How this page is sourced
Every price on this page is third-party: lifted from a named company’s own public page, linked so you can check it, read on 18 August 2026. The only first-party figures are the relative weights I see on real quotes and what model usage costs on the live systems I run, which is money paid to the model provider, not to me.
I don’t publish my own price list. Every project I take on is a fixed price, quoted after a short scoping chat with me. I have not surveyed the market, averaged anything, modelled anything, or written the phrase “we analysed N projects”. If a price is not cited, it is not on this page.
If you search what custom AI costs in 2026 you will be told, on page one, that it is $5,000 and that it is $2 million. Both pages look authoritative. Neither is lying. The problem is that nobody states which market they are pricing, so the buyer — usually a 10 to 500 person company — comes away with a range so wide it is useless for a board conversation.
So here is the tedious version. Published numbers only, each attributed, with the definitions attached. Use it as a reference, quote it, argue with it. Everything is checkable.
1. Why there is no AI Makers price list here
A price list for custom work is a guess dressed up as a commitment. The same “chatbot” can be one API and a knowledge base, or nine systems, a legacy ERP and an audit trail. So I don’t publish one. Every project I take on is a fixed price, quoted after a short scoping chat with me: you know the full number before anything starts, and if my estimate is wrong, that is my problem, not yours.
What I can publish is the part every buyer is trying to reverse-engineer: what moves the number. Then the market’s own published ranges, so you can check any quote — mine included — against something independent.
2. The two variables that move the number most
The type of project tells you roughly what kind of thing you are buying. What decides where you land — and whether you land outside the range you expected — is almost never the model. It is how many other systems the thing has to talk to, and where it has to live. These are the relative weights I see on real quotes.
| Variable | Option | Relative weight | What it buys |
|---|---|---|---|
| Systems involved | 1–3 systems | ×1.0 | One or two APIs, light mapping |
| 4–8 systems | ×1.25 | Several connectors, real data mapping | |
| 9+ systems | ×1.6 | Bespoke connectors, legacy formats, reconciliation logic | |
| Where it runs | Managed cloud | ×1.0 | Hosted, monitored, kept running for you |
| Your server / on-premise | ×1.15 | Deployed inside your infrastructure, full data control |
Put together: a project touching nine systems and deployed on your own servers costs roughly 1.8 times the same project wired to two cloud tools. That is most of the variance in any quote. If an agency will not tell you what moves their number, the reason is usually that it ends in a rate card rather than a commitment.
The reason the systems count dominates is that each integration is a small project: agree the contract, handle the errors, handle the retries, handle the day the vendor changes the API. I wrote up how that plays out in practice in connecting an AI assistant to your business tools.
Scope your custom AI project in 30 seconds
Three questions show the shape of your build and a realistic timeline, based on real shipped projects. After a short scoping chat with me you have a written fixed-price quote, so you know the full cost before anything starts.
Or build your own AI system piece by piece and send the design in for a written quote →
3. What the market publishes
Three sources, each linked, each quoted exactly. I have converted nothing between currencies and rounded nothing.
| Source | Tier as they describe it | Published range |
|---|---|---|
| Kellton | Starter / SMB | $5,000 – $50,000 |
| Simple implementations (chatbots, basic classification) | $50,000 – $150,000 | |
| Mid-complexity (predictive analytics, computer vision) | $150,000 – $500,000 | |
| Enterprise-grade (multiple models, real-time) | $500,000 – $2m+ | |
| EncodeDots | Basic AI chatbot (FAQ / support) | $5,000 – $20,000 |
| Custom LLM integration (RAG on internal data) | $15,000 – $50,000 | |
| AI agent (multi-step task automation) | $20,000 – $75,000 | |
| Predictive analytics / ML model | $30,000 – $100,000 | |
| Enterprise AI platform (multi-agent, multi-department) | $100,000 – $400,000+ | |
| SFAI Labs | Basic (3–6 weeks) | $15,000 – $40,000 |
| Moderate (8–14 weeks) | $40,000 – $120,000 | |
| Complex (14–20 weeks) | $120,000 – $250,000 | |
| Enterprise (20–30+ weeks) | $250,000 – $500,000+ |
Sources: Kellton, EncodeDots, SFAI Labs, all read 18 August 2026.
