AI consulting has a sales problem, and it is not the one you would guess. The problem is that the pitch usually arrives with no denominator. You hear what the automation will do, rarely what it costs against what it saves, and almost never the break-even point where the engagement pays for itself. So before anyone signs anything, here is the arithmetic, with real numbers.

What AI consulting actually costs in 2026

The market has settled into recognizable tiers:

• Hourly advisory: $150 to $300 an hour in regional markets like Arkansas. Higher on the coasts. Used for assessments, tool selection, and second opinions.
• Project implementations: $2,500 to $15,000 for a defined build: an intake automation, a quoting workflow, a customer-service assistant trained on your documents. Complexity drives the range.
• Ongoing retainers: $1,000 to $5,000 a month for businesses running several automations that need monitoring, updating, and expansion.

Below all of that sits the do-it-yourself tier: off-the-shelf subscriptions at $20 to $100 a month per tool. This tier matters because a fair number of consulting proposals are expensive wrappers around tools you could subscribe to directly. Part of the five-question test below exists to catch exactly that.

The break-even math

The formula is short: hours saved per month, times the loaded cost of the person who was doing the work, against the total cost of the automation.

Worked example. A three-person office spends about 10 hours a week on appointment scheduling, follow-up emails, and retyping information between systems: 40 hours a month of administrative work. At a loaded cost of $25 an hour, that is $1,000 a month in payroll going to tasks with fixed rules.

Suppose a consultant builds automations covering 70 percent of it for a $6,000 project fee plus $150 a month in software. Savings run $700 a month against $150 in carrying cost: $550 net. Break-even lands just before month eleven. Every month after is $550 back, indefinitely, and that is before counting the revenue effect of follow-ups that now happen every time instead of when someone remembers.

Now run the same formula on a bad candidate. A task that takes two hours a month has a theoretical maximum savings of $50 a month at that labor cost. No implementation price makes that math close. The formula is the filter, and it works in both directions.

One asterisk: capacity savings only become cash when the freed hours produce something, more jobs quoted, more calls answered, or reduced overtime. Ten minutes saved here and there evaporates. Whole roles' worth of routine work, reclaimed, compounds.

What to automate first

The best first automations share three traits: high frequency, fixed rules, and low stakes when something odd happens. In practice, the same four candidates come up in almost every small business:

• Lead follow-up. Speed decides outcomes here. An inquiry answered inside five minutes converts at a multiple of one answered tomorrow, and follow-up is the first thing humans drop when busy. This is reliably the highest-return automation a service business can build.
• Scheduling and reminders. Booking, confirmations, and reminder sequences. Cuts no-shows measurably and erases a phone-tag tax every office pays daily.
• Intake and data movement. Information arriving by phone, form, and email, then hand-typed into a CRM or invoice. Fixed rules, high frequency, pure toil.
• First-draft paperwork. Quotes, recurring reports, routine customer answers drawn from your own documents. The machine drafts, a human approves.

Notice what is not on the list: pricing judgment, hiring, anything a customer experiences as the relationship itself. The pattern that works is automating the follow-up, not the handshake. Businesses that hand the relationship to a bot save pennies and pay in churn.

The five-question test before you sign

1. Which specific hours does this save, and how many? If the proposal cannot name the tasks and count the hours, the break-even math cannot exist, and neither can accountability later.
2. What is the total carrying cost? Software subscriptions, API usage, and maintenance continue after the consultant leaves. Get the monthly number in writing.
3. Could a $50-a-month tool do 80 percent of this? Sometimes the honest answer is yes, and a good consultant will say so and charge you for the assessment, not the wrapper.
4. Who fixes it when it breaks? Automations touch real customers. A quoting bot with a bad price list is not a hypothetical. You want a named answer, a response time, and a human-approval step anywhere money or commitments move.
5. Where does our data go? Customer lists and financials flowing through third-party AI tools deserve a plain-English answer about storage, training use, and access before anything launches.

A proposal that survives all five is worth considering. A proposal that dodges question one is marketing.

The bottom line

AI consulting is neither a scam nor a miracle. It is an investment with unusually computable returns, which makes the vendors who refuse to compute them the easiest red flag in the category. Run the formula, start with follow-up, keep the judgment human.

Stone Path Consulting's AI consulting work for Arkansas businesses starts with the assessment, not the build: which hours you are losing, what they cost, and whether the break-even math deserves your signature, in writing, before any tool gets bought. Call 501.232.1017 or email info@stonepathconsulting.com and bring your messiest weekly task to the first call. That one is usually the goldmine.