# The State of the AI Build Market, Q4 2026

The typical AI build costs $2,500, but that median hides the market. 69% of the $272M posted sits in projects of $5,000 to $50,000, buyers have started naming the model rather than the workflow tool, and the systems a build connects to say more about its budget than the AI inside it.

Updated 2026-11-04. Published by VIZIO AI (https://vizio.ai/?utm_source=build-cost-index&utm_medium=referral&utm_content=pulse-md). Source page: https://aibuildindex.com/pulse. Data: https://aibuildindex.com/pulse/data.json.

## Key findings

- Posts of $5,000 to $50,000 are 32% of priced posts and 69% of posted value. Posts under $500 are 25% of posts and 0.6% of value.
- The largest 10% of posts hold 53% of posted value, and the largest 1% hold 13%.
- Hourly work is 57% of priced posts and about 90% of posted value, on our estimate of hourly posts.
- Claude rose from 1.6% of post titles in January to 9.9% in April, passing OpenAI and ChatGPT in February. It held 9.1% in July, against 3.2% for OpenAI.
- Workflow tools lost share of titles from January to July: n8n from 10.3% to 6.4%, Zapier from 3.8% to 2.1%. “Agent” rose from 9.2% to 13.1%.
- The system a build connects to tracks its budget: a median of $1,000 for posts naming Google Sheets and $5,760 for posts naming Salesforce.
- Posts that ask for evaluation or guardrails have a median of $5,000, twice the market's $2,500.
- Advisory and training posts have the highest median of the five kinds of work we track, $4,320, and the fewest applicants, 8.
- Inside a category, the AI label rarely changes the hourly rate. It changes the rest: 42% of AI posts ask for expert level against 32% of others, and they draw 13 applicants against 9.
- Buyers with $100,000 to $1M of past spend have a median of $5,600, about 4.5 times the $1,250 of buyers with under $1,000.

## 01. Where the money sits

Count posts and this looks like a market of small jobs. Count dollars and it looks like a market of serious projects.

- There are two markets in one feed. Fixed-price posts peak at $100 to $250, usually a small task. Hourly posts peak at $2,500 to $5,000 and describe sustained work.
- The value is concentrated. 53% of it sits in the largest tenth of posts, so the median tells you what is typical, not where budgets go.
- Read the hourly share with care. Hourly totals are our estimate from rate, length and workload, while fixed budgets are what buyers wrote down.

**Priced posts and posted value by budget band**

| Band | Priced posts | Share of posts | Posted value | Share of value | Median applicants |
| --- | --- | --- | --- | --- | --- |
| Under $500 | 9,580 | 24.7% | $2M | 0.6% | 9 |
| $500 to $2k | 8,071 | 20.8% | $8M | 2.9% | 13 |
| $2k to $5k | 8,164 | 21.1% | $28M | 10.4% | 15 |
| $5k to $15k | 7,802 | 20.1% | $67M | 24.7% | 17 |
| $15k to $50k | 4,516 | 11.7% | $120M | 44.0% | 21 |
| $50k and up | 617 | 1.6% | $47M | 17.3% | 21 |

**Priced posts by budget, fixed and hourly**

| From | Fixed | Hourly |
| --- | --- | --- |
| <$100 | 2,211 | 0 |
| $100 | 4,371 | 73 |
| $250 | 2,599 | 326 |
| $500 | 2,605 | 1,088 |
| $1k | 2,664 | 3,307 |
| $2.5k | 1,051 | 5,521 |
| $5k | 645 | 4,730 |
| $10k | 377 | 4,385 |
| $25k | 71 | 2,110 |
| $50k+ | 55 | 562 |

## 02. What is being built

Most demand attaches AI to something that already exists: a product, a process or a customer conversation.

- Products take the most value. AI SaaS and MVPs and custom agents carry more value than their post counts suggest, and over 60% of their posts are hourly.
- Packaged tasks sit lower. n8n workflows, WhatsApp bots and AI receptionists are frequent and small, which is what a commodity looks like.
- The named pages below give the full cost breakdown for each build.

