---
title: "AI agent pricing: how much to charge for custom agents"
canonical: "https://agenticup.dev/posts/ai-agent-pricing-how-much-to-charge-custom-agents/"
pubDate: "2026-06-01T00:00:00.000Z"
description: "Pricing is the hardest part of selling AI agent services. Here's exactly how I price — the numbers, the reasoning, and the mistakes that led to this model."
tags: [pricing, business, freelancing, ai agents, consulting, rates]
---

Gartner's [AI services forecast](https://www.gartner.com/en/newsroom/press-releases) indicates that fixed-price AI services projects have 40% higher client satisfaction than time-and-materials engagements, supporting the fixed-price model advocated in this post.

McKinsey's [State of AI](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai) shows that organisations are spending $50K-$500K on AI agent proof-of-concepts, providing a pricing reference for custom agent development services.

**TL;DR:** Pricing is the hardest part of selling AI agent services. This post breaks down my ₹1,20,000 ($1,440) fixed-price model — how I arrived at it, the 3-box scoping framework, and the 5 common mistakes that cost solo developers money.

I spent my first three months as an AI developer underpricing every project. I was converting dollars to rupees in my head, feeling guilty about charging "so much" for "just code," and consistently leaving money on the table.

Then I started talking to other developers about pricing. I realized something: most of us make up prices based on vibes. We don't have a model. We don't have a justification. We just pick a number that feels less embarrassing than the last one.

This post is my attempt to formalize what I've learned. I'll share my exact pricing model, how I arrived at it, and what I'd change.

> **Key takeaways:**
> - Fixed-price ₹1,20,000 ($1,440) for a vertical agent delivered in 14 days is the sweet spot
> - Use the 3-box scoping framework (Must Have / Nice to Have / Future) to protect your scope
> - Hourly billing penalizes efficiency — charge for value delivered, not time spent
> - Raise prices when your pipeline is consistently full, not when you feel confident enough

## The $1,400 fixed-price model

My current offering: **₹1,20,000 ($1,400) for a vertical AI agent delivered in 14 days.**

This isn't arbitrary. Here's how I arrived at this number.

### The cost side

14 days of work on a single project. Let's say 6 hours per day of focused work (generous, but sustainable). That's 84 hours of billable time.

₹1,20,000 / 84 hours = ₹1,428/hour ($17/hour)

That looks low, but:

- Not all my time is project work. I spend time on sales, marketing, learning, and operations. The 84 hours is the portion I dedicate to delivery.
- I work on 1-2 projects simultaneously. Effective hourly rate across all projects: ₹1,20,000 × 1.5 projects / 84 hours = ~₹2,100/hour ($25/hour).
- Projects are concentrated. The 14-day timeline means I spend 2 weeks fully focused, then 2 weeks lighter (sales, learning, maintenance) before the next project.

The hourly rate isn't the point. Fixed-price removes the friction of time tracking and lets me focus on delivering value.

### The value side

$1,400 is less than a US-based developer would charge for the same work. A US developer at $150/hour × 84 hours = $12,600.

$1,400 is more than an Indian freelance marketplace developer would charge. Upwork developers for "AI integration" often quote $15–$30/hour.

The price sits in a sweet spot: cheap enough that a US or European business doesn't think twice, expensive enough that Indian clients take it seriously.

### The market side

I tested prices from ₹40,000 to ₹2,00,000. Here's what happened:

- **₹40,000 ($480):** Too many inquiries, mostly unserious. Clients treated me as cheap labor. Projects had poor scope boundaries. I was overworked and underpaid.
- **₹80,000 ($960):** Better clients but still some price haggling. Projects were profitable but the margin was thin.
- **₹1,20,000 ($1,440):** Sweet spot. Serious inquiries only. Clients value the work. Good margin. Reasonable volume.
- **₹1,60,000 ($1,920):** Fewer inquiries but higher quality. Projects were very profitable but the pipeline was inconsistent.
- **₹2,00,000 ($2,400):** Almost no inquiries. Price exceeded what most solo clients would pay a solo developer without a proven track record.

₹1,20,000 is where the market response is strongest while maintaining healthy margins.

## What's included and what's not

The ₹1,20,000 covers:

- **Discovery call (45 min):** Understand the workflow and define the agent's scope
- **Agent development (8–10 days):** Build the agent with proper error handling and monitoring
- **Deployment (2–3 days):** Deploy to Railway, Fly.io, or Cloudflare Workers
- **Documentation:** What the agent does, how to use it, how to maintain it
- **Handoff call (30 min):** Walk through the system and answer questions
- **30 days of support:** Bug fixes and minor adjustments post-delivery

It does NOT cover:

- **Ongoing API costs:** Client pays for LLM API usage directly
- **Hosting infrastructure:** Client pays for server or platform costs
- **Extended maintenance:** Beyond 30 days, ₹15,000/month for monitoring and bug fixes
- **Significant scope changes:** If the project shifts fundamentally, we re-scope and re-price
- **Training data preparation:** If the agent needs domain-specific data, that's separate

## How to scope an AI agent project

The hardest pricing skill is scoping — defining what the project includes clearly enough that you don't lose money.

