Objection #5 in our series: "they all look the same". What if the problem is the way you compare them? Method, grid and real examples
This series looks at 7 good reasons a Head of eCommerce might have for not looking (or re-looking) into a conversational AI agent for their site. Objection #5: "We're unsure which one is right for us. There are many, they all look the same and we have other priorities." The stance: solutions only look the same when compared on features. Compared on the job they get done for the customer and for the business, they turn out to be very different.
In short: to choose an AI eCommerce assistant, start with the job you need it to do rather than the technology. Three criteria are enough to make the market readable: the job you give the assistant, the experience your customers expect, and the ROI metrics, including how they are measured.
Why all AI eCommerce assistants seem to look the same
The objection is understandable. On the Shopify App Store alone, the AI chatbot category lists more than 400 apps, according to AppJubilee. And most AI sales assistant vendors use the same lines: "recreate the in-store experience", "advanced intent detection", "guided selling".
The phenomenon goes beyond eCommerce. Gartner calls it "agent washing": many vendors rebrand existing chatbots, assistants or automation tools as "agentic AI" without any genuinely new capability. Of the thousands of vendors claiming agentic solutions, the firm estimates that only about 130 actually offer them (Gartner, June 2025).
Add speed to the mix: according to Rithum's 2026 Commerce Readiness Index, a survey of 200 retail and brand leaders in the US and UK, nearly three in four say AI is evolving faster than their organisation can adopt it.
This is the paradox of choice: when everything looks the same, doing nothing feels safer, and waiting for the market to settle looks like the prudent choice. That is exactly where the trap lies: the market won't settle on its own, and the most common comparison method, the feature list, is precisely the one that makes every solution look identical.
Three criteria for choosing an AI eCommerce assistant: start with the job, not the technology
"We need conversational AI" isn't a use case. "Help a customer confidently choose between 300 similar products" is. Before looking at the technology, and even before AI governance, three anchors help frame the need.
1. The job: what exactly do you need the assistant to do for your customers?
Each possible answer points to a very different tool. Do you need it to:
- help customers choose between 300 similar SKUs?
- guide them through a complex or high-value purchase?
- help a customer who knows what they need, but not what to search for?
- or answer order tracking and returns questions?
2. The expectation: retrieve a product or lead a decision?
What does a great experience look like for your customers? Should the assistant simply retrieve products, or ask questions, narrow the choice and explain why a product fits? Should it start the conversation itself, or wait to be asked? Should it act on the site (navigate, filter, add to cart) or stay within its chat window?
3. The ROI: which metrics, and measured how?
Several metrics can matter: conversion rate, average order value, customer service costs, or the share of visitors who come back because they got the right advice the first time. And above all: how are those metrics measured? It is the question asked least often, yet it is the one that lets you compare what vendors claim.
Chatbot, AI search, sales assistant: the five families of solutions
From a distance, the AI eCommerce assistant market looks uniform. In reality it splits into five families solving different problems:
- Conversational support, the eCommerce chatbot in the classic sense, to reduce repetitive questions.
- AI search and merchandising, to help visitors find what they are looking for.
- Algorithmic personalisation and recommendation, to grow the basket without conversation.
- Conversational sales assistants, to support the purchase decision.
- In the background, the agentic layer of platforms and consumer assistants (ChatGPT, Gemini, Perplexity), which captures part of the purchase intent away from the site.
The five players below all belong to the fourth family, sales assistants. They use very similar language. Read through the three anchors, however, they don't do the same job. Juliia is included in the grid, on the same terms, to allow comparison.
Six AI sales assistants put through the three criteria
For a detailed analysis of three of these solutions, see our Juliia, Dialog and iAdvize comparison.
The product page salesperson: Dialog
Dialog is probably the player whose messaging is closest to the idea of a "virtual salesperson". The assistant steps in where hesitation happens: product page, homepage, search bar. The Paris-based startup claims more than 300 client brands, including Delsey Paris and OhMyCream, and puts forward significant aggregate results. It is also preparing to surface its clients' products in consumer AI assistants. Its pricing is based on traffic (monthly unique visitors).
The AI and human advisor platform: iAdvize
A French pioneer of conversational commerce, iAdvize doesn't just sell an AI assistant but a platform orchestrating AI, human advisors, campaigns and conversation analytics. The job is different: scaling conversation for brands that already have advisory teams. That functional depth is reflected in the pricing, based on conversation volume and catalogue size.
Sales and support in one tool: Rep AI
Rep AI stands out on the experience anchor: its behavioural AI spots hesitant visitors and starts the conversation itself. The tool also covers post-purchase support (order tracking, returns). It is a relevant choice for a brand that wants a single tool across sales and customer service. Pricing combines traffic and number of products.
Choosing and buying without leaving the conversation: Dori
Launched in 2025, Dori quickly gained visibility on the Shopify App Store. Its strength is fluidity: comparison, recommendation and add-to-cart from the chat, with a module built into the product page. The vendor communicates on adoption (stores, conversations) rather than measured commercial impact.
A new entrant to watch: Advisor
Launched in September 2026, Advisor follows a mechanism close to other sales assistants: understand the need, recommend, guide to purchase. It is too early to assess its results: the app has no published reviews yet.
Expert advice before the purchase: Juliia
Juliia makes a deliberate scope choice: 100% pre-sale. It doesn't start the conversation, but once asked, it leads the sale. It qualifies the need, infers what the customer hasn't said, compares several options while explaining why each one fits, and navigates the site in step with the conversation. It doesn't handle post-purchase support and directs customers to customer service for human requests. This positioning fits large or technical catalogues, where advice is what makes the sale.
Two real conversations illustrate this depth of advice.
