Your Next Salesperson Could Be An AI Agent. Who Is Training It?

Your Next Salesperson Could Be an AI Agent. Who Is Training It?

AI agents can answer questions, recommend products, qualify leads and book appointments. Before one speaks for your business, it needs the same clarity, boundaries and supervision as any other member of your team.

Imagine hiring a salesperson who can answer every enquiry instantly, work through the night, recommend products, qualify leads and book appointments.

Useful.

Now imagine that same salesperson confidently quoting last year’s price, recommending the wrong service, making a promise your delivery team can’t keep, and refusing to call in a human when the conversation becomes complicated.

Still useful?

That’s the decision many businesses now face as AI agents move from internal productivity tools to customer-facing roles.

The question is no longer simply whether AI can help your team work faster. It’s whether you’re ready to let it speak for your business.

AI agents are increasingly moving into customer conversations

In June 2026, Meta announced the expansion of its Meta Business Agent – an AI agent designed to represent businesses across WhatsApp, Messenger and Instagram.

Meta says more than one million businesses are already using a Business Agent on WhatsApp and Messenger. The agent can answer business-specific questions, recommend products from a catalogue, book appointments, qualify incoming leads, allow a human team member to step in, and in some circumstances, close sales.

Google is heading in the same direction. It has announced Business Agent for Leads – a forthcoming format that places an AI-powered brand agent inside a Search advert. Instead of completing a static form, the customer can start a conversation and receive answers based on information from the advertiser’s website. Google describes the format as part of its next generation of Search advertising rather than something already available to every advertiser.

This is more than a slightly cleverer chatbot.

An AI agent can become part receptionist, part salesperson, and part customer-service assistant. It can answer questions and, depending on the system and permissions, take action.

For a small business, that’s a clear attraction: more customers can receive immediate help without the business needing somebody online every hour of the day.

But a salesperson who never sleeps can also repeat the wrong answer at scale.

Why should an SME care?

The opportunity isn’t difficult to see. An AI sales agent could:

• Answer routine questions outside normal working hours
• Direct customers towards the right service
• Arrange calls or appointments
• Collect the information your sales team needs
• Filter obviously unsuitable enquiries
• Help customers compare products
• Keep straightforward conversations moving when your team is busy

For a small team, that could remove a significant amount of repetitive work.

The risk is that marketing, sales and customer service are being compressed into one automated conversation. A customer may discover the business, ask questions, compare options, discuss pricing and decide whether to proceed – all without speaking to a person.

If the agent handles that conversation well, it makes the business feel responsive and straightforward to deal with.

If it handles it badly, the customer may never reach your human team.

An AI agent can scale your best sales conversation. It can also scale your weakest explanation.

Your AI agent can only work with what your business has told it

Google says its Business Agent for Leads will answer questions based on the advertiser’s website. Meta says its agent can use business-specific information and product catalogues to answer questions and make recommendations.

That makes the quality of your existing marketing material particularly important.

Suppose your website says your service starts from £500. Your current proposal template says £750. Your sales team says every project is individually priced. Your social media still promotes a package you stopped selling last year.

Which answer should the agent provide?

It won’t reconcile the disagreement through years of commercial experience and a quick chat with Sandra in sales. It will work with the information, rules and systems it’s been given. If your source material is inconsistent, the agent may produce a polished version of the inconsistency.

Before introducing a customer-facing agent, make sure you have a reliable set of marketing collateral that supports sales conversations, including current service descriptions, approved pricing, exclusions, policies and answers to common questions.

Where can an AI sales agent go wrong?

The technology doesn’t need to become dramatically rebellious to create a problem. Small inaccuracies and badly designed handovers are enough.

Answer pricing questions

It may quote an old price or leave out an important condition. The business should have one approved, dated source for pricing and exclusions.

Recommend a product or service

It may recommend the most obvious option rather than the best fit. The business needs clear qualification questions, suitability rules and known limitations.

Qualify leads

It could dismiss a valuable prospect because their situation doesn’t fit a simple rule. The business needs flexible criteria and a route for human review.

Book appointments

It may offer unavailable times or book the wrong type of meeting. The business needs live calendar information and clear appointment categories.

Handle complaints

It could keep generating sympathetic responses instead of resolving the problem. The business needs immediate escalation rules and a named human owner.

Explain policies

It may confidently interpret a policy in a way the business never intended. The business needs approved wording and clear limits on what the agent may decide.

Represent your tone of voice

It might sound cheerful when the customer is angry or informal during a serious issue. The business needs tone guidance that changes according to the context.

The agent doesn’t need to be perfect. It does need to understand the limits of its job.

That’s true of human employees too. The difference is that a human team member will often recognise the moment a conversation has gone off-script. An AI agent may continue with impressive confidence.