**Use cases by posted value**

| Build | Priced posts | Posted value | Median applicants | Expert requests |
| --- | --- | --- | --- | --- |
| AI SaaS and MVPs | 7,126 | $65.1M | 17 | 45% |
| Back-office automation | 5,056 | $31.6M | 12 | 36% |
| Custom agents | 2,831 | $24.9M | 14 | 45% |
| API integrations | 1,994 | $12.3M | 15 | 41% |
| n8n workflows | 2,396 | $11.8M | 14 | 41% |
| CRM automation | 1,715 | $10.3M | 11 | 42% |
| RAG and document search | 919 | $7.4M | 19 | 51% |
| Support chatbots | 1,344 | $6.5M | 13 | 32% |
| Voice agents | 1,116 | $6.3M | 13 | 45% |
| Make and Zapier | 1,614 | $5.8M | 11 | 35% |
| AI mobile apps | 1,010 | $5.8M | 16 | 45% |
| Vibe-code rescue | 552 | $3.9M | 13 | 39% |
| SEO content | 597 | $3.8M | 9 | 47% |
| Computer vision | 460 | $3.5M | 11 | 51% |
| Social media | 619 | $3.2M | 5 | 34% |
| Email automation | 569 | $2.7M | 10 | 34% |
| Trading bots | 413 | $2.7M | 6 | 44% |
| Text to SQL | 439 | $2.3M | 14.5 | 39% |
| MCP | 291 | $2.3M | 14 | 57% |
| Document extraction | 423 | $1.9M | 15 | 42% |
| WhatsApp bots | 411 | $1.1M | 10 | 34% |
| AI scraping | 276 | $0.9M | 13 | 31% |
| AI receptionists | 222 | $0.7M | 13 | 33% |

## 03. The stack is changing hands

Titles show how buyers think. In 2026 they moved from naming a workflow tool to naming a model and asking for an agent.

- Claude's rise was fast: under 2% of titles in January, near 10% by April. About a third of those posts ask for Claude Code specifically, at a median of $4,000.
- Workflow tools fell at the same time. That may mean less demand, or that automation has become assumed and no longer needs naming. Titles alone cannot separate the two.
- Chatbot is fading as a word while agent grows. Some of this is relabelling: a chatbot in January may be an agent in July.

**Share of post titles naming each term, by month. August onwards is on wider collection.**

| Month | Claude | OpenAI | Gemini | n8n | Zapier | GoHighLevel | Agent | Chatbot | Automation |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Jan | 1.6% | 3.1% | 0.8% | 10.3% | 3.8% | 2.6% | 9.2% | 4.3% | 38.6% |
| Feb | 3.7% | 3.4% | 0.6% | 8.8% | 3.3% | 2.7% | 10.9% | 3.9% | 36.9% |
| Mar | 6.8% | 3.2% | 0.6% | 7.6% | 3.4% | 2.7% | 12.3% | 3.9% | 36.6% |
| Apr | 9.9% | 3.4% | 0.4% | 7.1% | 3.2% | 2.8% | 11.2% | 4.0% | 34.5% |
| May | 9.3% | 2.8% | 0.5% | 6.3% | 2.4% | 3.3% | 11.1% | 3.4% | 34.1% |
| Jun | 8.4% | 3.2% | 0.6% | 6.2% | 2.6% | 3.7% | 12.0% | 3.4% | 33.7% |
| Jul | 9.1% | 3.2% | 0.7% | 6.4% | 2.1% | 4.1% | 13.1% | 2.7% | 33.3% |
| Aug | 6.7% | 2.9% | 0.4% | 5.8% | 1.9% | 5.3% | 13.6% | 2.9% | 33.6% |
| Sep | 5.1% | 2.2% | 0.3% | 5.1% | 1.4% | 6.3% | 15.5% | 2.1% | 32.3% |
| Oct | 5.7% | 2.6% | 0.3% | 4.7% | 1.0% | 5.5% | 17.8% | 2.4% | 29.0% |

**Tools and terms named in titles**

| Tool or term | Share of posts | Median cost |
| --- | --- | --- |
| AI agents | 13.0% | $3,200 |
| n8n | 6.9% | $1,200 |
| Claude | 6.1% | $3,500 |
| GoHighLevel | 4.0% | $1,600 |
| OpenAI and ChatGPT | 3.0% | $1,200 |
| Zapier | 2.5% | $1,200 |
| Make | 2.4% | $900 |
| RAG | 2.4% | $2,100 |
| Lovable and vibe coding | 1.3% | $2,900 |
| Vapi and Retell | 1.1% | $1,050 |
| MCP | 0.7% | $2,600 |
| LangChain | 0.6% | $3,000 |
| Gemini | 0.5% | $1,500 |

## 04. Scope is the real price signal

What a buyer asks for in the description separates a demo from a system. The words track the budget closely.

- Connecting to a system of record costs more than connecting to a spreadsheet. The AI may be the same; the permissions, data and failure costs are not.
- Only 8% of posts mention evaluation or guardrails and 6% mention compliance. These are the posts closest to production, and they sit well above the market median.
- These are associations, not causes. Longer, better-scoped posts mention more of everything, and a buyer cannot raise a project's worth by adding a word.

