AI Enquiry Assistants: Where They Fit and Where They Don't

Every business fielding customer enquiries has probably had the same thought at some point: could an AI assistant handle some of this?

It's a reasonable question. It's also a question that gets a much more useful answer once you separate what these tools are actually good at from what the marketing around them promises.

This isn't a case against using one. Davina Connect has built and deployed AI enquiry assistants as part of connected client systems, including Katie, our own reactive AI customer assistant.

It's a case for being precise about where they genuinely help and honest about where a person still needs to be in the loop.

What an AI enquiry assistant is actually being asked to do

Strip away the branding and most AI enquiry assistants deployed by small and medium-sized businesses today do a narrower job than the term suggests.

They pick up a message, work out roughly what the person wants, ask a few qualifying questions and either answer directly, book something or pass the conversation on.

That narrower reality shows up clearly in the UK data.

The Department for Science, Innovation and Technology's AI Adoption Research found that among UK businesses already using AI, 85% are using it for text and language tasks while only 7% are using anything that could be described as agentic or fully autonomous.

Most AI assistants handling enquiries today are closer to a well-briefed junior team member working within defined instructions than an independent decision-maker.

Adoption itself has moved quickly. The British Chambers of Commerce, working with Atos, found active AI use among UK firms reached 54% in 2026, up from 35% the year before.

But that headline figure hides real unevenness once you look at what businesses are actually using AI for.

DSIT's UK Business Data Survey 2026 found that customer-service chatbot use remains low among smaller firms: 4% of sole traders, 4% of micro businesses, 6% of small businesses and 7% of medium businesses compared with 21% of large businesses.

AI enquiry assistants are still a much newer habit for smaller businesses than the general AI adoption headlines might suggest.

Where they genuinely help

The clearest benefit is availability and consistency, not intelligence.

An enquiry arriving at 9pm on a Sunday can receive the same prompt first response as one arriving at 11am on a Tuesday. Qualifying questions can also be asked consistently rather than depending on which member of staff happens to be available.

Salesforce's 2026 State of Service research reported that AI agent adoption among service organisations had risen to 66% from 39% the previous year. It also reported that 70% of teams deploying one saw measurable value within 60 days, according to reporting by Tommaso Maria Ricci.

For a business relying on enquiries to generate booked calls or appointments, this is the kind of repetitive and well-defined task an assistant can take on.

It can capture details consistently, ask the same qualifying questions and reduce the chance of an enquiry disappearing between one reply and the next.

Where they reliably struggle

The gap starts to open around anything that requires judgement rather than a defined process.

Qualtrics' 2026 Consumer Experience Trends research found that nearly one in five consumers who used AI for customer service reported no benefit from the experience at all, according to reporting by IrisAgent.

The same research placed AI customer service among the weaker AI use cases for convenience and usefulness. That is a noticeably different picture from the adoption headlines.

AI can recognise that someone sounds frustrated, but genuine empathy and judgement still belong with a person. It can also follow a defined qualification journey accurately, but unusual conversations can quickly move outside that path.

That is why unusual circumstances, sensitive conversations and decisions requiring judgement still need a clear route to a person.

The point most businesses miss: who is responsible when it gets it wrong?

In 2024, a Canadian tribunal ruled on a case that has become a widely cited example of this issue.

A customer asked Air Canada's website chatbot about bereavement fares and the chatbot gave an incorrect answer. When the airline later refused to honour it, the customer took the case to the British Columbia Civil Resolution Tribunal.

As reported by the American Bar Association, Air Canada argued that the chatbot was effectively responsible for its own answers.

The tribunal rejected that argument and treated the chatbot as part of the company's website. Air Canada was ordered to pay the fare difference.

This was a Canadian case, not a ruling on UK law. The practical lesson is still useful: a business should not assume responsibility sits with the AI itself simply because the information came from an automated system.

An AI enquiry assistant communicating with customers therefore needs appropriate controls, clear boundaries and human oversight.

Why the handover matters more than how clever the AI is

The quality of an AI system isn't measured by how rarely it hands a conversation to a person.

What matters is whether the handover works when it needs to happen.

A well-designed handover preserves the information already gathered so the customer doesn't have to repeat the whole conversation to a person who is starting from nothing. This is a common principle in guidance on AI-to-human handover design.

This is also where the business's own judgement should sit.

Deciding when a conversation needs a person and ensuring the handover carries the right context are design decisions about the business's own customer relationships.

The AI can flag that a handover might be needed. What happens next should be determined by the process around it.

Where this fits into a connected business

This is really the same principle behind what we describe as business disconnection.

Adding a new piece of technology before understanding how the existing process works can simply create one more system to keep track of.

An AI enquiry assistant added without a clear view of how enquiries, information and follow-up already connect risks becoming a faster version of the same disconnection rather than resolving it.

An AI assistant works best as one part of a connected journey. It can work from the same CRM record as other channels, hand over cleanly when a conversation needs judgement and respond to what staff do manually elsewhere in the same system.

That's a different question from simply asking, “Should we get a chatbot?” It's much closer to looking at the three quiet signs of business disconnection before adding another system.

The UK data suggests this connected approach is still far from universal.

DSIT's UK Business Data Survey 2026 found that among businesses using AI, only 21% said their AI tools were integrated into existing business systems, such as being embedded within a CRM rather than sitting apart from it.

What this looks like in practice

In one Davina Connect implementation for Expert Growth Marketing, Katie wasn't built as a standalone chatbot.

She became part of the customer journey when a contact replied. The surrounding CRM and workflows recognised that engagement and adjusted what happened next. Human intervention remained part of the same connected process rather than sitting outside it.

AI can handle the repetitive first pass of a conversation and gather useful information. It doesn't replace the judgement of the person who ultimately owns the customer relationship.

It works best when the system around it knows the difference.

Questions worth asking before deploying one

• What specific, repeatable task is this assistant being asked to do rather than simply “handle enquiries” in general?
• What happens when a conversation moves outside the expected path? Does it hand over cleanly or try to continue?
• Who checks what it says and how often?
• Does it write to the same CRM record as every other channel or become one more disconnected source of customer information?
• If a person steps into the conversation directly, does the wider system recognise what has happened and adjust accordingly?