**The one-sentence test:** If you can't describe the agent's purpose in one sentence, you haven't scoped it well enough.

> Bad: "Build an AI agent to help with customer support."
> Good: "Build an AI agent that reads incoming support tickets, categorizes them (billing/technical/account), drafts a response, and triages to the right team member."

The well-scoped version tells you:
- What goes in (support tickets)
- What processing happens (categorization, drafting)
- What comes out (categorized tickets with draft responses)
- Who uses it (support team)

With this scope, you can estimate effort accurately. Without it, you're guessing.

### The 3-box scoping framework

I use this framework in every discovery call:

| Box 1: Must Have | Box 2: Nice to Have | Box 3: Future |
|------------------|-------------------|---------------|
| Agent runs on new tickets | Agent auto-assigns priority | Agent resolves tickets autonomously |
| Three category labels | Custom label support | ML-based dynamic categories |
| Draft response in email format | Draft in multiple formats | Integrated with Zendesk API |

Box 1 is the scope for ₹1,20,000. Box 2 is discussed as a potential addition. Box 3 is a conversation about future phases.

This framework serves two purposes: it establishes clear boundaries (so you don't build more than you priced), and it gives the client a sense of progression (they see what they could get next).

## Fixed-price vs hourly vs value-based

I've experimented with all three models. Here's my honest assessment:

**Hourly billing ($90/hour):** Works for consulting and advisory work where scope is genuinely unclear. Terrible for agent development because it penalizes efficiency. The faster you build, the less you earn. This creates perverse incentives.

**Fixed-price ($1,400/project):** Ideal for well-scoped agent projects. The client knows what they'll pay. I know what I'll earn. The incentive is aligned: build fast, build well, move to the next project.

**Value-based ($5,000–$10,000 based on ROI):** The best model economically but the hardest to sell. It requires quantifying the agent's ROI before building it. For mature businesses with clear metrics (cost per ticket handled, hours saved per week), value-based pricing works. For most clients, it's too abstract.

My recommendation: start with fixed-price. Graduate to value-based when you have case studies that prove your agents deliver measurable ROI.

## Pricing for different project types

Not every project fits the $1,400 template. Here's how I price variations:

| Project Type | Price (INR) | Price (USD) | Duration |
|-------------|-------------|-------------|----------|
| Simple automation agent (one workflow, one tool) | ₹80,000 | $960 | 7–10 days |
| Standard vertical agent (medium complexity) | ₹1,20,000 | $1,440 | 14 days |
| Complex multi-agent system (2–3 agents) | ₹2,40,000–₹3,60,000 | $2,880–$4,320 | 21–30 days |
| RAG pipeline + agent | ₹1,60,000 | $1,920 | 14–18 days |
| AI consulting (per hour) | ₹7,500 | $90 | Varies |
| Monthly maintenance retainer | ₹15,000 | $180 | Ongoing |

The prices aren't fixed — they're starting points. I adjust based on the client's budget, the project's complexity, and my current workload.

## How to raise prices

The progression I recommend:

1. **First 3 projects:** ₹60,000–₹80,000 per project. Goal: build portfolio, get case studies, develop process.
2. **Next 5 projects:** ₹1,20,000. Goal: establish market rate, refine scoping, get testimonials.
3. **After 10 projects:** ₹1,60,000–₹2,00,000. Goal: fewer but better projects, higher margins, more time for product work.

Each price increase happens when your pipeline consistently fills at the current price. If you're turning down work or your calendar is full 3 weeks out, raise prices.

## Common pricing mistakes

I made all of these. You can skip some.

**1. Charging by the hour.** Hourly billing caps your income at your available hours. AI development rewards efficiency — you want to be rewarded for building fast, not punished.

**2. Discounting for "experience."** New developers underprice because they lack confidence. Clients don't know what you lack confidence in. They evaluate your price against the value they expect, not your experience level.

**3. Not raising prices after successful projects.** After a good delivery, raise your rate for the next client. You've proven you can deliver. Charge accordingly.

**4. Including API costs in the price.** Your ₹1,20,000 quote should be for development only. API costs are a pass-through: the client pays them directly or reimburses you. If you include them, a complex agent run eats your margin.

**5. Over-customizing the scope.** Every client wants features that are "small changes." They're never small. Protect your scope with the 3-box framework.

---

*Related: [How to make money as a solo AI developer](/posts/how-to-make-money-ai-solo-developer/) — practical strategies for building a sustainable income, and [How to start an AI development business from India](/posts/how-to-start-ai-development-business-india/) — everything about finding clients, pricing, and delivery.*

*Related: [AI agent business models: how to build a sustainable agency](/posts/ai-agent-business-models/) — how different business models affect pricing strategies for custom AI agent development.*

## The honest truth about AI agent pricing

The market for AI agent development is still forming. Prices are all over the place. Some developers charge $500 per project. Some charge $20,000. Both are getting clients.

Your price communicates something. $500 says "I don't know what this is worth." $20,000 says "I solve serious problems." ₹1,20,000 says "I'm a professional who delivers real value at a fair price."

Choose what you want to communicate. Price accordingly.

And remember: pricing is not the thing clients care about most. They care about trust — can you deliver what you promise? The developer who communicates clearly, sets boundaries confidently, and delivers reliably will always win over the cheaper option with less clarity.