**Median cost by system named in the post**

| System | Posts | Median cost |
| --- | --- | --- |
| Google Sheets | 3,324 | $1,000 |
| WhatsApp | 2,542 | $1,165 |
| Airtable | 2,179 | $2,000 |
| Twilio | 2,103 | $2,000 |
| GoHighLevel | 3,572 | $2,400 |
| Gmail | 1,412 | $2,440 |
| Shopify | 2,118 | $3,000 |
| Stripe | 2,324 | $3,200 |
| Notion | 1,067 | $3,600 |
| Supabase | 3,086 | $3,600 |
| HubSpot | 1,812 | $4,310 |
| Slack | 2,056 | $4,400 |
| QuickBooks or Xero | 783 | $4,800 |
| AWS | 3,196 | $5,760 |
| Salesforce | 665 | $5,760 |

**Median cost by what the description asks for**

| Asks for | Posts | Share of posts | Median cost |
| --- | --- | --- | --- |
| Evaluation or guardrails | 4,406 | 7.7% | $5,000 |
| Production | 10,620 | 18.6% | $4,720 |
| Security | 4,457 | 7.8% | $4,400 |
| Compliance | 3,315 | 5.8% | $4,080 |
| Ongoing work | 19,658 | 34.4% | $4,000 |
| Past work or portfolio | 12,554 | 22.0% | $3,000 |
| Documentation or handover | 10,851 | 19.0% | $2,500 |
| Self-hosted or private | 1,301 | 2.3% | $2,000 |

**How titles frame the work**

| Framing | Posts | Share | Median cost | Median applicants |
| --- | --- | --- | --- | --- |
| Hire a person | 29,864 | 52.2% | $3,600 | 14 |
| Build a thing | 19,516 | 34.1% | $1,400 | 11 |
| Fix or extend | 3,170 | 5.5% | $880 | 11 |
| Advise or train | 2,955 | 5.2% | $4,320 | 8 |
| Prove it works | 1,710 | 3.0% | $2,180 | 19 |

## 05. Talent: rates flat, competition uneven

AI work does not command a higher rate inside its category. It commands more seniority and draws a bigger crowd.

- The premium is in scope, not rate. Compared within a category, posted rates rarely move, while fixed budgets and expert requests do.
- A bigger budget does not buy a proportionally bigger field: a median $150 post draws 9 applicants and a median $24,000 post draws 21.
- RAG and document search is the most crowded large category, with 19 applicants a post. Crowding says little about quality, so shortlisting is where the work is.

**AI posts against other posts in the same category**

| Category | Fixed: other | Fixed: AI | Top rate: other | Top rate: AI | Expert: other | Expert: AI | Applicants: other | Applicants: AI |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| All categories | $250 | $350 | $35 | $40 | 32% | 42% | 9 | 13 |
| Web development | $300 | $500 | $35 | $35 | 36% | 46% | 12 | 16 |
| Mobile development | $250 | $500 | $30 | $30 | 26% | 37% | 12 | 16 |
| Scripts and utilities | $200 | $250 | $40 | $40 | 27% | 35% | 7 | 11 |
| Ecommerce development | $250 | $300 | $27 | $27 | 30% | 42% | 10 | 9 |
| Data analysis | $200 | $275 | $45 | $50 | 33% | 34% | 8 | 9 |
| Digital marketing | $250 | $215 | $25 | $30 | 32% | 41% | 3 | 5 |

## 06. Who is buying

New accounts make up much of the demand, but experienced buyers bring the larger budgets.

- 46% of posts with a known account date come from accounts opened in 2025 or 2026. Recent accounts may simply be more active, so this is not a growth figure.
- Experience raises budgets. Median cost climbs with every step of past spend, from $1,250 to $6,240.
- Small teams dominate what we can see: 94% of posts that state a size come from teams of up to 10. Large companies rarely buy through public posts.

**Budgets by the buyer's past spend**

| Past spend | Posts | Median cost | Median applicants |
| --- | --- | --- | --- |
| Under $1k | 9,028 | $1,250 | 11 |
| $1k to $10k | 12,411 | $1,800 | 14 |
| $10k to $100k | 9,987 | $3,120 | 18 |
| $100k to $1M | 3,492 | $5,600 | 19 |
| $1M and up | 479 | $6,240 | 21 |

## 07. Where the demand is

One country carries half the money. Elsewhere, budgets and competition vary more than rates do.