[Image: screenshot of the conversation. Suggested alt text: "Juliia conversation: a complete first bivouac kit for carp fishing, €1,000 budget"]
On a specialist fishing site, a customer simply writes: "something to bivouac for the first time on a carp fishing trip, max budget €1,000". A keyword search would have returned a list of shelters. Juliia understands that a first bivouac requires a complete set of equipment. It builds a coherent bundle of five products:
- a spacious, quick-to-pitch bivvy (shelter);
- a bedchair with a memory foam mattress;
- a sleeping bag designed for cool spring and autumn nights;
- a chair for the long hours of waiting;
- a head torch for setting up at night, a need the customer hadn't mentioned.
Total: €982.97, within budget.
Each product is justified against the customer's situation. The products are also chosen to work together: the bedchair is "perfectly compatible with the dimensions of the selected bivvy". They appear in the conversation with their availability and an add-to-cart button. The customer didn't name a single product: they described a situation, and Juliia built the answer a specialist salesperson would give.
On another site, a customer states a budget of €1,500 to €3,000, then narrows it to €2,000 mid-conversation. Juliia adjusts its selection without losing the context of the conversation. It then compares three models on the criteria the customer expressed, rather than pushing a single product. In another exchange, asked about stock in a store it has no access to, it says so clearly and suggests an alternative (online reservation, in-store pickup) rather than making up an answer.
That difference, between retrieving a product and leading a purchase decision, is exactly what feature lists don't show.
What the grid reveals
The same words don't describe the same job. "Sales assistant" covers very different tools:
- a hybrid AI-human platform (iAdvize);
- a combined sales and support tool (Rep AI);
- a product page advisor (Dialog, Dori);
- an expert pre-sale advisor (Juliia).
A brand looking to cut support tickets and a brand selling 3,000 technical products don't need the same tool.
The experience is in the details of the scope. Who starts the conversation, the assistant or the customer? Does the assistant act on the site or stay in its window? Does it cover post-purchase? Does it hand over to a human? These choices, rarely highlighted on homepages, deeply change what the customer experiences.
Evidence exists, but it doesn't all measure the same thing. Most published figures compare visitors who used the assistant with those who didn't: "8x", "10x", "+300%". These gaps are real, but they blend two effects: the assistant's, and the fact that visitors who start a conversation are often already further along in their purchase intent. So the question to ask any vendor, Juliia included, is: how much of this result is genuinely incremental? Gartner itself recommends pursuing agentic AI only where it delivers clear value or ROI.
Pricing models reflect the job. Pricing per visitor, per conversation or per catalogue sends different signals: with usage-based billing, the assistant's success drives the bill up. With a subscription that includes unlimited conversations, it stays stable as usage grows. We look at this question in more detail in our article on the budget objection.
"We have other priorities": a matter of weeks, not months
That leaves the second half of the objection: even with a clear way to choose, the topic would still come after everything else on the roadmap.
The idea that an AI assistant is a heavy project tying up teams for months belongs to an earlier generation of tools. Juliia's latest customer went live in under 30 days, and their team spent no more than a few hours here and there testing it. On that site, visitors who interact with Juliia convert 30% better, with average order value up 25% and assisted revenue accounting for 6% of online revenue.
A targeted win that requires neither a site rebuild nor extra headcount can fit into an already full roadmap. Go-live effort then becomes a selection criterion like any other, to check with each vendor.
Five questions to ask every AI eCommerce assistant vendor
- What specific job is your assistant designed for, and for what type of catalogue?
- Who starts the conversation, and what can the assistant do beyond its chat window?
- How are your results measured: assisted vs. unassisted visitors, or incremental gain?
- How much time and effort will our teams need for a first go-live?
- How does the price change if usage takes off?
The right choice starts with the right question
AI eCommerce assistants only look alike as long as you view them through the technology. Once the job, the expected experience and the ROI metrics are clear, the market becomes readable, and so does the choice. And since deployment is now measured in weeks, the priority isn't to find the best AI on the market, but to quickly test, on your own site, the one that best does the job you need.
If your catalogue is large or technical and advice is what makes the sale, a demo of Juliia on your own use cases is a good place to start asking these five questions.
FAQ
What is an AI eCommerce assistant?
An AI eCommerce assistant is a conversational tool built into an online store to help visitors, for example to find a product, choose between several options or get an answer about their order. Depending on the solution, it falls under customer support, product search or guided selling. These families meet different needs.
What is "agent washing"?
It is the term Gartner uses for rebranding existing chatbots, assistants or automation tools as "agentic AI" without any genuinely new capability. The firm estimates that only about 130 of the thousands of vendors claiming agentic solutions actually offer them.
What criteria should you use to choose an AI eCommerce assistant?
Three criteria help frame the choice:
- the job you need the assistant to do (help customers choose, guide a complex purchase, answer order tracking questions);
- the experience you want for your customers (who starts the conversation, does the assistant act on the site);
- the ROI metrics, and how they are measured.
How long does it take to deploy an AI eCommerce assistant?
It depends on the solution and the scope. Current tools often go live within a few weeks. At Juliia, the latest customer went live in under 30 days, with a few hours of testing here and there from their team. It is a question worth asking every vendor.
How can you tell whether a vendor's results are reliable?
Ask how they are measured. Most figures compare visitors who used the assistant with those who didn't, and the former are often already further along in their purchase. Only a measure of incremental gain, for example against a control group, isolates the assistant's real effect.
Keep exploring
Dive into more articles to explore the future of conversational commerce and AI for conversion.
AI eCommerce assistants: how to choose when they all look the same
No budget for conversational AI? Wrong question
Reason #4 in the series: budget. What if the real question isn’t what the tool costs, but what the problem it solves is already costing?
