The short answer: hire a conversational-AI specialist for the storefront, a platform for the help desk, and a product engineer for anything visual

If you sell online and need a chatbot that answers product questions, tracks orders and handles returns, the strongest fits on this list are Master of Code Global and BotsCrew, both US-based with named retail work. For a first bot under $50,000, LeewayHertz publishes a $10,000 minimum on Clutch and 10Clouds reports 120-plus AI deployments. If the bot has to live inside a 3D product page or configurator, the Culver City studio at position five builds both halves. Choose carefully: Gartner predicts more than 40% of agentic AI projects will be cancelled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls (Gartner, press release, June 2025).

How this list was built: the criteria, and what was deliberately ignored

Every company here was checked against its own website and, where one exists, its Clutch profile on September 20, 2026. Nothing comes from memory, sales decks or other people's rankings. When a company does not publish a figure, the comparison table says "not published" instead of an estimate.

Four filters decided who made the cut. First, the company builds or supplies conversational AI for customer-facing commerce: product discovery, pre-sale questions, order status, returns and exchanges, post-purchase support. Second, it names retail, marketplace or consumer-brand work on its own site, or publishes a minimum project size that a mid-sized store can meet. Third, it describes systems in production rather than demos; a chatbot that only works in a sales call is not a chatbot. Fourth, the list mixes regions and company sizes, because a US retailer will realistically compare a domestic agency against a Polish or Indian one on price, and a services firm against a platform.

Two entries are products, not agencies: Botpress, a support-agent platform, and Syte, a visual product-discovery engine. They are here because a large share of ecommerce chatbot projects end up as "buy a platform, hire a builder", and a buyer who reads only agency lists never learns what the platform half costs.

The order is by fit to an ecommerce chatbot project, not by headcount or revenue. A 6,000-person firm can build a returns bot, but it is rarely the right call for a $40,000 first project, so size counted against a company as often as for it. Awards, "top developer" badges, hourly rates in isolation and social-media followings were ignored.

Why the bias toward production work: Gartner predicted at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025 (Gartner, press release, July 2024), and in McKinsey's State of AI 2026 survey of 1,719 respondents, 40% of large organizations reported scaling AI agents while smaller organizations stayed flat at 22% (McKinsey / QuantumBlack, 2026). The gap between a pilot and a bot that runs your returns queue at 2 a.m. is the whole job.

Comparison table: ten vendors by focus, published budget, timeline and region

Budget and timeline columns show only figures the company publishes on its own site or Clutch listing. "Not published" means exactly that; it says nothing about how expensive or slow a vendor is. Ask each one for a written estimate against the same scope, and use the questions in the section after the company entries.

The ten companies, one at a time: what they build, what they have shipped, and when to look elsewhere

1. Master of Code Global: conversational AI for retail and beauty, Redwood City

Master of Code Global calls itself a consulting-led AI engineering partner: it assesses, builds and deploys conversational and generative AI, plus the custom software around it. Founded in 2004 and now 150-plus people, it is headquartered in Redwood City, California, with teams in Winnipeg, Poland and Kyiv. The site names Burberry, Estée Lauder, La Mer and Tom Ford Beauty alongside smaller ecommerce brands such as BloomsyBox and Zipify, which is the closest thing on this list to a pure retail chatbot portfolio.

The stack is unusually wide for a services firm: Claude and OpenAI models, Rasa, Parloa, Salesforce, AWS and Amazon Connect, Google Cloud, Apple Messages for Business, RCS and SMS, described as "15-plus conversational AI platforms". That matters when a beauty brand wants the same assistant in web chat, iMessage and a contact-center queue.

Pick them when the bot is customer-facing, brand-sensitive and multichannel, and when you want a partner to choose the platform rather than sell you one. Do not pick them if you need a published price before the first call, or if your project is a single-channel FAQ bot that a platform license and two weeks of configuration would cover. With teams in four countries, also ask which office would staff your project and in which time zone your daily contact sits.

2. BotsCrew: bespoke AI agents and campaign chatbots, San Francisco

BotsCrew, founded in 2016, builds custom AI agents, generative AI and conversational AI for enterprises and startups, with retail, healthcare, HR and marketing named as focus verticals. Headquarters are in San Francisco, with US offices in Austin, San Diego, Cincinnati and Minneapolis and an engineering base in Lviv, Ukraine. The stack listed is GPT-5, Llama 3, RAG, NLP and AI agents.

