Can AI respond to website enquiries?

Yes. AI can help respond to enquiries coming through your website.

It can potentially read what somebody has submitted, understand what they're asking for, check relevant business information, identify anything missing, prepare an appropriate response and make sure the enquiry reaches the right next step.

Depending on the boundaries you set, that response could be reviewed by a person before it's sent, or routine enquiries could potentially be handled automatically.

And no:

THAT DOESN'T MEAN YOU NEED A CHATBOT.

The customer can complete the same ordinary contact form they use today.

The AI can do its work after they press Send.

The problem with the traditional contact form

Imagine somebody visits your website at 7:14pm.

They've spent 20 minutes looking through your services.

They're interested.

They complete the form:

"Hi, we're looking to replace our current website. We're a property company with offices in London and Manchester. The current site is WordPress. We'd ideally like the new one running before January and we'd like to discuss whether you can integrate it with our existing property system."

They press:

SEND.

What happens next?

In plenty of businesses:

An email arrives in an inbox.

That's it.

The customer sees:

"Thanks for your enquiry. A member of our team will be in touch shortly."

And the actual sales process waits for a person.

What could happen instead?

The form submission can become the trigger for a workflow.

• UNDERSTANDWhat does this person actually want?
• EXTRACTWhat useful information have they already given us?
• CHECKExisting customer? Previous enquiry? Relevant service?
• FINDWhat approved business information is relevant?
• IDENTIFY GAPSIs anything important missing?
• PREPAREWhat is the appropriate response?
• AUTHORITY CHECKCan this be sent? Does somebody need to approve it?
• RECORDCreate or update the CRM.
• NEXT ACTIONBook / Gather information / Assign / Follow up / Escalate

Now the contact form isn't simply sending an email.

It's starting a process.

The AI doesn't need to answer immediately

This is worth saying because "AI responding to enquiries" can quickly turn into:

INSTANT RESPONSES! 24/7! NEVER MISS A LEAD!

Speed can be useful.

But speed isn't the only objective.

A bad response sent in four seconds isn't automatically better than a useful response sent later.

The goal should be:

THE RIGHT ENQUIRY GETS THE RIGHT NEXT ACTION.

Sometimes that means responding immediately.

Sometimes it means preparing something for the team tomorrow morning.

Sometimes it means asking one missing question.

And sometimes it means:

This needs a person.

What can AI understand from an enquiry?

Potentially quite a lot.

Take our example:

"We're looking to replace our current website. We're a property company with offices in London and Manchester. The current site is WordPress. We'd ideally like the new one running before January and we'd like to discuss whether you can integrate it with our existing property system."

AI could potentially structure that into:

Already, the person receiving the enquiry has a better starting point.

But we can go further.

It can check what your business already knows

Suppose the email address belongs to somebody already in your CRM.

The workflow could potentially find:

• an existing company record,
• previous enquiries,
• earlier conversations,
• current opportunities,
• customer status,
• or other information relevant to this particular job.

Now imagine that the same person contacted you six months ago.

A human might remember.

They might not.

The system doesn't have to rely entirely on memory.

That context can be surfaced before the response is prepared.

It can work out what information is missing

This is one of the more useful capabilities.

Traditional forms have a problem.

You either ask very little and spend time gathering information afterwards.

Or you build a monster form:

Fine, perhaps not the last one.

But you know the kind of form.

AI gives us another possibility.

DON'T ASK EVERYONE EVERYTHING.

Use what they've already told you.

Then work out what's actually missing.

Could AI ask the customer for more information?

Potentially.

Suppose the enquiry is clear except for one thing:

the business needs to understand which property management system requires integration.

The AI could prepare:

"Thanks for getting in touch. The integration you've mentioned is something we'd want to understand before advising properly. Could you let us know which property management system you're currently using?"

That's considerably more useful than:

"Thanks for your enquiry. Someone will be in touch."

And more respectful than sending a generic questionnaire asking ten things the customer already explained.

But should AI send the response itself?

Maybe.

This is where capability and authority need separating.

The AI may be perfectly capable of preparing the response.

That doesn't automatically mean it should have permission to send it.

A business might start with:

• READAI reads incoming enquiries.
• RECOMMENDAI recommends what should happen.
• PREPAREAI drafts the response.
• HUMAN APPROVALSomeone checks and sends.

