How to hire an AI chatbot developer in 2026: freelance, agency or in-house

Here's the short version. If you need one chatbot built on top of an existing model (GPT, Claude or Gemini answering questions from your own documents), hire a freelance AI chatbot developer with production RAG experience; if you need several integrations, compliance paperwork and someone accountable for uptime, hire an agency; and only hire in-house once the chatbot is a product you'll keep changing every month. Rates run from about $10 to $350 an hour depending on where the developer lives and how senior they are, so the hiring model you pick moves your budget more than almost anything else.

I've put this guide together from published rate data, salary ranges and agency cost breakdowns, all checked in early November 2026. Nobody at the site hired a developer for this piece, and the vendor sources below sell hiring services themselves, so I've flagged that where it matters.

Key takeaways

A freelance chatbot developer in the US typically charges $119 to $185 an hour at mid-level, going by Second Talent's 2026 benchmarks.
Agencies cost roughly twice as much per hour as freelancers, but they take on project management, QA and handover.
An in-house conversational AI developer in the US earns around $110,000 to $160,000 a year before benefits.
The skills that separate good candidates from weak ones are retrieval (RAG), evaluation and integration work, not prompt writing on its own.
Budget for running costs too: a RAG knowledge-base bot costs roughly $300 to $1,200 a month to run once it's live.

The work an AI chatbot developer does

An AI chatbot developer in 2026 rarely trains a model. They wire a hosted large language model to your data and your systems, then make it behave. Upwork's guide to chatbot careers describes the conversational AI developer's job as building LLM-powered agents, "implementing RAG (Retrieval-Augmented Generation) systems to allow bots to access custom data", fine-tuning open-source models and writing the prompts that steer behaviour.

In practice, a typical build has four layers. There's the model API, a retrieval layer that searches your help centre or PDFs and feeds the right passages into the prompt, the integrations (CRM, order lookup, ticketing), and the front end, which might be a website widget, WhatsApp or Slack. Most of the hours go into the middle two.

That's also why the job title is a bit slippery. Upwork splits the work into conversational AI developers, chatbot UX designers, AI integration specialists and machine learning engineers, and a small project probably needs one person who covers the first three reasonably well.

If you don't need custom code at all, a no-code builder may do the job for a fraction of the price; our list of the best AI chatbot builders covers those.

Freelance vs agency vs in-house at a glance

The table below uses US hourly figures from Second Talent's cost-to-hire page (updated 24 September 2026) for AI agent developers, which is the closest published category to LLM chatbot work. Second Talent places offshore developers, so read the agency column with that in mind.

Model	Junior (1-3 yrs)	Mid-level (3-6 yrs)	Senior (7-10 yrs)	Best for
Freelance or contract	$95-$142/hr	$119-$185/hr	$155-$235/hr	One well-scoped bot, prototypes
In-house employee	$105-$155/hr	$130-$205/hr	$170-$260/hr	Chatbots that are core to the product
Agency or dev shop	$180-$295/hr	$230-$380/hr	$295-$490/hr	Multi-system builds, regulated industries

Second Talent doesn't break down what's inside the in-house figure. I'd guess it includes some employer overhead, since it sits above the freelance band, and that's a fair reminder that a salary number on its own understates what an employee costs.

Hiring a freelance chatbot developer

Freelancing is where the price spread is widest. goLance's chatbot developer rate guide puts the worldwide market at $40 to $350 an hour, with mid-level developers averaging about $93 and seniors about $160. On its own marketplace, though, most chatbot developers charge $10 to $30 an hour, and it says about 83% charge under $40.

That gap between $20 and $160 is the whole story of freelance hiring. Location explains part of it: Second Talent lists mid-level rates of $26 to $41 an hour in South Asia, $70 to $109 in Eastern Europe and $74 to $115 in the UK. Experience explains the rest, and goLance says a developer who has shipped to millions of users charges two to three times more than one with only notebook experience.

Vetted networks sit at the top end. Arc's chatbot developer listings show profiles describing multi-agent RAG platforms built with LangGraph and pgvector, and one developer's profile gives $150 an hour as his usual rate. Toptal's chatbot page shows similar senior profiles, including one for the author of the NLP.js framework, whose profile says he's deployed more than 20 chatbots and call centre virtual agents.

Weak spots of freelance hiring

One person is a single point of failure. If they get sick, take another contract or simply stop replying, your bot stops improving and nobody else understands the retrieval setup. I'd insist on a written handover (architecture notes, prompt versions, the evaluation set) as a paid milestone, not a favour.

