Agency, Productized Service, or Consulting: Which AI Business Model Fits You

You've had fifteen tabs open for three weeks. One creator says agencies are dead, the next says they're printing money. Someone on LinkedIn swears productized services are the only path, while a podcast guest just made consulting sound like the obvious play. You're not confused about whether AI businesses work — you're confused about which one works for someone with your specific skills, schedule, and bank account.

Here's what you actually need: a framework that evaluates three proven models against four variables that are unique to you. Not opinions. Not someone else's success story. A structured way to stop comparing in the abstract and start measuring against your reality.

By the end of this article, you'll have scored yourself across all three models and identified one frontrunner. That's the goal — not to learn about all three, but to eliminate two.

What These Three Models Actually Are — In Operational Terms

Before you can choose, you need shared definitions. Not marketing definitions — operational ones. What do you actually do on a Tuesday afternoon in each of these businesses?

AI Agency

An AI agency delivers custom AI solutions to clients, with you managing the project and — eventually — a team doing the execution. You sell outcomes ("we'll build you an AI system that cuts your customer support response time by 60%"), then figure out how to deliver them.

Your days look like: sales calls, scoping projects, managing contractors or employees, communicating with clients, troubleshooting delivery issues, and chasing invoices. The AI is one component of what you sell. The rest is project management, client management, and operations.

You are building a services firm. The AI is the specialization, not the business model.

Productized AI Service

A productized service is a fixed-scope, fixed-price offering delivered repeatedly. Instead of custom projects, you do the same thing for every client — "We build AI-powered lead qualification chatbots for real estate brokerages. $3,500. Delivered in 10 business days."

Your days look like: marketing to attract the right clients, running an intake process, executing a repeatable delivery workflow, and refining your system to get faster without sacrificing quality. You trade the high ceilings of custom work for the predictability of doing one thing well.

You are building a production line. The constraint is the product — and that's the point.

AI Consulting

AI consulting means selling your expertise and judgment, not your execution. Clients pay you to assess their operations, identify where AI creates real value, and either recommend solutions or guide their internal teams through implementation.

Your days look like: discovery calls, stakeholder interviews, building assessment frameworks, writing recommendations, and sometimes coaching teams through execution. You rarely build the AI system yourself — you tell people what to build and why.

You are selling your brain. The deliverable is a decision, not a deployment.

The Four Variables That Actually Determine Your Fit

Every piece of advice you've consumed has been generic because it ignores the four variables that make your situation different from everyone else's. Here's what actually matters — and why.

Variable 1: Domain Expertise

This is what you already know about a specific industry or function — not AI knowledge, but knowledge that AI can amplify. A former logistics manager understands supply chain pain points. A marketing director knows which workflows eat the most hours. A healthcare administrator knows compliance constraints.

Why it matters: Domain expertise is the difference between "I can build AI tools" and "I know exactly which problem to solve for exactly which buyer." The second one closes deals. The first one gets you ghosted.

Rate yourself 1–5:

• 1 — No specific industry or functional expertise. General professional background.
• 2 — Some experience in one field, but not deep enough that people seek your opinion.
• 3 — Solid experience in one domain. You could hold your own in a room of practitioners.
• 4 — Deep expertise. People in this field already ask you for advice.
• 5 — You're a recognized expert. You've led teams, solved hard problems, and built a reputation.

Variable 2: Available Hours Per Week

Not the hours you wish you had. The hours you can actually commit, consistently, for the next six months — after your job, your family, and the non-negotiable parts of your life.

Why it matters: Each model has a minimum viable time commitment below which you're just playing business, not building one. Underestimating this is the most common reason people stall out in month two.

Rate yourself 1–5:

• 1 — Under 5 hours per week. Genuinely squeezed.
• 2 — 5–10 hours per week. Evenings only, and not every evening.
• 3 — 10–20 hours per week. Consistent blocks of focused time available.
• 4 — 20–30 hours per week. Part-time job level of commitment.
• 5 — 30+ hours per week. Full-time or close to it.

Variable 3: Financial Runway

How many months of living expenses do you have saved — or how long can you sustain reduced income while building? This isn't about startup capital. It's about how long you can operate before the business needs to feed you.

Why it matters: Some models generate revenue in weeks. Others take months. If your runway doesn't match your model's time-to-revenue, you'll quit before the model has a chance to work — and blame the model instead of the mismatch.

Rate yourself 1–5:

• 1 — Less than 1 month. You need income now.
• 2 — 1–3 months. Tight, but some breathing room.
• 3 — 3–6 months. Enough to build without panic.
• 4 — 6–12 months. Comfortable runway for a deliberate build.
• 5 — 12+ months. Money isn't the constraint.