That alone may remove a lot of work.

Later, you might decide certain types of routine enquiry can move to:

ACT WITHIN LIMITS

For example:

• Known service
• Approved information
• No unusual pricing
• No complaint
• No contractual commitment
• No uncertainty requiring expertise

The AI can respond.

Anything outside those limits:

ESCALATE.

A real-looking example

Imagine a glazing company receives:

"Hi, we're building a rear extension in Ealing and need a flat rooflight approximately 2m x 1m. We already have a builder and only need supply. Do you make them to size and can you deliver?"

A useful system could potentially understand:

If those questions can be answered from approved company information, the system can prepare an accurate response.

If an exact price requires specification details:

DON'T INVENT THE PRICE.

Ask for what's missing or pass it to the appropriate person.

That's the difference between useful AI and an enthusiastic hallucination with access to your inbox.

Where does the information come from?

This is crucial.

AI can't provide reliable business-specific answers simply because it is "smart".

It needs access to appropriate information.

Depending on the enquiry, that might include:

• service information,
• product information,
• FAQs,
• locations served,
• availability rules,
• pricing information,
• technical specifications,
• delivery information,
• CRM records,
• previous conversations,
• and business policies.

The important principle is:

THE AI NEEDS THE RIGHT INFORMATION FOR THE JOB.

Not every document your company has ever produced.

Website content alone may not be enough

This is an important implementation point.

Imagine your website says:

"We offer bespoke solutions for businesses of all sizes."

Lovely.

Completely useless for answering:

"Can you integrate with version 8.2 of our property management system?"

The AI needs information that can actually support the decision or response.

Sometimes building a useful agent reveals that important business knowledge is:

• in people's heads,
• in old emails,
• in PDFs,
• in spreadsheets,
• spread across systems,
• or not documented at all.

That's not primarily an AI problem.

It's an information problem.

And it's often worth fixing.

What if the enquiry needs a price?

Now authority becomes more important.

There are several different jobs hiding inside:

Can AI answer pricing enquiries?

Don't bundle all five together under:

AI CAN DO PRICING.

Be precise about the job.

What if the customer asks something the AI doesn't know?

This is where a good system becomes noticeably different from a bad chatbot.

The wrong approach is:

Try really hard to produce an answer anyway.

The better approach is:

STOP → EXPLAIN → ESCALATE.

For example:

STOP → EXPLAIN → ESCALATE

That's useful.

The AI has still done work.

It understood the enquiry.

Checked the information.

Identified the problem.

And routed it correctly.

"I DON'T KNOW" CAN BE A SUCCESSFUL OUTCOME.

Can AI route website enquiries?

Yes, and this may be useful even if you don't want AI replying to customers.

Imagine your website receives:

• sales enquiries,
• support questions,
• job applications,
• supplier messages,
• existing customer requests,
• partnership enquiries,
• and spam.

A single hello@ inbox can become a miniature sorting office.

AI can potentially interpret the message and recommend or trigger the appropriate route.

For example:

• New sales opportunity → Sales
• Existing customer support → Support
• Technical question → Technical team
• Invoice query → Accounts
• Job application → Recruitment
• Uncertain → Human review

The agent doesn't have to talk to the customer to create value.

Can AI create the CRM record too?

Potentially.

Once the enquiry has been understood, a workflow could create or update the appropriate record.

For example:

Now information gathered during the enquiry doesn't need to be manually copied into another system.

But again:

NOT EVERY FIELD NEEDS THE SAME AUTHORITY.

Routine contact information may be straightforward.

Changing an existing opportunity value or customer status may deserve more control.

Can AI book the meeting?

Potentially, yes.

If the enquiry is appropriate for a sales conversation, the workflow could potentially:

• offer suitable availability,
• allow the customer to choose a time,
• create the calendar event,
• update the CRM,
• and prepare the meeting context.

So the process becomes:

• Enquiry
• Understand
• Qualify
• Respond
• Book
• Prepare
• HUMAN CONVERSATION

That's a good example of agentic selling.

AI handles more of the work around the conversation.

The person still handles the part where the person is valuable.

What about chatbots?

Chatbots can absolutely be useful.

