No budget for conversational AI? Wrong question

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 #4: “We don’t have the budget to invest in additional gimmicks for our eCommerce.” The stance is simple: the question isn’t what an AI assistant costs, but what the problem it solves is already costing.

Gimmick or growth lever

The word “gimmick” is deliberately provocative. No one in the eCommerce ecosystem seriously thinks of technology as a gimmick. But the word does capture something real: a gimmick is fun, and it adds no value.

The budget objection often rests on that unspoken suspicion. An AI assistant would be one more layer on the site, impressive in a demo, with no measurable impact on revenue.

So the question to ask is: what value does an eCommerce AI assistant actually create, and can it be measured?

Feature or investment: changing the budget category

When technology is seen as the main cost centre of an eCommerce site, cost control naturally becomes a key lever to protect margin. In that frame, the objection is perfectly rational.

If conversational AI is treated as one more feature, like a new filter or a widget, the reasoning holds: why spend more on a feature?

But an assistant that influences conversion rate, average order value, customer service costs and customer knowledge no longer belongs in a “feature” budget. It belongs in an investment decision, judged on its return, like any other growth lever.

Early market data points the same way. According to Salesforce, retailers that had deployed their own AI agent saw average sales growth of 6.2% over the 2025 holiday season, compared with 3.9% for the rest. On the consumer side, 31% of French online shoppers already use generative AI for their purchases, according to a Fevad/Odoxa study cited by KPMG and Fevad in September 2026.

The hidden costs of the status quo

Rather than asking “what does an AI assistant cost?”, it is more useful to look at what the current situation already costs:

• Customer hesitation: the visitor who can’t decide between two products and leaves without buying.
• Inefficient product discovery: the customer who knows what they need, but not what to search for, and doesn’t find it.
• Abandoned decisions: the basket that doesn’t convert for lack of reassurance at the right moment.
• Repetitive support: the same questions handled over and over by customer service teams.

These costs are real, but not immediately visible. They are spread across lost conversions, customer service headcount and CRM. No budget line shows them, which makes them easy to ignore.

A few benchmarks make them tangible. According to the Baymard Institute, 70.22% of online shopping carts are abandoned on average (based on 50 studies). On the service side, Pandora’s AI agent has handled 60% of service cases without human intervention since its launch in January 2025, according to Salesforce.

Running the numbers on a concrete case

Take a site doing €4M in online revenue. An AI assistant involved in 5% of that revenue accounts for €200,000 in assisted revenue.

Within that segment, average order value rises by a measured 25%. Without the assistant, those same orders would have been worth around €160,000. The incremental gain is therefore around €40,000 a year: revenue created by the assistant, not simply revenue it happened to be present for.

Against that, assume an average cost of €590 a month, or €7,080 a year: about 18% of the incremental value generated. At that rate, the assistant pays back its annual cost in just over two months.

And this is a conservative calculation: it only counts the effect on average order value, leaving out higher conversion and customer service savings.

In the field, the results follow. For a Juliia client in outdoor retail, with a catalogue of several thousand technical products, the assistant generated €155,000 in assisted revenue within the first three months, or 3.6% of online revenue, with average order value up 38%. And 70% of interactions take place outside customer service hours: questions that, without an assistant, would have gone unanswered at the very moment the customer was hesitating. At that level, the question is no longer what the tool costs, but how to increase the number of interactions with Juliia to sell more!

Visitors who reach a site after chatting with an AI already convert better. According to Adobe Analytics, over the 2025 holiday season, visits to US retail sites from generative AI tools converted 31% more than other traffic sources.

Traffic acquisition costs are measured rigorously, down to the click. The same discipline can apply to the technology designed to convert that traffic.

Test, measure, then fund

The budget objection remains valid until the incremental economics are proven. The point isn’t to invest on the promise of a return, but to test whether conversational AI creates measurable incremental value on your own site, and to fund it once that value is established.

In practice, that means setting the metrics before the test (conversion rate, average order value, share of assisted revenue) and comparing visitors who use the assistant with those who don’t.

The real budget question

So the real budget question isn’t affordability, it’s measured gain on your own site. A gimmick costs money. A measured lever makes money.