Variable 4: Risk Tolerance

Not theoretical risk tolerance — your actual comfort with uncertainty, inconsistent income, and the possibility that your first three months produce zero revenue. Be honest. There's no wrong answer, but there is a wrong match.

Why it matters: Agencies have feast-or-famine revenue cycles. Productized services have slow ramp-ups. Consulting can produce early income but requires confidence in selling yourself. Your risk profile determines which revenue pattern you can emotionally survive.

Rate yourself 1–5:

• 1 — Very low. You need predictability and can't handle months of zero income.
• 2 — Low. Some uncertainty is fine, but you want early signals that it's working.
• 3 — Moderate. You can handle a few dry months if the trajectory is clear.
• 4 — High. You're comfortable with significant uncertainty for a bigger upside.
• 5 — Very high. You've done this before and know how to sit with discomfort.

Model Breakdown: The Honest Numbers and Real Requirements

Here's where we get specific. Each model is evaluated with the same structure so you can compare directly.

AI Agency: The Full Picture

What you actually do: Sell custom AI projects, manage delivery (initially yourself, eventually through contractors), handle client relationships, and continuously prospect for new work.

The honest numbers:

• Startup costs: $500–$2,000. Mostly software subscriptions, a basic website, and potentially one paid AI API for prototyping. No significant capital required upfront — but you'll reinvest early revenue into contractors and tools.
• Minimum viable hours: 20–30 per week. Agencies are operationally heavy. Below 20 hours, you can't manage clients and prospect simultaneously.
• Time to first paying client: 4–8 weeks if you have an existing professional network you can mine. 8–16 weeks if you're building from scratch.
• Revenue range, months 1–6: Likely $0 in month one. $2,000–$8,000/month by months 3–4 if you land 1–2 clients. Potential for $10,000–$25,000/month by month 6 — but this is the high end and depends on project size and closing ability.

Who this fits best: Someone with a 4–5 in domain expertise and a 4–5 in available hours. Agencies require both deep credibility (to win projects) and significant time (to deliver them). A 3+ in risk tolerance helps — agency revenue is lumpy, especially early.

Red flags — this model is wrong for you if:

• You have fewer than 15 hours per week. You'll drop balls with clients, and dropped balls destroy agency reputations.
• You have no network in a specific industry. Cold outreach for custom projects is brutally slow without warm introductions.
• You dislike managing people. Agencies that stay solo hit an income ceiling fast. Growth means hiring — and managing — other people.
• You rated yourself a 1–2 on risk tolerance. The gap between projects can be financially and emotionally rough.

The first 90 days — what you'd actually be doing:

Weeks 1–2: Define your niche (industry + AI application). Draft 3 case study concepts based on problems you know exist. Set up basic infrastructure — website, calendar booking, CRM.

Weeks 3–4: Reach out to 20–30 people in your network with a specific, relevant message — not "I started an AI agency" but "I'm solving [specific problem] for [specific type of company], who do you know dealing with this?" Have 5–10 discovery calls.

Weeks 5–8: Close your first project — likely at a discounted rate ($2,000–$5,000) to build a case study. Deliver it while continuing outreach. Document everything about your delivery process.

Weeks 9–12: Use your first case study to pursue full-price projects. Formalize your scoping and pricing. Identify which parts of delivery you can hand off to a contractor. Pipeline should have 3–5 active conversations.

Productized AI Service: The Full Picture

What you actually do: Deliver one specific AI solution, the same way, to every client. Your job is to find the right clients, run them through your process, and deliver a consistent result. You refine the process over time to increase margins.

The honest numbers:

• Startup costs: $300–$1,500. Lower than an agency because you don't need to scope custom projects. Your costs are a website, one or two AI tools, and potentially a template or system you build once.
• Minimum viable hours: 10–15 per week. Once your delivery system is built, each client takes less time than a custom project. The real time investment is in marketing and sales.
• Time to first paying client: 6–12 weeks. Longer than consulting because you need to build the delivery system before you can sell it. But once it exists, each subsequent client is faster to land and serve.
• Revenue range, months 1–6: Likely $0 in months 1–2 while building the service. $1,500–$4,000/month by month 3–4 with 1–2 clients. $4,000–$12,000/month by month 6 if you've nailed your marketing and can handle 3–5 concurrent clients.

Who this fits best: Someone with a 3–4 in domain expertise (enough to identify a repeatable problem), a 2–3 in available hours (the model is designed for efficiency), and a 3+ in financial runway (the build phase before first revenue is real). This is the strongest model for someone with a full-time job who wants to build alongside it.