But don't confuse:

AI WEBSITE ENQUIRIES

with:

PUT A CHATBOT ON THE HOMEPAGE.

They're not the same thing.

The chat window is the interface. The workflow is the useful bit.

A chatbot is an interface.

The agentic workflow is the process underneath.

A customer could interact through:

• a contact form,
• email,
• live chat,
• a chatbot,
• a booking form,
• or potentially another channel.

The interesting part is what the system can do after information arrives.

• Can it understand?
• Can it retrieve context?
• Can it decide what it needs?
• Can it use tools?
• Can it prepare an action?
• Can it continue the process?
• Can it recognise when it should stop?

That's much closer to what we mean by an AI sales agent.

When should AI NOT automatically answer?

There should be boundaries.

Complaints

A customer expressing anger or dissatisfaction may deserve immediate human attention.

Sensitive situations

Don't automate conversations simply because they're technically classed as enquiries.

Unusual commercial requests

Discounts, unusual terms, bespoke agreements and exceptions may require judgement.

Unclear information

If the AI can't establish the facts required for a reliable answer, it should not improvise.

High-value or important relationships

Sometimes the personal response is part of the value.

Anything outside its information

If the answer isn't supported by reliable business information:

DON'T MAKE IT UP.

Escalate.

Fast isn't always better

This deserves repeating.

A lot of AI sales technology is marketed around:

Respond in seconds.

But imagine receiving this four seconds after submitting a thoughtful enquiry:

"Hi Sarah! Great question! We'd absolutely LOVE to help you transform your business journey!"

Nobody needed that faster.

A useful response may take slightly longer because the system:

• checked the account,
• found the relevant information,
• identified a missing detail,
• or passed something to a person.

Measure:

USEFUL RESPONSE TIME.

Not merely:

AUTOMATED RESPONSE TIME.

What should you measure?

If you introduce AI into website enquiry handling, don't measure success by:

AI responded to 94% of enquiries.

That doesn't tell you whether the responses were any good.

Look at things like:

• How quickly do genuine enquiries reach a useful next action?
• How many require manual sorting?
• How much information has to be copied manually?
• How many enquiries need repeated questions?
• How often does the AI escalate?
• Are escalations appropriate?
• Are customers receiving accurate information?
• Are more enquiries reaching the right person?
• Are fewer genuine opportunities being forgotten?

And ultimately:

DID THE ENQUIRY PROCESS GET BETTER?

How should a small business start?

Start with the inbox, not the chatbot.

Take your last 30 or 50 website enquiries.

Look at them.

• What types do you receive?
• What did somebody have to do after each one arrived?
• What information was already present?
• What did they need to look up?
• What questions did they ask repeatedly?
• Which enquiries were easy?
• Which required judgement?
• Which went to the wrong person?
• Which sat unanswered?

Then map the process.

You may discover that AI isn't required for every step.

Good.

Use:

• Automation for predictable movement.
• AI for interpretation.
• Humans for consequential judgement.

A sensible first version

You don't need to begin with autonomous responses.

I'd start here:

• NEW ENQUIRY
• AI READSUnderstands and categorises it.
• AI PREPARESSummary + recommended next action + draft response.
• HUMAN APPROVESPerson checks and sends.
• SYSTEM UPDATESCRM + next action.

Run that for a while.

See what the AI gets right.

See what it gets wrong.

See which responses are so routine that approval adds little value.

Then decide whether some narrowly defined cases should move to automatic action.

EARN THE AUTONOMY.

Don't start by giving it away.

So, can AI respond to website enquiries?

Yes.

But the interesting opportunity isn't simply generating an instant reply after somebody submits a form.

AI can potentially understand the enquiry, retrieve relevant information, identify what's missing, prepare the right response, update your systems, route the opportunity and make sure there is a next action.

Sometimes it can respond automatically.

Sometimes it should prepare the response for a person.

Sometimes its most useful action is:

THIS ONE NEEDS YOU.

That's the difference between putting AI on your website and putting AI into your sales process.

Quick answers

The enquiry has arrived. What happens after that?

AI can also help determine whether an enquiry is ready for a sales conversation and what information is still missing.

Can AI qualify sales leads? (coming soon)

Or see what happens later in the process:

Can AI follow up sales leads automatically? (coming soon)

In the meantime, start with the bigger picture: