AI drafts or changes records directly, which cases go to a person, how mistakes get noticed and how quickly an automation can be paused.

4. Who approves each automation before it runs?

Your team should. Ask how approval works in practice, and whether you can let some steps run on their own while others always wait for a person.

5. How will we measure the result?

Expect a baseline before anything starts: how often the work happens, how long it takes and how many errors and exceptions there are. Then the same numbers afterward, counted the same way.

6. Who owns what you build, and what happens if we stop?

Find out what you keep (the automations, the rules, the logs and your data), what you lose and how access is shut off. These answers belong in the contract, not just in a conversation.

7. How do you handle our data?

Ask where your data is processed and stored, how long it’s kept, who can see it and whether it’s used to train AI models. Get the answers in writing before anything is connected.

8. Will you meet our team and watch the work?

The people who do the work know the exceptions that matter. A firm that maps the workflow with them, in person when possible, has a better chance of getting the rules right the first time.

9. What exists today, and what is still in development?

AI products change fast. An honest firm tells you what’s finished, what’s coming and whether anyone has used it yet. Ask to see the parts that exist.

10. How do you price it, and what does the first year cost?

You may be offered a fixed fee per workflow, a monthly subscription, usage-based pricing, hourly work or a mix. Ask what’s included, what costs extra, what happens as volume grows and what the first year costs in total. What drives the cost of AI automation explains what moves the price.

How to compare the answers

Put the ten questions in the rows of a spreadsheet and give each firm a column. Fill it in from what each firm sends in writing, not from your memory of the meeting. Gaps and hedges are easy to spot that way, and so are the firms that answered the question you actually asked.

Red flags

• A plan to automate everything at once.
• No review step, or a claim that the AI doesn’t make mistakes.
• Vague answers about where your data goes.
• No baseline, so no way to know whether it worked.
• A long contract before a single workflow has been proven.
• A demo of features that turn out to be plans.

How Kestrel answers these

Kestrel is an AI-native consulting firm in St. Louis that deploys an AI agent inside the software a company already uses. Here are our answers to the same ten questions, including the ones we can’t fully answer yet:

1. First workflow: the one that costs your team the most time, chosen with you.
2. Access: only the tools you authorize, through a dedicated account or role where the software allows it. You can revoke access at any time.
3. When it’s wrong: the agent drafts, exceptions go to a person, every run is logged and any automation can be paused.
4. Approval: your team’s. Nothing runs until your team switches it on.
5. Measurement: before anything runs, we count the work with you (how often it happens and how long it takes), so there’s a baseline to compare against.
6. Ownership and exit: our contract terms, including what you keep if you stop, aren’t published yet because the agent is in development. Ask us for them in writing before anything is connected. What’s already true is that you can revoke the agent’s access at any time.
7. Data handling: our terms for data handling, retention and model training aren’t published yet either, for the same reason. Ask for them in writing before anything is connected.
8. In person: yes. We’re based in St. Louis and can meet your team in person to agree what the agent may access and which steps stay with people, and to review what it finds.
9. Status: the agent is in development, and early access is open. We have no client results to show yet.
10. Pricing: not published yet while the agent is in development.

See how we work in St. Louis or request early access.