4. Why the ranges disagree by more than 10x
Set the tables next to each other and the gap looks absurd. Kellton calls a chatbot $50,000 to $150,000. EncodeDots calls a chatbot $5,000 to $20,000. That is a factor of ten on the same noun. Three things explain almost all of it.
The word describes different objects. Kellton’s chatbot tier sits alongside computer vision at $600,000 and recommendation engines at $1.5 million — it is an enterprise deployment with security review, procurement, integration to systems of record and a support organisation behind it. EncodeDots’ chatbot is a support bot on a website. The same word, two different products, both priced correctly.
The buyer decides the price more than the build does. Notice that Kellton also publishes a starter tier at $5,000 to $50,000 for smaller companies. The engineering is not ten times harder for the enterprise version; the process around it is. Requirements sign-off, security questionnaires, vendor onboarding, steering meetings, a project manager and a delivery lead who never write code. If you are a 40-person company you are not buying that, and you should not be paying for it.
Nobody says which currency of risk they are quoting. A range on a marketing page is a starting point for a conversation. A fixed-price quote is a commitment. That is the only kind of number I give: a fixed price, quoted after a short scoping chat, not a range on a web page.
Once you filter for tier, the SMB numbers across independent sources converge: $5,000–$20,000 for a basic chatbot (EncodeDots) and $5,000–$50,000 for Kellton’s starter tier. Two companies with no reason to agree, describing the same slice of the market, land in the same place. That is the number an SMB should plan against, and the rest of the spread is a different industry’s pricing accidentally sharing a keyword.
5. Rate and salary benchmarks
Two ways to sanity-check any quote: what the hours cost, and what the people cost.
| Benchmark | Published figure | Source |
|---|---|---|
| Blended agency rate | $150 – $350 / hr | SFAI Labs |
| Senior ML engineer | $200 – $300 / hr | SFAI Labs |
| AI architect | $275 – $400 / hr | SFAI Labs |
| AI specialist rate | $150 – $300 / hr | Kellton |
| AI/ML engineer, median salary (US) | $189,500 / yr | Stack Overflow 2025 |
| AI/ML engineer, median salary (global) | $89,427 / yr | Stack Overflow 2025 |
| Full-stack developer, median salary (US) | $138,000 / yr | Stack Overflow 2025 |
Salary figures: Stack Overflow Developer Survey 2025. Rates: SFAI Labs and Kellton, as linked above.
The salary line is the one worth sitting with, because it is the honest floor under the whole market. A US AI/ML engineer’s median total cost to an employer is meaningfully above the $189,500 headline once you add employer taxes, equipment and benefits. Hire one and you have committed to that number for a year before a line of code is written. Even the top of EncodeDots’ published agent range, $75,000, is well under half of one year of that person — which is the actual comparison an SMB is making, and I made the long version of it in AI agency vs in-house.
6. What it costs after launch
The build number is the one everyone quotes and the smaller of the two over three years. Here the published sources agree closely, which is unusual enough to be worth trusting.
| Running cost | Published figure | Source |
|---|---|---|
| Annual maintenance, as % of build | 20 – 30% | Kellton |
| Annual maintenance, as % of build | 15 – 25% | SFAI Labs |
| Annual maintenance, as % of build | 15 – 20% | EncodeDots |
| LLM API costs, monthly | $200 – $5,000 | SFAI Labs |
| Cloud infrastructure, monthly | $500 – $5,000 | SFAI Labs |
| Model API usage on a live SMB system (paid to the model provider) | €50 – €500 / mo | Observed on systems I run |
The gap between the €50–€500 I see on SMB systems and SFAI’s $200–$5,000 is a volume gap, not a disagreement: an SMB support assistant handling a few thousand messages a month and an enterprise system serving a million requests are not the same meter reading. Take the percentage figures as the planning number and the monthly figures as a range to place yourself inside. The mistake I see most often is a business budgeting the build and nothing else, then treating year-two maintenance as an unpleasant surprise rather than the 15 to 30% every published source told them to expect.