The representative projects on its homepage are Honda's HR-V launch campaign chatbot, work with Samsung NEXT and a chatbot for the FIBA Basketball World Cup. These are high-traffic, short-window, consumer-facing deployments, which is a useful signal for a retailer planning a product launch or a peak-season assistant that must not fall over on Black Friday.

Pick BotsCrew when you need a consumer-grade conversational experience with a US contract and engineering hours priced from Eastern Europe. Do not pick them if you need order-management depth documented in a case study; the showcase work is campaign and engagement bots, and neither team size nor pricing is published, so budget fit has to be established on a call. If your traffic is seasonal, ask how the launch-campaign bots were load-tested; the work suggests they have an answer, but the site does not document it.

3. 10Clouds: agentic commerce and AI assistants from Warsaw

10Clouds has been operating for 17 years and describes its AI practice as custom AI agents and assistants, workflow orchestration and "agentic commerce", with credit automation on the fintech side. It is an Anthropic (Claude) partner, runs its own AIConsole tooling and reports 120-plus AI deployments. The about page names Displate, a print marketplace, alongside PZU, an insurer, and Trust Stamp, an identity-verification company.

The ecommerce relevance is the marketplace work and the explicit agentic-commerce positioning: agents that do not just answer but act, placing reorders or resolving a return within rules you set. Pricing and team size are not published. Poland sits in the $50–99 hourly band most commonly reported on Clutch, the same band as the United States, so the case for Warsaw is talent and process rather than a discount (Clutch software development pricing guide, data updated September 2026).

Pick 10Clouds when you want an assistant that takes actions in your order system and a partner already working with a frontier-model vendor. Do not pick them if you require a US office or a portfolio weighted toward retail; the named cases lean fintech and insurance. One procurement detail: the legal entity on the about page is 10Clouds Financial Institutions sp. z o.o., so confirm which entity signs your contract and under which law.

4. LeewayHertz: a $10,000 entry point and its own ZBrain platform

LeewayHertz, founded in 2007 and now part of The Hackett Group, is an AI consulting and development firm with 50–249 staff, based in Gurugram, India according to its Clutch profile. It builds custom AI platforms, generative AI solutions, AI agents and enterprise software, and runs ZBrain, its own orchestration platform. The published stack is deliberately model-agnostic: GPT-5.2, Claude, Gemini, Llama 4, Grok 3 and Mistral, with crewAI and AutoGen Studio for multi-agent work, which matters if you want a bot that is not welded to one model vendor's price list. The site names Scrut (an LLM-powered compliance app), app work for O'Reilly Auto Parts and an app enhancement for Siemens.

The number that puts it at position four is the published minimum: $10,000 per Clutch, at $50–99 per hour. That is the lowest entry point among the agencies here, and it makes LeewayHertz the practical answer for a store that wants a scoped first bot, say order-status and returns-policy questions, before committing to a wider program.

Pick LeewayHertz for a defined first project with a hard budget ceiling, or when you already know you want an orchestration layer rather than a single bot. Do not pick them if you need daytime overlap with a US team as a hard requirement; the site does not publish an office list, and the Clutch HQ is India. Also ask whether a proposal steers you onto ZBrain and what leaving it would cost.

5. A Culver City studio that puts the assistant inside the product page

Most chatbots on this list sit in a corner widget. When the assistant has to work inside a visual product experience, a configurator, a 3D catalog or a WebAR preview, Todor3D builds both halves under one roof. Founded in 2020, the studio has 40-plus engineers across three continents and 300-plus delivered projects, with 25 reviews and a 5.0 rating on Clutch. Its three practice areas are 3D and Immersive, Custom Software Engineering and AI Solutions, on a stack of WebGL, Three.js, React Three Fiber and WebAR/WebXR, with a presence in Culver City, California.

The studio publishes its brackets: $10,000–50,000, $50,000–100,000 and $100,000–250,000-plus, with timelines of 4–10 weeks, 3–6 months and 4–12 months. The homepage lists a closet configurator, in the $10,000–50,000 bracket and delivered in four months, and a jewelry configurator in the same bracket. Neither is described as a chatbot project, so treat them as evidence of the visual side, not the conversational side.

Pick the studio when the bot must answer against a configurator's rules, guide a shopper through options on a 3D model, or hand a rep a quote with the configured product attached. Do not pick them for a text-only support bot with no visual surface; AI is one of three practices rather than the whole business, and founded in 2020, the track record is shorter than the 2004 and 2007 vintages on this list. Ask which engineers would staff the conversational half and for a live example of it, since the two configurators on the homepage show the 3D side only.

