Should AI answer your customers?
AI that helps people reply is well supported: at a large support centre, an AI assistant raised issues resolved per hour by 15%, and by 34% for newer agents. AI replacing people is contested. Most customers say they would rather companies didn’t use AI in service, mainly for fear of not reaching a person. Angry customers react badly to humanlike bots, and how a bot is disclosed changes whether people buy. In the EU, telling people they are talking to an AI becomes law in 2026.
What to do
- Let AI answer routine questions instantly, with a person one tap away.
- Be open that it’s the business’s assistant, in one friendly line.
- Send complaints and unusual questions to a person quickly.
- Lead your marketing with the benefit, like “answers in seconds, day or night”, rather than with “AI”.
What the research says
- Peer-reviewed or official
Generative AI at Work
Brynjolfsson, Li & Raymond, QJE 2025; NBER WP 31161
Giving support agents an AI assistant raised issues resolved per hour by 15% on average. Novice and low-skilled agents improved by 34% (working-paper figure). Customers became more polite and asked for a manager less often.
What it means for you: AI-suggested replies help a small team, especially newer staff, answer faster and better. Keep a person in the loop.
Keep in mind: One large support operation where people sent the AI’s suggestions. The most skilled agents gained little.
Read the sourceMethod and quote
Staggered roll-out of a generative-AI conversational assistant across 5,172 customer-support agents (5,179 in the working paper). Quasi-experimental.
“From the journal abstract: "Access to AI assistance increases worker productivity, as measured by issues resolved per hour, by 15% on average, with substantial heterogeneity across workers… customers are more polite and less likely to ask to speak to a manager." From the NBER abstract: "…including a 34% improvement for novice and low-skilled workers but with minimal impact on experienced and highly skilled workers.”
- Peer-reviewed or official
AI Act, Article 50
European Union, 2024
Providers must design chatbots so that people are told they are talking to an AI, unless this is obvious. Deployers must disclose deep-fake image, audio or video. The Act applies generally from 2 August 2026.
What it means for you: Telling people they’re talking to an AI is becoming the law. Do it in a friendly way, with a person one tap away.
Keep in mind: EU law, applying from 2 August 2026. Check how it applies where you sell.
Read the sourceMethod and quote
Regulation.
“Providers shall ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system, unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect…" Article 113: "It shall apply from 2 August 2026.”
- Peer-reviewed or official
Business Trends and Outlook Survey: AI use
US Census Bureau, 2026
17–20% of US businesses used AI (Dec 2025–May 2026), against 37% of firms with 250+ employees. Less than 20% of firms with four or fewer employees use AI, and use did not grow significantly among firms with fewer than 20 employees.
What it means for you: Most small businesses haven’t adopted AI themselves; fewer than one in five of the smallest firms use it.
Keep in mind: US Census Bureau survey, December 2025 to May 2026. US only.
Read the sourceMethod and quote
Biweekly, nationally representative BTOS, 14 Dec 2025–3 May 2026. The question changed in Nov 2025 to "in any business function", so it is not comparable with earlier waves.
“The BTOS data (December 2025 to May 2026) show that overall AI usage hovered between 17% and 20%…" Also: "37% of firms with at least 250 employees reported using AI…" And: "Less than 20% of firms with four or fewer employees reported using AI.”
- Peer-reviewed or official
Generative AI and the SME Workforce: New Survey Evidence
OECD, 2025
31% of SMEs use generative AI, ranging from 24% in Japan to 39% in Germany. Non-adopters cite unsuitability (57%), legal and copyright concerns (54%), data concerns (52%) and lack of staff skills (50%).
What it means for you: Owners’ main AI worries are fit, legality and data. A good AI tool should answer those up front.
Keep in mind: OECD countries; India, the UAE and the US weren’t covered.
Read the sourceMethod and quote
A representative late-2024 survey of more than 5,000 SMEs in Austria, Canada, Germany, Ireland, Japan, Korea and the UK.
“It is used (by the respondent or a colleague) in 31% of SMEs, ranging from 24% in Japan to 39% in Germany." The barriers listed are: "unsuitability to the SME's work (reported by 57% of non-adopters); concern about copyright, legal or regulatory issues (54%) and about what happens to the information fed into generative AI models (52%); and a lack of skills among employees (50%).”
What cuts the other way
Research that complicates the picture, and what to take from it.
- Peer-reviewed or official
Machines vs. Humans: The Impact of AI Chatbot Disclosure on Customer Purchases
Luo, Tong, Fang & Qu, Marketing Science 2019
Undisclosed chatbots sold as well as proficient human agents. Disclosing the bot before the conversation cut purchase rates by more than 79.7%. Late disclosure and customers' prior AI experience softened the effect.
What it means for you: How you disclose a bot matters. Be open that it’s an assistant, let it show it can help, and keep a person one tap away.