7. What actually moves your number
Ranked by how much I see each one swing a real quote, largest first.
- Systems involved. Up to ×1.6 on its own, and the effect is the same everywhere. Integration is the work; the model is a line item.
- Blast radius. Software that drafts for a human to approve is cheap. Software that acts on its own needs evaluation, guardrails, audit and a fallback path, and that is a different budget.
- The state of your data. If it lives in one clean system, fine. If it lives in three spreadsheets, a legacy database and someone’s head, the first weeks go on making it usable before anything intelligent happens.
- Regulation. Financial services, healthcare and public sector work carries an audit-and-evidence tail that is real engineering, not paperwork.
- Interface. A chat window is the cheap answer. Something embedded in your own product is design and front-end work that can be a third of the project.
- Where it runs. Around ×1.15 for on-premise. Usually worth it if it is what gets the deal through your own security review.
Two adjacent references if you are costing a specific shape: what a custom AI agent costs for anything that takes actions rather than answers questions, and what an AI employee really costs for the subscription-shaped end of the market, where the sticker price and the total are furthest apart.
How to use this page
- Find your tier, not your keyword. Decide whether you are the 40-person company or the 4,000-person company first. Every published range is only meaningful inside that answer.
- Take the published range for your tier, then place yourself in it. Count the systems honestly — people forget the accounting package and the thing the warehouse uses.
- Add 15 to 30% a year. Every independent source in the table says so. Put it in the business case at the start and the project survives its second year.
- Then get a real number. Every project I take on is a fixed price, quoted after a short scoping chat with me. The estimator gives you a first read in about thirty seconds, or design the system yourself in the builder and send it in for a written fixed-price quote. The spec is yours either way — take it to anyone.
Citing this page
Use anything here. If it is useful in a report, a board paper or an article, the attribution I would like is: Austen, M. (2026) Custom AI Pricing Benchmarks 2026, AI Makers. aimakers.co/blog/custom-ai-pricing-benchmarks-2026/. The price figures belong to the sources named beside them, so link those directly.
Common questions
Why do published custom AI prices vary by more than 10x?
Because the phrase covers two different industries. A published range of $500,000 to $2 million describes a multi-model enterprise programme with a procurement process attached; a published range of $5,000 to $50,000 describes a boutique shop wiring one workflow for a 40-person company. Both are honest about their own market. Read every published range together with the size of company the publisher sells to, and most of the apparent disagreement disappears.
What is a realistic 2026 budget for an SMB custom AI project?
Plan against the SMB tier of the market, not the enterprise one. EncodeDots publishes $5,000–$20,000 for a basic chatbot and $20,000–$75,000 for a task-automation agent, and Kellton publishes a $5,000–$50,000 starter tier for SMBs. Independent sources broadly agree at that end, which is the useful signal in this data. I don’t publish my own price list: every project I take on is a fixed price, quoted after a short scoping chat with me.
How much of the budget is the AI itself?
Less than buyers expect. The model is an API call. What you pay for is integration with the systems you already run, error handling, an evaluation harness so you know the thing still works next month, permissions, and an admin view. On my own projects the number of systems involved is the single biggest driver of the final number.
What should I budget for running costs after launch?
Three published sources put annual maintenance at a similar level: Kellton says 20–30% of initial development cost, SFAI Labs says 15–25%, EncodeDots says 15–20%. On the live SMB systems I run, model API usage, paid to the model provider, is typically €50–€500 a month. Budget the percentage, not the rumour.
Are hourly rates or fixed price better for an AI project?
For a scoped SMB build, fixed price. SFAI Labs publishes blended agency rates of $150–$350 an hour and AI architects at $275–$400. At those rates an estimate that slips by three weeks costs you five figures, and on time-and-materials that risk is entirely yours. I quote a fixed price after a short scoping chat for exactly that reason — the estimating risk should sit with the person doing the estimating.
Related reading
- Custom AI development cost — the interactive estimator
- Custom AI development in 2026: real cost, timeline and how to pick a developer
- What an AI employee really costs in 2026
- Custom AI agent development cost 2026
- How to estimate cost and time for any custom AI project
- AI agency vs in-house: the real comparison