- The United States is 45% of posts and 50% of posted value, at a median of $3,360.
- Rates barely move by buyer country, mostly staying between $25 and $30 an hour, which fits a global talent pool. Budgets move far more.
- Posts from India, Pakistan and Nigeria draw about 5 or 6 applicants against 15 for the United States.

**Buyer countries**

| Country | Posts | Median cost | Median rate | Median applicants | Posted value |
| --- | --- | --- | --- | --- | --- |
| United States | 25,638 | $3,360 | $29.5/hr | 15 | $136.3M |
| United Kingdom | 4,401 | $1,800 | $27.5/hr | 14 | $17.3M |
| India | 3,716 | $1,200 | $25/hr | 5 | $14.1M |
| Australia | 2,777 | $2,160 | $26.5/hr | 15 | $12.0M |
| Canada | 2,544 | $2,160 | $25.5/hr | 14 | $11.6M |
| Pakistan | 1,490 | $1,100 | $27.5/hr | 6 | $5.3M |
| United Arab Emirates | 1,174 | $1,800 | $29.5/hr | 15 | $4.8M |
| Germany | 1,106 | $2,600 | $29.5/hr | 15 | $5.7M |
| Nigeria | 912 | $1,000 | $22.5/hr | 5 | $1.6M |
| Netherlands | 845 | $3,200 | $25/hr | 13 | $4.5M |
| France | 682 | $2,000 | $26.5/hr | 11 | $3.1M |
| Israel | 519 | $3,600 | $28.5/hr | 10 | $2.6M |

## What this means for buyers

- **Budget from the system, not the feature.** The same idea costs more against a system of record. Ask what it must read, write and never break, then price from there. Medians run from $1,000 to $5,760.
- **Decide if you want a person or a result.** 52% of posts ask for a person, at a median of $3,600. 34% ask for a thing, at $1,400. The first is a capacity decision and the second is a scope decision.
- **Write the test before the brief.** Posts asking for evaluation have a median of $5,000. If you cannot say how you will know it works, the price you get is for a demo.
- **Do not buy production at prototype prices.** A quarter of posts are under $500, and they hold 0.6% of the value. That money buys a task. A system that runs your business sits in a different band.
- **Plan for shortlisting.** A larger budget brings a slightly larger field, not a much better-sorted one. Ask for evidence of similar work, as 22% of buyers do.
- **Do not judge by hourly rate.** Inside a category, AI work pays about the same rate as other work. Length and seniority explain more of the bill. Small jobs stay small too: the median to fix or extend something is $880.

## What to watch

- **Does Claude's lead hold?** Its title share eased from 9.9% in April to 9.1% in July, still well above OpenAI. We will know more once there are six months of like-for-like data.
- **How much of “agent” is real?** The word now appears in 13.1% of titles, and it also catches human roles. A proper classifier is the next step for this measure.
- **Does advice become a market?** Advisory titles were 3.9% of posts in January and 4.2% in July, then 6.1% in October on wider collection.

## Questions

### How much does an AI build cost in 2026?

The median across 38,732 priced job posts is $2,500. Half fall between $500 and $7,200, and one in ten reaches $19,200 or more. Most of the money sits in projects of $5,000 to $50,000.

### Which AI model do businesses ask for most?

Among models named in post titles, Claude has led since February 2026, with 9.1% of titles in July against 3.2% for OpenAI and ChatGPT and 0.7% for Gemini. This measures what buyers name in public posts, not what is deployed.

### What do AI developers charge per hour?

Posted rates mostly run from $20 to $40 an hour, with a median of $28. Expert-level posts have a median of $36. These are open-market rates; agencies usually charge more.

### Does AI work pay more than other software work?

Not by the hour, once you compare within a category. The difference shows up in scope: 42% of AI posts ask for expert level against 32% of others, and median fixed budgets are $350 against $250.

### What makes an AI project cost more?

Length first, then what it connects to and how close it is to production. Posts naming evaluation or guardrails have a median of $5,000 and posts naming production $4,720, against $2,500 for the market.

### Is demand for AI builds growing?

We do not publish a growth rate yet. Our collection widened in August 2026, so comparing volumes across that date would mostly measure the change in collection. Within January to July, AI posts rose from about 20% to over 24% of posts in the same categories.

## How to read this

Pulse reads public job posts from December 2025 to November 2026. Posted value is what posts offered, not what was paid: fixed budgets as written, plus our estimate for hourly posts from rate, length and workload. Collection widened in August 2026, so trend claims stop at July and later months are shown as a marked band. Terms are matched in post titles and descriptions by keyword, so they measure what buyers wrote, not what was built. Public posts lean towards smaller, open-market work and do not measure enterprise spending.