Keep in mind: Outbound sales calls with 2018-era voice bots, not customers’ own questions.
Read the sourceMethod and quote
Field experiment with more than 6,200 customers randomized to outbound sales calls from voice chatbots or humans.
“undisclosed chatbots are as effective as proficient workers and four times more effective than inexperienced workers in engendering customer purchases. However, a disclosure of chatbot identity before the machine–customer conversation reduces purchase rates by more than 79.7%." The abstract adds: "such negative impact can be mitigated by a late disclosure timing strategy and customer prior AI experience.”
- Independent research
Gartner Survey Finds 64% of Customers Would Prefer That Companies Didn't Use AI For Customer Service
Gartner, 2024
64% of customers would prefer firms didn't use AI in customer service, and 53% would consider switching. The top fear is that it will be harder to reach a person.
What it means for you: Customers fear AI will make it harder to reach a person. Make handover to a human easy.
Keep in mind: Stated attitudes in a survey, not behaviour.
Read the sourceMethod and quote
Survey of 5,728 customers, December 2023 (countries not given in the release).
“Sixty-four percent of customers would prefer that companies didn't use artificial intelligence (AI) in their customer service… Furthermore, 53% of customers would consider switching to a competitor if they found out a company was going to use AI for customer service. A Gartner survey of 5,728 customers conducted in December 2023…" The release also says: "The top concern that consumers have about AI in customer service is that it will get more difficult reaching a person…”
- Peer-reviewed or official
Blame the Bot: Anthropomorphism and Anger in Customer–Chatbot Interactions
Crolic, Thomaz, Hadi & Stephen, JM 2022
For angry customers, a humanlike chatbot lowers satisfaction, firm evaluation and purchase intentions. Non-angry customers are unaffected.
What it means for you: Send angry customers to a person quickly; a humanlike bot makes them angrier.
Keep in mind: Experiments. Customers who weren’t angry were unaffected.
Read the sourceMethod and quote
Five studies: a large real-world telecom chatbot dataset and four experiments.
“when customers enter a chatbot-led service interaction in an angry emotional state, chatbot anthropomorphism has a negative effect on customer satisfaction, overall firm evaluation, and subsequent purchase intentions.”
- Peer-reviewed or official
Adverse impacts of revealing the presence of 'Artificial Intelligence (AI)' technology in product and service descriptions on purchase intentions
Çiçek, Gürsoy & Lu, JHMM 2024
Mentioning "artificial intelligence" in product or service descriptions lowered purchase intention, through lower emotional trust. The effect was stronger for high-risk products.
What it means for you: Leading with “AI” in product copy can lower trust. Lead with the benefit instead.
Keep in mind: Stated purchase intent in experiments.
Read the sourceMethod and quote
Six experiments, analysed with t-tests and PROCESS mediation models.
“the inclusion of the 'Artificial Intelligence' term in descriptions of products and services decreases purchase intention, and that emotional trust mediates this relationship.”
- Peer-reviewed or official
The AI-authorship effect
Kirk & Givi, Journal of Business Research 2025
When consumers believe an emotional message was written by AI, positive word of mouth and loyalty fall. The effect is weaker for factual messages and when AI only edits.
What it means for you: Use AI for factual replies; keep emotional messages like thank-yous and apologies human.
Keep in mind: Experiments; the effect is weaker when AI only edits.
Read the sourceMethod and quote
Seven preregistered experiments.
“Seven preregistered experiments demonstrate that when consumers believe emotional marketing communications are written by an AI (vs. a human), positive word of mouth and customer loyalty are reduced… this 'AI-authorship effect' is attenuated for factual (vs. emotional) messages (Study 2); when an AI only edits the communication (Study 3)…”
- Peer-reviewed or official
Navigating the Jagged Technological Frontier
Dell'Acqua et al., HBS WP 24-013, 2023
Consultants with GPT-4 completed 12.2% more tasks, 25.1% faster and at higher quality on tasks inside AI's competence. On a task outside it, they were 19% less likely to produce a correct solution.
What it means for you: AI is great at routine questions and risky on unusual ones. Hand off the edge cases.
Keep in mind: Consultants at one firm, on writing and analysis tasks.
Read the sourceMethod and quote
Preregistered experiment with 758 BCG consultants.
“subjects using AI outperformed those not using AI, completing 12.2% more tasks and completing them 25.1% more quickly on average while also delivering solutions of significantly improved quality. However, for a complex managerial task selected to be outside the frontier, subjects using AI were 19% less likely to produce correct solutions compared with those without AI…”
Sources checked as of October 25, 2026. Links go to the original publication.
More questions
- Does replying fast really win customers?
- Why lead with WhatsApp in India?
- Is keeping customers worth more than finding new ones?
- Do reminders really cut no-shows?
- Can you trust your marketing numbers?
- Does automating the admin really save time?