6. Markovate: generative and agentic AI with a $50,000 minimum, San Francisco

Markovate, founded in 2015, is a San Francisco firm with a 50-plus core team and four locations, building custom generative and agentic AI: chatbots, machine learning, computer vision and MLOps on Azure, Google Cloud and AWS. Its named industries are manufacturing, healthcare, insurance, construction and real estate, and its proprietary tools, CADIAM and an AI Blueprint Classifier, are aimed at CAD and quotation automation. Client references include MPP Innovation, CodmanAI, Standard Textile, Civil Takeoff and LegalAlly.

Its Clutch profile lists a $50,000 minimum project size at $50–99 per hour. That is the highest published floor here, and it tells you what Markovate is for: a bot that is really a workflow, such as a B2B storefront where the assistant reads a spec sheet, checks configuration rules and produces a quote. A textile or building-products seller with a catalog full of engineering constraints will find more relevant experience here than at a retail-first shop.

Pick Markovate when your ecommerce is B2B, technical and quote-driven, and when a $50,000 floor fits. Do not pick them for a consumer fashion or beauty assistant; none of the named work is in that space, and the minimum will price out a small first pilot. The four locations are not named on the site either, so ask where the team on your account would sit.

7. SoluLab: 250-plus developers across four countries

SoluLab positions itself as an AI-native product development partner for enterprises and startups, covering enterprise AI platforms, generative AI products, chatbots and custom software, alongside a large blockchain and Web3 practice on Ethereum, Solana, Polygon, Hyperledger and Hedera. The company cites 11-plus years of experience and 250-plus developers, with offices in the United States, Canada, India and Australia and delivery across North America, Europe, the Middle East and APAC. It does not name a headquarters city.

The case studies on its about page are Aman Bank, Ambetter Insurance and Morpheus.Network, a supply-chain platform. None of these is a retailer, so the relevance to a store is capacity and breadth: a team big enough to run a chatbot build in parallel with, say, a loyalty program on a token ledger, and to staff around the clock across time zones.

Pick SoluLab when the chatbot is one workstream inside a larger platform build and you want one vendor for all of it. Do not pick them if you want a partner whose published cases are in commerce, or if you need pricing before a scoping call; neither is on the site. A blockchain practice that large also raises a fair question: what share of the 250 developers work on AI rather than Web3, and has that team shipped a commerce integration before.

8. Quantiphi: conversational AI at enterprise scale, Massachusetts

Quantiphi is an AI-first digital engineering firm founded in 2013, with 1,000–9,999 staff per Clutch and US offices in Marlborough, Massachusetts, Princeton, New Jersey and San Jose, California. It sells conversational, generative and agentic AI, document AI, cloud modernization and data analytics, and packages some of that into products: baioniq, Codeaira, Dociphi and Qollective.cx, the last a customer-experience platform.

What it does not publish is as telling as what it does. There is no minimum project size on Clutch, no retail case study on the homepage, and the testimonials name Sullivan County and Illinois Tech without project detail. This is a firm built for enterprise engagements where the chatbot is one piece of a contact-center or data program, with the compliance and procurement paperwork that implies.

Pick Quantiphi when you are a large multi-brand retailer consolidating support, search and analytics under one AI program, and when procurement expects a vendor of this size. Do not pick them for a single-store assistant or a fixed-price pilot; the engagement model and the absence of published commerce work both argue for a smaller shop. If you do engage, ask whether the proposal builds on Qollective.cx or baioniq and what leaving those products would cost later.

9. Botpress: the platform option, priced at $9,000 a year

Botpress is a product, not a services firm: an enterprise AI agent platform for customer support that handles multi-step workflows such as refunds, account updates and ticket deflection. Its homepage lists American Express, Kia, Polestar and Converse as customers and, according to Botpress's own data, reports a 66% AI resolution rate at Super Dispatch and a 30% reduction in calls at Extendly. Neither the LLM vendors it uses nor its headquarters are named on the pages checked.

The reason it earns a place on an agency list is the published price: $9,000 per year for four seats and 1,200 conversations per month, with no per-seat cost, per the pricing shown on the site on September 20, 2026. That is a number you can put next to every agency quote above. If a vendor proposes a $60,000 custom build, the honest question is what it does that a $9,000 platform plus a few weeks of configuration would not.

