What AI Consultants Actually Earn (The Unglamorous Version)

A GTM consultant who posted his full earnings breakdown on Reddit last year wasn't trying to sell anything. No course link. No affiliate code. Just a spreadsheet in prose: roughly $75,000 in revenue over twelve months, eighteen clients, eight of them on monthly retainers. Then this sentence, buried near the end: "After tool costs and taxes, it's a solid income — but not a windfall."

That's probably the most useful thing anyone has written about AI consulting for small businesses in 2025. Not because $75,000 is the number you'll hit. It might be more, it might be less, it might be zero for a long time. It's useful because it includes the ugly parts: a $2,000 project that ate six weeks because scope wasn't defined upfront, and a business that only stabilized when he stopped doing custom work for everyone and started selling one specific outcome to one type of client.

Before deciding whether this makes sense for you personally, you need to understand the structural case for why the demand is real — and why the gap between "demand exists" and "you will profit from it" is wider than most YouTube thumbnails admit.

The Implementation Gap Is Real — and That's Where the Work Lives

The exact AI adoption rate among small businesses is genuinely contested, but the implementation gap those numbers reveal is not, and that gap is where the consulting opportunity lives.

Goldman Sachs surveyed 1,256 small-business owners in early 2026 and found that 76% said they were using AI — but only 14% had fully integrated it into core operations. Meanwhile, NFIB's random email survey of small employers, conducted in March 2025, found just 24% currently using AI tools. Different populations, different definitions — the honest answer is "somewhere between 24% and 76% are experimenting, but almost none have embedded it operationally." The implementation gap is large regardless of which number you trust.

What makes this a business opportunity rather than just an interesting statistic: 73% of Goldman's participating small businesses said they'd benefit from additional training and resources to implement AI successfully. That's not curiosity — that's a stated willingness to seek outside help. RSM's 2025 middle-market survey found the same pattern at scale: 92% of respondents experienced implementation challenges, and 70% admitted needing outside help to get real value from generative AI.

The demand for AI consulting isn't a hype artifact. It's a measurable mismatch between adoption and operational integration, confirmed across multiple surveys with different methodologies.

Whether you come from marketing, HR, finance, customer service, or operations, the underlying problem is the same: businesses are using AI tools the way people use a treadmill — they own it, it's running, and they're not sure they're doing it right.

Knowing demand exists is necessary but not sufficient. The harder question is what operators who pursued this actually earn — and what the business looks like six months in, when the novelty has worn off.

What the Earnings Evidence Actually Shows

There is no defensible median income for AI consultants serving SMBs — but the range of credible public cases reveals what drives the difference between "solid income" and "minimum-wage math on a six-figure invoice."

The Reddit consultant's $75,000 year broke down to roughly $4,200 average project, 60% of revenue from retainers, and an honest admission that the early flat-fee model produced minimum-wage effective rates on some projects. A 12-hour build and a 40-hour build cost the client the same $2,500, which meant some projects were fine and others consumed full attention for substandard pay.

A separate published case offers a more modest but instructive comparison: a bookkeeping-firm owner who sold a document-intake workflow to accounting peers — a problem he'd already solved inside his own three-person firm — reached 10 clients at $650 per month average by month four, roughly $93,000 annualized. He never hired a developer. The product was narrow, reusable, and built from domain knowledge rather than AI novelty.

Self-reported revenue claims on platforms like Medium reach $23,000–$30,000 per month, with one author claiming 70%-plus margins. These numbers are not independently audited, do not disclose the full economic cost of founder labor, and should be treated as upper-tail possibilities rather than forecasts.

What changed for me was niching down hard: instead of 'AI automation,' I now specialize in automating one specific thing for one specific industry. That let me build reusable components and standardize pricing.
— jdrolls, AI Agency Founder

Before you run the math on what you could earn, run the math on what you'd actually collect: total hours including sales, discovery, rework, and support — divided into cash received after software, taxes, and contractors. That number is your real hourly rate, and it tells you whether this is a business or an expensive hobby.

This calculus applies whether you're a marketer building lead-qualification systems, an HR professional automating onboarding workflows, or an ops person streamlining document processing. The tool changes; the unit-economics test is identical.

Why Projects Fail (And It's Almost Never the AI)

Most AI consulting engagements fail not because the technology doesn't work, but because the people-and-process problems were never addressed — which is exactly what makes domain expertise more valuable than tool fluency.

