Survey: 65.0% of those using AI to select B2B products put AI's comparison into internal approval and proposal documents | an Itera survey

Itera, Inc. (head office: Shinjuku-ku, Tokyo; Representative Director: Takayuki Muto) ran two surveys: one on selecting B2B products among 5,000 working people aged 23 to 65 across Japan, and one on the use of AI in that selection among 300 people with experience of using generative AI for it.

60.3% of those involved in selection and comparison used generative AI. 68.7% have dropped or changed a candidate because of an AI answer, and 57.0% of the users were the decision-maker or the person responsible for the selection.

Key findings

• 24.4% were involved in selecting or comparing products or services at work in the past year (n=5,000), and 60.3% of them used generative AI or AI search for it (n=1,219)
• 65.0% have used AI’s comparison or recommendation in an internal approval or proposal document (n=300)
• 57.0% of users are the final decision-maker or the person responsible for the selection (n=300)
• Generative AI was used most at the stages of drawing up candidate products and companies (42.3%), organising the requirements (39.0%) and building the comparison table (38.0%) (n=300, multiple answers)
• 63.7% have been shown a product, service or company they did not know about (n=300)
• Of those, 50.8% requested materials or made an enquiry, 33.0% added it to the internal candidate list or comparison table, and 29.8% actually adopted or contracted it (n=191, multiple answers)
• 68.7% in total have dropped a candidate or switched to another because of an AI answer (n=300)
• The triggers were quoted negative reviews or reputation (51.0%), negative statements about the company or product (35.0%), an unfavourable assessment against competing products (29.1%) and information being out of date or contrary to fact (27.7%) (n=206, multiple answers)
• The AI used were ChatGPT 73.7%, Gemini 53.0%, Copilot 29.0% and Google’s AI Mode 24.7% (n=300, multiple answers)
• 77.3% think companies providing products and services should understand and improve how generative AI answers about them (n=300)

The findings in detail

1. 60.3% of those involved in selection used generative AI or AI search (n=1,219)

We asked 5,000 working people aged 23 to 65 across Japan whether, in the past year, they had used generative AI or AI search when selecting or comparing products and services to adopt or order at work.

1,219 people (24.4%) had been involved in selection or comparison, and the 735 among them who used generative AI represent 60.3%. As a share of all working people, that is 14.7%.

By company size, those 735 split as 14.4% at 1–10 employees, 13.6% at 11–50, 26.0% at 51–300, 16.5% at 301–1,000 and 29.5% at 1,001 or more. It is used regardless of company size.

2. 65.0% have used AI’s comparison in an approval or proposal document (n=300)

Generative AI’s output is not staying as reference material — it is being taken into the documents on which decisions are made.

3. 57.0% of users are the decision-maker or the selection lead (n=300)

4. Generative AI is used for drawing up candidates, organising requirements and building comparison tables (n=300, multiple answers)

Drawing up candidate products and companies came first at 42.3%, organising the requirements second at 39.0% and building the comparison table third at 38.0%.

All three are early stages, before the candidates have been narrowed. A product whose name does not come up at this stage loses the chance to appear in the comparison table at all.

5. The AI used were ChatGPT 73.7% and Gemini 53.0% (n=300, multiple answers)

Copilot at 29.0% and Google’s AI Mode at 24.7% follow.

6. The most compared category was business software and SaaS at 44.3% (n=300, multiple answers)

7. 63.7% were shown a product, service or company they did not know about (n=300)

We asked the 191 people who had been shown something new what action they took (multiple answers). 50.8% requested materials or made an enquiry, 33.0% added it to the internal candidate list or comparison table, and 29.8% actually adopted or contracted it.

Those who adopted or contracted represent 19.0% of all 300 people using generative AI.

8. 68.7% dropped a candidate or switched because of an AI answer (n=300)

68.7% in total had one of those experiences. Given the sample size, the 95% confidence interval runs from 63.4% to 73.9%.

9. The most common trigger was quoted reviews or reputation, at 51.0% (n=206, multiple answers)

Information about features or pricing being out of date or contrary to fact accounted for 27.7%. Barely any information appearing accounted for a further 13.1% — the state of a company’s official information is a direct reason for being dropped.

10. 77.3% think companies should understand and improve how AI describes them (n=300)

What the results suggest

The survey shows generative AI working in both directions when products and services are selected at work: adding candidates and cutting them. 63.7% were shown a product they did not know about; 50.8% of those requested materials or enquired and 29.8% adopted or contracted it. Meanwhile 68.7% have dropped a candidate or switched because of an AI answer.

What stands out is that generative AI’s output is not staying with the individual gathering information. It is used at the early stages — drawing up candidates, organising requirements, building the comparison table — and 65.0% of that output goes into internal approval or proposal documents. 57.0% of users are the decision-maker or the selection lead. How generative AI describes a company or product goes straight into the documents on which decisions are made.

The leading trigger for dropping a candidate was quoted reviews and reputation at 51.0%, with incorrect information about features or pricing accounting for 27.7%.

From these results, we believe companies providing products and services need to:

• Check regularly how generative AI answers about their products and services
• Inspect whether the pricing, features and adoption record in those answers match their current official information
• Understand which review sites and comparison sites AI refers to
• Keep pricing and feature information on the official site current, and publish it in a form AI can read

About the survey

LLMOcheki https://llmocheki.com/