Pick Botpress when your use case is support, your volume fits the plan, and you have someone in-house or a small agency to configure it. Do not pick it, on its own, when the bot must drive a visual product experience or take actions deep in a custom order system; that is where the builders on this list come back in. Because neither headquarters nor the underlying model vendors are named on the pages checked, settle where conversation data is stored and which model answers your shoppers before you sign a data-processing agreement.

10. Syte: visual product discovery to sit behind the chatbot

Syte, shown on its site as "Syte by Nayax", is a visual AI product-discovery engine for ecommerce: image and text search, recommendations and automated product tagging with deep tags, thematic tags and tag analytics. It is here because the weakest part of most retail chatbots is not the conversation but the retrieval; a bot that cannot find "the blue linen shirt like this one" fails on the first real question.

The results on its homepage are vendor-reported: a 40% average-order-value uplift at Decathlon and a 7.1-times conversion rate at Coleman, with Prada and Farfetch named as customers. Pricing, headquarters, founding year and team size are not published, so this is a demo-and-quote conversation.

Pick Syte when your catalog is large and visual, apparel, home or sporting goods, and your chatbot vendor needs a product-discovery layer they will not have to build. Do not pick it as a substitute for a chatbot developer; it does not design conversations, connect order systems or handle returns, and none of the published figures come from an independent study. The footer reads "Syte by Nayax", so as with Botpress, establish who signs the contract and where your catalog images and shopper queries are processed.

How to choose for your own store: six questions to ask a vendor before you sign

Every vendor above will say yes to "can you build an ecommerce chatbot". The questions below are the ones that produce different answers, and the differences are what you are paying for.

1. Which of your live bots can I talk to right now, and what does it do when it fails? Ask for a URL, not a video. Type in a question the bot should not be able to answer, such as a return outside the policy window, and watch whether it hands off to a human, invents an answer or loops. Gartner's finding that only about 130 of the thousands of vendors calling themselves "agentic AI" are genuinely agentic (Gartner, June 2025) applies to chatbot shops too: a demo that only works on a script is a demo.

2. What does the bot cost to run per month, at my volume, and who pays the model bill? Development is a one-off; inference is forever. Ask the vendor to price it against published list rates. On September 20, 2026, OpenAI's gpt-5.6-luna listed at $0.20 per million input tokens and $1.20 per million output, Anthropic's Claude Haiku 4.5 at $1 and $5, and Google's Gemini 3.8 Flash at $0.75 and $3.75 with a scheduled increase to $1.50 and $7.50 on January 1, 2027 (OpenAI, Anthropic and Google developer pricing pages, checked September 2026). Output tokens cost four to six times input tokens across all three vendors, and cached input is roughly ten times cheaper than fresh input, so a vendor who designs for short answers and prompt caching is saving you money every month.

Add the retrieval layer: Pinecone's Starter tier is free up to 2 GB of storage, Builder is $20 per month flat and Standard starts at a $50 monthly minimum (Pinecone pricing page, checked September 2026). About 20% of respondents in McKinsey's 2026 survey said AI-related operating costs had constrained their AI use (McKinsey State of AI, 2026); a vendor who cannot show you a monthly cost model before you sign is asking you to join that group.

3. Which systems does the bot read from and write to, and which of those integrations have you built before? A useful commerce bot reads the product catalog, inventory, order status and returns policy, and writes at least one thing: a ticket, a return authorization or a reorder. Ask for the list of platforms and order-management systems the vendor has integrated in production, and how the bot behaves when one of them is down. "We can integrate with anything" is a true statement about APIs and a useless one about risk.

4. How will the bot handle sizing, fit and product-detail questions my product pages do not answer? Baymard's product-page benchmark found 83% of apparel sites provide insufficient sizing information and 25% of ecommerce sites lack product images with enough resolution or zoom (Baymard Institute, product page benchmark, 2025–2026). Shoppers ask the bot what the page did not tell them. A vendor should explain where the bot's answers will come from, a size chart, a PIM, a human, and what it will say when it does not know. This is also where a visual layer earns its place: a bot that can show the product at the angle the shopper asked about answers a question text cannot.