Scope creep is the other failure mode. goLance suggests hourly billing when scope may change and fixed price when deliverables are clear, which sounds obvious but matters more for chatbots than for most software, because "make the answers better" has no natural end.

Hiring an AI chatbot development agency

An agency gives you a team: usually a lead engineer, a project manager and QA. RaftLabs, which sells this service, says on its cost page that its lean pod (senior engineer, part-time PM, QA) starts at $12,000 to $15,000 a month, with mid-complexity projects running $36,000 to $90,000 over three to six months.

Project totals vary a lot between agencies. RaftLabs lists $8,000 to $25,000 for an API-based chatbot and $20,000 to $50,000 for a RAG knowledge base. Monocubed, another agency, has a chatbot cost guide with much higher bands: $75,000 to $300,000 for a generative LLM chatbot and $150,000 to $500,000 or more for an enterprise agentic system, and it says vendor quotes "vary by 5x to 10x for what looks like the same project". Honestly, that line is the most useful thing on either page.

Agencies probably make the most sense for an enterprise AI chatbot development service for websites that has to plug into five systems and pass a security review. Monocubed says the upper end of its bands assumes multi-channel deployment, five or more integrations and HIPAA, SOC 2 or PCI-DSS controls. If that's you, our roundup of AI chatbot development services compares specific firms; this guide sticks to how to hire, not whom.

Weak spots of agency hiring

You pay for overhead you may not need. A two-week FAQ bot doesn't need a project manager, and at $230 an hour or more the meetings add up fast. I'd also ask who'll actually write the code (the senior engineer on the sales call isn't always the one on the build) and get that name in the contract.

RaftLabs makes another point I agree with: data preparation is 20% to 40% of total project cost and it's the line item most often missing from low quotes. A suspiciously cheap agency quote has likely left it out.

Hiring an in-house AI chatbot developer

Upwork lists average US salaries of $110,000 to $160,000 or more for conversational AI developers, $100,000 to $150,000 for AI integration specialists and $120,000 to $170,000 or more for machine learning engineers, citing ZipRecruiter, Indeed and Glassdoor. Our AI developer salary explainer goes further into pay by role.

An employee makes sense when the chatbot is the product, or close to it. Models change every few months, retrieval quality drifts as your documents change, and someone has to watch the logs. A full-time hire who knows your data will catch problems an outside contractor won't see. Demand is high, too: Upwork says its 2026 In-Demand Skills report found demand for AI chatbot development up 71% since 2024, so expect to compete for good candidates.

The downside's simple. You're paying all year for work that may be front-loaded into three months. For a single customer support bot, I think a full-time hire is usually overkill.

A worked example: one RAG support bot, three ways

Say you want a support chatbot that answers from 400 help-centre articles, looks up order status in Shopify and hands off to a human in your ticketing tool. Assume about 300 hours of engineering, which fits inside RaftLabs' 8 to 14 week timeline for a RAG knowledge base. Using Second Talent's mid-level US rates:

Option	Rate used	300-hour build	Notes
US freelancer	$119-$185/hr	$35,700-$55,500	You manage the project yourself
South Asia freelancer	$26-$41/hr	$7,800-$12,300	Time zones and vetting are on you
US agency	$230-$380/hr	$69,000-$114,000	Includes PM and QA overhead
In-house hire	$110k-$160k salary	Full year's pay	Only worth it if work continues

Then add running costs. RaftLabs puts a RAG knowledge base at $300 to $1,200 a month after launch, so over a year you're looking at roughly $3,600 to $14,400 on top of the build. Monocubed says build cost is typically 60% to 70% of first-year spend. Those two sources don't line up perfectly, and I'd plan for the higher share of running costs, since the cheaper build options leave more of the tuning to later.

The offshore line looks like a bargain, and it can be. It's also where vetting matters most, because a bad hire at $30 an hour that you replace after six weeks costs more than a good one at $90.

Skills to look for in a chatbot developer

Most candidates will list the same frameworks. Upwork's list for conversational AI roles is Python, NLP, LangChain, LlamaIndex and prompt engineering. Those are the floor. What you want to know is whether they've shipped something people used and fixed it when it broke.

Retrieval design: chunking documents, choosing embeddings, using a vector store such as pgvector or Pinecone, and knowing when keyword search beats vectors.
Evaluation: a written test set of real questions with expected answers, run after every prompt or model change.
Integration work: API calls into your CRM, order system or ticketing tool, with sensible error handling when those calls fail.
Cost control: tracking tokens per conversation and choosing a smaller model where it's good enough.
Guardrails: refusing out-of-scope questions, handing off to a human and not inventing refund policies.
Data handling: experience with GDPR, HIPAA or SOC 2 if you're in a regulated field (goLance says this adds 20% to 40% to rates).