BCG's 2024 survey of 1,000 senior executives across 59 countries found that 74% of companies had not yet shown tangible value from AI. Approximately 70% of implementation challenges were people-and-process related — compared with just 10% attributable to the AI algorithms themselves. Gartner forecast that at least 30% of GenAI projects would be abandoned after proof of concept by year-end 2025, most often due to poor data quality, unclear business value, or inadequate risk controls. Not bad models. Bad implementation conditions.

Here's where the Reddit consultant's scoping disaster becomes structurally important. That six-week $2,000 project wasn't a technical failure. The automation worked. The problem was that scope wasn't written down: the client kept adding adjacent requests, the consultant kept saying yes, and neither party had defined what "done" meant. That's a process failure. It's also why he now sends a scope document before touching anything — a behavioral change, not a technical one.

The consultant who wins is not the one who knows the most about language models. It's the one who can diagnose whether a client's data is clean enough to automate, whether there's a human review step for high-stakes decisions, and whether anyone in the organization actually owns the workflow being changed. These are operations and change-management skills — the kind built inside an industry over years, not from a six-week AI course.

A former HR manager who understands hiring workflows will catch data-quality problems that a generalist developer will miss. A marketing operations specialist who knows HubSpot cold will scope a lead-qualification system accurately where a newcomer will overbuild and underdeliver.

Understanding the failure modes tells you what to avoid. But it doesn't tell you whether you're positioned for the version of this business that's durable — or the one that's already being commoditized. That distinction is where the decision actually lives.

Two Paths: Which One Are You On?

There are two versions of this business: one that's durable because it's built on domain expertise and measurable outcomes, and one that's being commoditized in real time because it's built on generic tool access. Which path you're on is usually determinable in four questions.

Path A is the Workflow Specialist. The bookkeeping-firm owner illustrates this cleanly: he identified that document intake was costing his own firm hours every week, automated it, then sold the same productized workflow to 40 other bookkeeping firms. He didn't know AI — he knew bookkeeping. The AI was the tool; the domain knowledge was the product. This path has a defensible niche, reusable components, and clients who can't easily replicate the domain diagnosis.

Path B is the Generic Automator. One Reddit analysis of AI automation specialists found that 70% of those who had worked at or owned an agency saw that agency close or pivot to selling courses and education. The most common pattern: sold general automation services, competed on technical novelty, faced price compression as tools improved and clients built internal capability. This path has a shrinking price floor and a client who is one platform update away from doing it themselves.

Running an agency — even an automation agency — becomes a human resources + client management job very fast.
— oyodeo, AI Automation Agency Founder

Here are the four questions to determine your path. Can you name a specific industry where you understand the most painful manual workflows from the inside — not from reading about them? Can you reach at least 10 decision-makers in that industry through existing relationships or targeted outreach, without cold-pitching strangers? Can you define a baseline measure — hours saved, leads converted, errors caught — before you build anything, and commit to reporting it afterward? Are you willing to do the maintenance: monitoring, updates, support calls, and client education, not just the initial build?

Four "yes" answers: Path A is available to you. Fewer than three: you're starting from a position that looks more like Path B, which means longer time to a first paying client and more exposure to commoditization.

The four questions apply identically whether you're coming from marketing, finance, HR, customer service, or operations. The variable is the industry knowledge — not the AI knowledge.

What to Do This Week

The Reddit consultant said the business only stabilized when he stopped doing custom work for everyone and started selling one specific outcome to one specific type of client. That's the entire playbook compressed into one sentence.

AI consulting for SMBs is not an AI business. It's an operations improvement business that uses AI as a tool. If that reframe excites you — if the operations problem sounds more interesting than the AI novelty — you're probably in the right place. If it deflates you, that's useful information too.

This week: identify one workflow in an industry you understand from the inside. Find three businesses where that workflow is currently manual and costly. Offer to map it for a small fixed fee — $500 to $1,500, bounded to two weeks, with a written deliverable — before building anything. That diagnostic is your market test, your first paid engagement, and your proof-of-concept simultaneously.

Build nothing until someone pays for the diagnosis.

The median income is unknown. The implementation gap is real. The only way to find out which side of it you land on is to sell the diagnosis before you write a single line of code.

Recommended Tools & Resources