5. What is the escalation design, and how much of the work do you expect a human to keep? Gartner predicts that by 2030, 75% of B2B buyers will prefer sales experiences that prioritize human interaction over AI (Gartner, August 2025), while its 2025 survey of 646 B2B buyers found 67% prefer a rep-free buying experience and 45% used AI during a recent purchase (Gartner, March 2026 release). Both are true: shoppers want self-service until it fails, then they want a person fast. Gartner also reports that buyers who use a supplier's digital tools together with a rep are 1.8 times more likely to complete a high-quality deal, and that purely self-service purchases are far more likely to end in regret (Gartner, B2B buying journey research, checked September 2026). Ask the vendor to show the handoff, including what context the human receives, and whether the bot can book a callback rather than dumping a transcript.

6. What do maintenance, retraining and content updates cost after launch, and who owns the prompts and data? Agencies surveyed by GoodFirms budget maintenance at 15–25% of build cost per year (GoodFirms app development cost survey, 2026). For a chatbot, that includes re-testing after every model version change and updating the knowledge base every time the returns policy or catalog shifts. Get in writing that you own the prompts, the evaluation set, the conversation logs and any fine-tuned artifacts, so that switching vendors or models later is a migration, not a rebuild.

To get comparable quotes, send every vendor the same one-page brief. It should contain:

• Your monthly conversation volume today and the peak you expect in season, split by channel (web chat, app, messaging, phone).
• The systems the bot must read from and write to: commerce platform, order management, returns tool, PIM or product feed, help desk.
• The ten questions shoppers ask most, with the answer you would want and the source of that answer.
• Where humans must take over, and the hours a human is available.
• Your budget bracket for the build and a separate ceiling for monthly running cost.
• Whether the bot has to work inside a visual surface such as a configurator or 3D catalog.

A vendor that answers this brief with a fixed scope, a monthly cost model and a named integration list has done the work. One that answers with a discovery-workshop invoice has not, yet.

FAQ: what buyers ask before hiring an AI chatbot development company

How much does it cost to develop an AI chatbot for an ecommerce store?

Published floors on this list run from $10,000 (LeewayHertz, per Clutch) to $50,000 (Markovate, per Clutch), with the Culver City studio's brackets starting at $10,000–50,000 and a platform alternative at $9,000 per year (Botpress). For context, Clutch reports typical software projects at $10,000–49,000, and GoodFirms' 2026 survey of 100-plus software companies puts an AI-powered MVP at $50,000–125,000 and medium projects at $125,000–250,000. The spread depends on integrations, channels and how much action the bot takes, not on the chat interface.

How long does it take to build an ecommerce chatbot?

Only one agency here publishes timelines: 4–10 weeks, 3–6 months or 4–12 months depending on bracket. As a working rule, a support bot on a platform such as Botpress is a configuration project measured in weeks, while a custom assistant that writes to your order system, handles returns and is tested against real conversations is a months-long build. The longest part is rarely the model; it is integration, evaluation and the content nobody has written down yet.

Should I build a custom chatbot or use a platform?

Use a platform when the job is support deflection, the volume fits a published plan and nobody needs the bot to act inside custom systems; Botpress lists $9,000 per year for 1,200 conversations a month. Build custom when the bot must drive a product experience, quote against configuration rules or write into an order system that has no off-the-shelf connector. Many projects combine both: a platform for the help desk, a custom layer for product discovery or a 3D configurator.

Can a chatbot connect to our order management, ERP or product catalog?

Yes, if those systems expose an API or a data export the vendor can reach, and this integration is where most of the budget goes. Ask which platforms the vendor has connected in production, whether the bot will read only or also write (create a return, change an address), and what happens when the source system is slow or offline. A bot that reads inventory in real time and one that reads a nightly export are different products at different prices.

What does it cost to run the chatbot every month?

Three things: model tokens, retrieval and hosting, and people. At list prices checked September 20, 2026, low-cost models run from $0.20 per million input tokens (OpenAI gpt-5.6-luna) to $1 (Claude Haiku 4.5), with output four to six times more, and a vector database starts at $0–20 per month on Pinecone's Starter and Builder tiers. Maintenance is budgeted at 15–25% of build cost per year by agencies surveyed by GoodFirms (2026). Ask every vendor for a monthly model at your conversation volume before you compare build quotes.

Will an AI chatbot replace our support team?

Not on the evidence so far. In McKinsey's State of AI 2026 survey, only 14% of AI-using organizations said AI reduced overall headcount in the past year, although 39% expect a decrease next year (McKinsey / QuantumBlack, 1,719 respondents, 2026). The pattern in ecommerce is deflection of repetitive questions, order status, policy, sizing, with humans handling exceptions and complaints. Design the handoff as carefully as the bot; that is the part shoppers remember.