Red flags — this model is wrong for you if:

• You can't narrow down to one specific deliverable. If "but I could also do X and Y" keeps pulling you sideways, you'll never ship a productized offer.
• You rated yourself a 1 on domain expertise. Without knowing an industry's problems, you'll guess at what to productize — and guess wrong.
• You need revenue in the first 30 days. The build phase is real and unavoidable.
• You get bored doing the same thing repeatedly. Productized services reward consistency, not variety.

The first 90 days — what you'd actually be doing:

Weeks 1–3: Research your target market. Talk to 10–15 people in your chosen niche about their biggest operational pain points. Identify the one problem that's painful enough to pay for, repetitive enough to productize, and solvable with current AI tools.

Weeks 4–6: Build your delivery system. Create the workflow, templates, and AI-powered process that takes a client from intake to deliverable. Test it on 1–2 free or heavily discounted "beta" clients.

Weeks 7–9: Package and price your service. Write your sales page. Create 3–5 pieces of content showing the problem you solve and the result you deliver. Launch to your beta clients' networks first.

Weeks 10–12: Run your first full-price clients through the system. Measure delivery time, client satisfaction, and your actual margins. Adjust pricing or scope based on real data. Begin building a repeatable lead generation process.

AI Consulting: The Full Picture

What you actually do: Sell strategic advice and assessment. Clients pay for your ability to evaluate their business, identify AI opportunities, and recommend (or guide) implementation. You're a trusted advisor, not a builder.

The honest numbers:

• Startup costs: $100–$500. A consulting business has almost no overhead. Your costs are a professional LinkedIn presence, maybe a simple website, and your existing expertise.
• Minimum viable hours: 5–15 per week. Consulting engagements are time-bounded and often part-time compatible. A single client might need 5–10 hours per week for 4–8 weeks.
• Time to first paying client: 2–6 weeks. Fastest of all three models because you're selling what you already know. No system to build first. No team to assemble. Just your expertise, packaged into a clear offer.
• Revenue range, months 1–6: $1,000–$5,000 in month 1–2 from a single engagement. $3,000–$10,000/month by month 3–4 with 2–3 concurrent clients. $8,000–$20,000/month by month 6 if you've established a referral pipeline. Hourly or project rates range from $150–$400/hour depending on your domain and client size.

Who this fits best: Someone with a 4–5 in domain expertise and a 1–3 in available hours. Consulting monetizes what you already know without requiring you to build systems or manage teams. It's the fastest path to revenue and the most compatible with a full-time job — but it has a ceiling unless you raise rates or add leverage.

Red flags — this model is wrong for you if:

• You rated yourself a 1–2 in domain expertise. Consulting without deep expertise is just giving opinions — and nobody pays for opinions from someone who doesn't have the track record.
• You're uncomfortable selling yourself. Consulting requires confidence in pricing your knowledge. If you'd rather hide behind a product or a team, this model will feel painful.
• You want to build something that runs without you. Consulting is you. When you stop, the revenue stops. There's no asset being built beyond your reputation.
• You want a business that scales beyond your personal capacity. Consulting scales only by raising rates or adding junior consultants — both have limits.

The first 90 days — what you'd actually be doing:

Weeks 1–2: Define your consulting offer in concrete terms: "I help [type of company] assess where AI can reduce [specific cost or time] in their [specific function]." Update LinkedIn. Reach out to 15–20 former colleagues or industry contacts.

Weeks 3–4: Have 8–12 conversations. Not sales calls — diagnostic conversations where you demonstrate your expertise by asking better questions than they expected. At least 2–3 of these should surface a potential engagement.

Weeks 5–8: Deliver your first paid engagement. Create a repeatable assessment framework you can use across clients. Document your process and the results you helped produce.

Weeks 9–12: Ask your first client for referrals and a testimonial. Publish 2–3 pieces of content based on insights from your engagement (anonymized). Begin positioning yourself for higher-value or longer-term engagements. Raise your rates by 20% for new clients.

The Comparison at a Glance

Factor	AI Agency	Productized Service	AI Consulting
Startup costs	$500–$2,000	$300–$1,500	$100–$500
Min. weekly hours	20–30	10–15	5–15
Time to first client	4–16 weeks	6–12 weeks	2–6 weeks
Month 6 revenue range	$10,000–$25,000	$4,000–$12,000	$8,000–$20,000
Domain expertise needed	High (4–5)	Moderate (3–4)	Very high (4–5)
Full-time job compatible?	Difficult	Yes — best fit	Yes
Revenue without you?	Eventually, with team	Partially, with systems	No
Scales beyond you?	Yes, with hiring	Yes, with automation	Limited

Self-Assessment Worksheet

This is the part you actually do. Grab a piece of paper, open a notes app, or just fill this in mentally. Score yourself honestly — optimism here only hurts you later.

Step 1: Record your scores from the four variables above.

