12 questions to ask before hiring an AI automation consultant

A managing partner about to sign a $12,000 automation contract usually has one page of notes and a good feeling about the person across the table. That is a thin basis for the decision, because most of these projects do not work. More than 80 percent of AI projects fail, roughly twice the rate of IT projects that do not involve AI, according to a 2024 RAND study on why AI projects fail. The questions you ask before hiring an AI consultant are the cheapest way to find out which side of that number this one will land on.

Quick Answer: Ask an AI consultant what business problem they will fix first, how they will measure it, who owns the finished automations, and what happens when one breaks. The strongest signal is simple: a good consultant asks about your workflows before pitching any tool. The twelve questions below sort operators from tool-sellers.

What questions should you ask before hiring an AI automation consultant?

Ask about outcomes, proof, ownership, process, cost, and support. The point is not to test their vocabulary. It is to find out whether they understand your business before they sell you technology, and whether you will still own the result after they leave. Here are the twelve, with what each answer tells you.

1. What business problem will you fix first, and how will we measure it? Reveals whether they lead with your problem or their tools. A clear metric means they have done this before.
2. Which of my workflows did you look at before proposing anything? Separates discovery-first consultants from ones running the same template on every client.
3. Can you walk me through a system you built for a business like mine? A real practitioner can describe a specific build, what broke in the first month, and how they fixed it.
4. Who owns the automations, the code, the written instructions, and the accounts when we are done? The answer should be you. Anything else is a dependency you will pay for later.
5. What happens when an automation makes a mistake? Every live system fails sometimes. How errors get caught and handled is the real test of whether they build for production.
6. What does this depend on to keep running, and what breaks it? Good consultants name the fragile parts up front: a tool that changes, a login that expires, a form that gets renamed.
7. How do you price this, and what is included? You want to know whether you are buying a fixed scope or an open-ended hourly commitment, and what counts as extra.
8. What do you need from me and my team, and how many hours? Your people's time is a real cost. A vague answer here usually means a stalled project later.
9. How will my data be handled, and who can see it? For a law firm or any regulated business, this is not optional. Ask where the data goes and how access is controlled.
10. What does support look like after go-live? Models drift, tools update, edge cases surface. Ask what happens in month two, not just launch week.
11. What would make you tell me not to automate something? A consultant who will never say no is selling, not advising. The honest ones talk you out of low-value work.
12. What is your plan if the first build does not deliver the result? You want to hear accountability, not a shrug. Ask how they de-risk the work before you commit the full budget.

You do not need to fire all twelve in one call. The first three tell you most of what you need within ten minutes.

Why do the questions you ask matter this much?

Because the odds are against the project, and the reasons are almost never technical. Failed automation traces back to unclear goals, a weak data foundation, and no plan for what happens after launch. Every question above is built to expose one of those failure points before you sign, while it still costs you nothing.

The numbers are worth sitting with. Gartner expected at least 30 percent of generative AI projects to be abandoned after the proof-of-concept stage by the end of 2025, citing poor data quality, unclear business value, and rising cost. In S&P Global Market Intelligence's 2025 survey of more than 1,000 leaders, the share of companies scrapping most of their AI projects before production jumped from 17 percent to 42 percent in a single year. And McKinsey's 2025 State of AI report found that while 88 percent of organizations use AI regularly, only about 6 percent can tie a large share of company profit to it.

Most of that spending went to projects that stalled before they changed anything. Asking these questions up front is how you avoid funding one.

How do you tell a real automation consultant from someone selling tools?

Watch what they lead with. Someone who opens with a list of AI tools before asking what is costing you time is selling technology. Someone who asks how a client goes from first call to signed matter, and where that process stalls, is trying to solve a problem. The second person is rarer and worth more.

Here is the pattern that shows up in almost every good and bad first call:

None of these require you to understand the technology. They only require you to notice whether the person is curious about your business or eager to start building.

Who should own the automations a consultant builds for you?

You should. That means the code, the written instructions, the configurations, the documentation, and the logins to any accounts the system runs on. If a consultant builds everything inside their own tools and keeps the keys, you have not bought an asset. You have rented one, and the rent goes up the moment you want to change anything or work with someone else.

Ask the ownership question directly and listen for specifics. A good answer sounds like: "Everything runs on your accounts, we hand over the documentation and a walkthrough, and you can hire anyone to maintain it." A weak answer talks about their platform, their subscription, and their ongoing role in every small change. That dependency is fine for them and expensive for you.