Treat your AI agent like a new employee

Most businesses wouldn’t put a new salesperson in front of customers on day one with nothing more than access to the website.

They’d explain:

• What the business sells
• Who it’s for
• How pricing works
• What’s included
• What can be promised
• What must never be promised
• When a manager needs to become involved
• How the business expects customers to be treated

Your AI agent needs the same induction. It won’t need a welcome lunch or a tour of the fire exits, but the rest still applies.

The STORY22 AI Agent Induction Checklist

1. Give the agent a clear job description

Start by deciding what the agent is actually there to do.

“Help customers” is far too broad. A clearer job might be:

• Answer routine questions about three defined services
• Help customers choose the correct appointment
• Collect five pieces of information before a sales call
• Recommend products from an approved catalogue
• Explain the returns process without making exceptions

Define what success means. Is the objective to reduce response time, increase booked calls, improve lead quality or remove repetitive enquiries from the team? Without a clear job, the agent will be judged on activity rather than value.

2. Create one source of truth

Bring together the information the agent is allowed to use. This may include:

• Current product and service descriptions
• Pricing and what affects it
• Package inclusions and availability
• Delivery times
• Refund and cancellation policies
• Eligibility rules
• Common customer questions and known limitations
• Approved case studies and proof

Give every document an owner and review date. A folder full of old PDFs is not a knowledge base. It’s an archaeological site.

3. Give it a clear, approved message

Your AI agent needs more than factual information. It needs to understand:

• Who the customer is
• What problem they’re trying to solve
• How your business helps
• What makes your approach different
• Which claims you can prove
• What action the customer should take next

A clear message also helps prevent the agent overwhelming customers with every feature your business has ever offered.

STORY22’s StoryBrand messaging and BrandScript service helps businesses create a central customer-focused message that can be used consistently across websites, campaigns, sales materials and customer conversations.

4. Define permissions and red lines

Write down what the agent may and may not do.

The agent may:

• Explain published prices
• Book a standard introductory call
• Recommend from an approved product list
• Ask basic qualification questions
• Share approved resources

The agent must not:

• Negotiate a price
• Promise a result
• Give legal, medical, financial or regulated advice
• Make an exception to a policy
• Confirm something that depends on a human assessment
• Respond to a serious complaint without escalation

Don’t rely on the system to infer your boundaries. If something matters, state it.

5. Design the human handover before launch

Meta says businesses can decide when a team member should step into a Business Agent conversation. That’s not a minor feature. It’s a core part of the customer experience.

A handover should happen when:

• The customer asks for a person
• The agent lacks current information
• The customer is unhappy
• The question involves a complaint
• The conversation involves a regulated or sensitive issue
• The customer wants to negotiate
• The agent has failed to answer twice
• The value or complexity of the opportunity crosses an agreed threshold

The customer shouldn’t have to repeat the entire conversation after the handover. The human team member needs the context, the question and the information already collected.

Otherwise, you haven’t created a smooth transfer. You’ve created a digital obstacle course.

6. Test it like a difficult customer

Don’t test only the easy questions. Ask:

• Can you guarantee this will work?
• Can you make an exception?
• Your website says one price, but I was told another. Which is correct?
• Which package should I buy when I’m unsure what I need?
• I want to cancel. What happens now?
• I’m unhappy. Can I speak to someone?
• Is your competitor better for my situation?
• What information are you storing about me?

Test spelling mistakes, vague questions, irritated customers and unusual situations. Your customers will.

7. Assign a human owner

Someone in the business needs to be responsible for:

• Reviewing conversations
• Correcting inaccurate answers
• Keeping source information up to date
• Checking failed handovers
• Monitoring complaints
• Reviewing lead quality
• Deciding when the agent’s role should expand

“Marketing, sales and customer service all own it” sounds collaborative. In practice, it can mean nobody notices that the agent has been quoting an expired offer for six weeks.

Give someone clear ownership.

Speed alone is not good service

A fast response is useful only when it helps the customer move forward.

The agent should reduce uncertainty for the customer, remove unnecessary effort and make it easier to reach the right person. It should not be used as a wall between customers and your team.

That distinction matters because the first interaction tells the customer what kind of business they’re dealing with.

Does the agent:

• Answer clearly?
• Recognise the customer’s actual question?
• Admit when it doesn’t know?
• Hand over quickly when human judgement is needed?
• Preserve the conversation so the customer doesn’t repeat themselves?
• Treat a complaint differently from a routine enquiry?

At STORY22, we treat customer experience as something that needs to be designed, not left to chance. Our Unreasonable Hospitality coaching helps organisations identify the customer moments that matter, create clearer handovers and turn routine interactions into experiences that build trust and loyalty.

AI can handle routine interactions. The human team should use the time it saves to deal more thoughtfully with the moments that need judgement, empathy or creativity.