Here's what a filled-in version looks like for a real situation — a marketing director with 12 years of experience, currently employed full-time:

Domain Expertise: 4 — Deep marketing operations experience, knows the pain points Available Hours: 2 — Can do 8–10 hours per week, evenings and Saturday mornings Financial Runway: 3 — Six months of expenses saved, partner's income covers basics Risk Tolerance: 2 — Needs to see early traction or will lose motivation

Step 2: Calculate your fit score for each model.

Each model weights the four variables differently. Multiply your self-score by the weight, then add them up.

AI Agency Fit Score:

Domain Expertise (your score) × 2 = ___ Available Hours (your score) × 3 = ___ Financial Runway (your score) × 2 = ___ Risk Tolerance (your score) × 2 = ___ Total: ___ out of 45

Productized Service Fit Score:

Domain Expertise (your score) × 2 = ___ Available Hours (your score) × 1 = ___ Financial Runway (your score) × 3 = ___ Risk Tolerance (your score) × 2 = ___ Total: ___ out of 40

AI Consulting Fit Score:

Domain Expertise (your score) × 3 = ___ Available Hours (your score) × 1 = ___ Financial Runway (your score) × 1 = ___ Risk Tolerance (your score) × 1 = ___ Total: ___ out of 30

Step 3: See our example marketing director's scores.

Agency: (4×2) + (2×3) + (3×2) + (2×2) = 8 + 6 + 6 + 4 = 24 out of 45 (53%) Productized: (4×2) + (2×1) + (3×3) + (2×2) = 8 + 2 + 9 + 4 = 23 out of 40 (58%) Consulting: (4×3) + (2×1) + (3×1) + (2×1) = 12 + 2 + 3 + 2 = 19 out of 30 (63%)

Her strongest fit is consulting — which makes sense. Deep expertise, limited hours, moderate runway, and low risk tolerance all point toward a model that monetizes what she already knows, starts generating revenue fast, and doesn't require 25 hours a week she doesn't have.

Step 4: Interpret your results.

Your highest percentage score is your frontrunner. But pay attention to the gap:

• Clear winner (10%+ gap over second place): Move forward with confidence. Your variables strongly favor one model.
• Close race (within 5–10%): Look at your red flags section for each model. The tiebreaker is usually which red flags apply to you — one model might score similarly but have disqualifying constraints.
• Three-way tie: You're either a generalist who could make any model work, or you scored yourself too conservatively. Re-examine your domain expertise score — if it's a 2 or below, focus on building expertise in one area before choosing a model.

The One Thing Most People Get Backwards

Most people choose their AI business model based on which one sounds most exciting. Then they discover their situation doesn't support it — three months in, when they've already invested time and money.

The model doesn't matter if it doesn't match your constraints.

A consulting business built on deep expertise and 8 hours a week will outperform an agency started by someone with 8 hours a week and no network — every single time. Not because consulting is better, but because the fit was better.

The AI is the easy part. The hard part is matching the model to the operator — and that's what this framework is for.

Your Next Step Based on Your Result

You've done the assessment. You have a frontrunner. Here's what to do with it.

If consulting scored highest: Your next move is to write a one-sentence consulting offer using this format: "I help [specific type of company] [specific outcome] using AI-powered [specific approach]." Then reach out to five people in your network who work at or with that type of company. Not to sell — to test whether the problem you've identified resonates.

If productized service scored highest: Your next move is to identify one repeatable problem in your domain that you could solve with a consistent AI-powered process. Talk to five potential buyers and ask: "If I could deliver [specific result] for a fixed price in [specific timeframe], would that be worth a conversation?" Their responses tell you if you've found a real service.

If agency scored highest: Your next move is to map your network. List every professional contact who either runs a business that could use AI or knows someone who does. Your first client almost certainly comes from a warm introduction, not a cold pitch. Start those conversations this week.

In all three cases, the pattern is the same — talk to real people about a real problem before you build anything. The assessment got you to one model. Five conversations will tell you if your instinct is right.

From Framework to Working System

You now know which model fits your situation. That's further than most people get — they stay stuck in the comparison loop for months.

But knowing which model to build and actually building it are different problems. The framework above handles the first one. The second one — building a repeatable, profitable system around your chosen model — is where most solo builders stall. Not because the work is impossible, but because doing it alone means solving every problem for the first time.

That's what NextBuild's cohort program is for. You come in with your model identified, and you leave with a working system — your pipeline, your offer, your delivery process, built alongside operators who've done this before. Not a course. Not content. A build environment.

If you'd rather figure it out alone, the framework above gives you everything you need to start. If you'd rather build alongside someone who's been through it, that option exists too.

Either way — you've chosen your model. Now go talk to five people.