How much should an AI automation project cost, and how should it be priced?

Expect two separate numbers: one for a scoped diagnostic, and one for the build. The bigger cost driver is rarely the software. It is scoping. Firms that map their workflows first avoid paying to automate a process that should have been deleted instead. A fixed-scope, fixed-price start protects you from open-ended hourly bills.

The cost usually breaks into three parts:

At Eleventh AI, our audit runs $3,500 to $5,000 and produces a ranked build order with hour estimates per opportunity, so you know what implementation involves before you commit to it. Implementation is then scoped by hours against the specific opportunities the audit found. In the law-firm audits we run, each one has turned up ten or more automation opportunities, and the highest-return ones are usually not the ones the partners expected walking in. That is the whole argument for scoping before building. You can read more on what an audit costs and what an AI automation audit includes.

What should you expect for support after the automation goes live?

Launch is the start of the work, not the end. Tools push updates, forms get renamed, a vendor changes a login, and a process that ran clean for two months quietly breaks. Ask what the support arrangement is before you sign, and get it in writing: response time, what is covered, and what counts as new work.

A reliable consultant offers a defined support structure, usually a monthly arrangement or a clear per-fix rate, and can tell you what is most likely to break in your specific setup. If support sounds like an afterthought, assume you will be the one keeping the system alive, and price that into your decision.

Should you start with an audit or jump straight to building?

Start with an audit if you have more than a handful of workflows and no clear ranking of which one to fix first. The RAND research is blunt on this: the projects that fail usually failed at the foundation, with fragmented data and no agreed definition of success, long before anyone wrote a line of automation. An audit is how you catch that while it is still cheap to fix.

Getting the sequence right is not theoretical. When we built a single source-of-truth data layer for a 500-employee New York real estate company, the work that made everything else possible was getting the data foundation right first. Skip that step and you are automating on sand. A short, scoped diagnostic tells you what is worth building, what your data can actually support, and where the first dollar of return will come from. See where to start with AI and automation for how to sequence it.

---

Not sure which of your workflows to automate first? That is the exact question our AI assessment answers. Six questions, about two minutes, and you get a personalized preview of where automation would pay off in your business, before you spend anything on an audit or a build. Start assessment.

Frequently asked questions

What is the single most important question to ask an AI consultant?

Ask what business problem they will fix first and how they will measure it. A consultant who answers with a specific outcome and a metric has done this before. One who answers with a list of tools is selling technology, not solving your problem. This one question filters most of the field.

How much does an AI automation consultant cost?

A scoped audit or diagnostic commonly runs a few thousand dollars, and a build is priced on top based on how many systems connect. At Eleventh AI the audit is $3,500 to $5,000, and implementation is scoped and costed separately against what the report found. Be cautious of open-ended hourly work with no fixed deliverable, which leaves you paying for time without a guaranteed result.

What are the biggest red flags when hiring an AI consultant?

The main red flag is leading with tools instead of your problem. Others include sending a proposal after a single call, promising vague outcomes like 10x efficiency, keeping the build on their own accounts, and never telling you not to automate something. Any consultant who claims guaranteed results before understanding your business is overselling.

Should I hire an AI consultant or an AI automation agency?

It matters less than how they work. A good solo consultant and a good agency both start with your workflows, hand you ownership of the result, and offer support after launch. A weak version of either leads with technology and locks you into their tools. Judge the fit by the answers to the twelve questions, not the label.

Who owns the automations after the project ends?

You should own the code, the written instructions, configurations, documentation, and the accounts the system runs on. If the consultant keeps everything on their platform, you are dependent on them for every change and every fix. Ask the ownership question directly before you sign, and make the handover part of the agreement.

How long does an AI automation project take?

A scoped audit usually takes a few weeks. A first build depends on how many systems connect and how clean your data is, and often runs several weeks to a couple of months. Anyone quoting a firm timeline before reviewing your workflows is guessing, which is itself a useful answer about how they work.

Do I need an audit before hiring someone to build automations?

If you have more than a few workflows and no clear priority, yes. Most failed AI projects fail at the foundation, with scattered data and no agreed measure of success, before any building starts. An audit ranks the opportunities, checks whether your data can support them, and tells you where the first return will come from, which keeps you from automating the wrong thing.