Evaluation is the one I'd weight highest. A developer who can't tell you how they'll measure answer quality before writing code is probably going to tune by feel, and you'll be the one who finds the bad answers.

Interview questions that sort candidates

Skip the trivia. Ask about things that only come up in production.

Walk me through a chatbot you shipped. How many users or conversations did it handle, and what broke first?
How would you split our help-centre articles for retrieval, and how would you know if the split was wrong?
What goes in your evaluation set, and how many questions would you start with?
Our bot gives a confident wrong answer about refunds. What do you check, in order?
Roughly what would 10,000 conversations a month cost us in model fees, and how would you cut that?
What do I get at handover, and could another developer pick it up without you?

On the cost question, RaftLabs gives a rough benchmark: 10,000 conversations at about 500 tokens each is 5 million tokens, which it put at $25 to $75 a month. Model prices move often, so don't hold anyone to that number; you're checking whether they can do the arithmetic at all.

Choosing the right hiring model

Start freelance if the scope fits on one page

A single bot, one or two integrations and a clear list of questions it must answer is freelance territory. Pay for a short discovery phase first (a week or two) and make the evaluation set a deliverable. Fixed-price milestones work well here: one for the retrieval prototype, one for integrations and one for launch plus two weeks of fixes. If the developer won't agree to a written list of questions the bot must pass, that's a warning sign.

Pick an agency when compliance or integrations dominate

Five integrations, a security questionnaire and a launch date your CEO has announced? That's agency work. Ask for the named engineer, the handover format and whether data preparation is in the quote. For context on the finished products, see our list of enterprise AI chatbots.

Hire in-house once the bot changes monthly

If you're shipping new flows every few weeks, or the chatbot sits at the centre of your product, the knowledge stays with a staff engineer. Many teams start with a contractor and convert the role later, which seems to me the lowest-risk path.

Consider building it yourself first

A weekend no-code prototype shows you what customers really ask. Our guide on how to build AI agents is a reasonable starting point.

FAQ

How much does it cost to hire an AI chatbot developer?

Freelancers charge from about $10 to $350 an hour worldwide, according to goLance, with mid-level US contractors at $119 to $185 an hour in Second Talent's data. Agency projects range from about $8,000 for a simple API bot to $500,000 or more for enterprise systems, depending on integrations and compliance needs.

What's the difference between a chatbot developer and a chatbot builder?

A chatbot builder is a software product with templates and a visual editor that you configure yourself. A chatbot developer is a person who writes custom code: retrieval pipelines, API integrations and evaluation tests. You'll probably want a developer once a builder can't reach your internal systems or meet your data rules. Plenty of teams use both: a builder for the widget and simple flows, plus a developer for the custom retrieval or the one integration the builder doesn't support. Our AI chatbots for business list covers the off-the-shelf side.

How long does it take to build a custom AI chatbot?

RaftLabs estimates 6 to 10 weeks for an API-based chatbot and 8 to 14 weeks for a RAG knowledge base. Monocubed's estimates are longer, up to nine months for enterprise agentic systems.

Should I choose the best AI chatbot development company or a cheaper freelancer?

It depends on scope, not reputation. A well-defined bot with one or two integrations suits a vetted freelancer. A multi-system build with compliance requirements suits an enterprise AI chatbot development company that can supply project management, QA and a named team. Compare quotes line by line, especially data preparation.

Do I need a machine learning engineer to build a chatbot?

Usually not. Most business chatbots use a hosted model through an API, so the work is retrieval, integration and testing rather than training. You'd need a machine learning engineer if you plan to fine-tune or self-host an open-source model, which RaftLabs says makes sense mainly at very high volumes.

Final verdict

For most small and mid-sized companies, I'd hire a vetted freelance AI chatbot developer for the first version, pay them to write an evaluation set and a handover document, and decide on agency or in-house help only after the bot has a few months of real conversations behind it. The rate data makes agencies look expensive, and for a single support bot they probably are. But if you're facing a compliance review and five integrations, the extra $30,000 to $60,000 is likely cheaper than a failed launch. Whoever you hire, the question that tells you most is how they'll measure whether the answers are right.

How to hire an AI chatbot developer in 2026: freelance, agency or in‑house