What about customer data?

A customer-facing AI agent may process names, contact details, questions, purchase information or other personal data.

The Information Commissioner’s Office says organisations using AI to process personal data need to identify an appropriate lawful basis and provide transparent information about why the data is being used, how long it will be kept and who it may be shared with. Current ICO guidance is being reviewed following the Data (Use and Access) Act, so businesses should check the latest position before deployment.

This doesn’t mean every small business needs to become an AI lawyer. It does mean privacy can’t be left until the agent is already talking to customers.

Before launch, confirm:

• What information the agent collects
• Why it collects it
• Where it is stored
• Who can access it
• Whether it is shared with suppliers
• How long it is retained
• How the customer is informed

Obtain appropriate legal or data-protection advice where the use is complex, regulated or high risk.

Start with one clearly defined job

There’s no prize for connecting an AI agent to every part of the business on day one.

Begin with:

• One channel
• One clear task
• One approved information source
• One named human owner
• One defined handover process
• One short review period

For example, allow the agent to answer common questions and book introductory calls. Review what customers ask, where the agent struggles and whether the booked calls are suitable. Then decide whether to expand its role.

A limited agent that performs one job reliably is more valuable than an impressive demonstration that creates uncertainty across the whole customer journey.

Your next salesperson still needs human oversight

AI agents are moving into the places where customers discover, compare and choose businesses. That creates a genuine opportunity for SMEs. A well-designed agent can make a small team feel more available, help customers receive answers sooner and reduce repetitive work.

But turning one on isn’t the same as preparing it to represent your business.

Your next salesperson may not need a desk or salary review, or a company mug. It still needs:

• A clear job
• Accurate information
• An approved message
• Defined boundaries
• A human handover
• Regular supervision

The technology may be new. The management responsibility is not.

Prepare your business before the agent speaks on its behalf

Before you add an AI agent to your website, advertising or messaging channels, make sure the agent has a clear message worth repeating and a customer journey it can support.

STORY22 can help you align your brand message, sales collateral and customer experience so your human and AI representatives give customers the same clear, credible answer.

Book a free 30-minute strategy call with STORY22 to review what an AI agent would need to know, where it could add value and which conversations should remain human.

For a lower-commitment next step, join the Mastering AI for Business programme waitlist for practical guidance on using AI in your marketing and wider business processes.

AI Sales Agents: Your Questions Answered

What is an AI sales agent?

An AI sales agent is software that can hold customer conversations using information about your business. Depending on the system and its permissions, it may answer questions, recommend products, collect lead information, book appointments or take other actions. Unlike a basic FAQ chatbot, an agent may use the context of the conversation and information from connected business systems rather than relying only on a fixed list of pre-written replies.

Is Meta Business Agent available to every UK business?

Not necessarily. Meta announced a global expansion in June 2026, but said the rollout was beginning with a selected number of businesses using the WhatsApp Business app, Instagram Pro, Messenger and Meta Business Suite. Meta also said that getting started was free at launch, with paid subscription options planned for the future. Availability and pricing should be checked directly before making plans.

Can an AI agent really close a sale?

Meta says Business Agent can close sales, as well as recommending products, booking appointments and qualifying leads. What that means in practice will depend on the business, the products being sold, payment integrations, catalogue information and permissions given to the agent. A complex professional service is unlikely to have the same automated sales process as a straightforward retail product.

What is Google Business Agent for Leads?

Google Business Agent for Leads is a planned AI-powered feature within Search advertising. A person will be able to press “Chat” inside an advert and receive answers based on information from the advertiser’s website. Google presented it as a forthcoming format rather than a universally available product.

Should customers be told they are speaking to AI?

Clear disclosure is sensible because it helps customers understand the nature and limits of the interaction. Where personal data is being processed, the business must also provide appropriate privacy information. The ICO says this should include the purpose of the processing, retention periods and who the data will be shared with. The exact legal requirements depend on the use, data and circumstances, so obtain advice where needed.

What conversations should always go to a human?

Handover should normally be considered when:

• A customer asks for a person
• The issue involves a complaint
• The agent lacks reliable information
• The conversation requires negotiation
• A regulated or sensitive judgement is needed
• The customer is distressed or vulnerable
• The agent has failed to answer clearly
• The potential opportunity requires individual assessment

The goal isn’t to keep humans out of the conversation. It’s to use human time where it adds the most value.

How should a small business begin using an AI agent?

Start with one narrow, repeatable job. Routine questions, initial lead qualification or appointment booking are usually more manageable starting points than allowing an agent to advise on every service. Give the agent one approved source of information, test it thoroughly and assign a named person to review the conversations.

Compliance note: This article provides general marketing and customer-experience commentary. It does not constitute legal, data-protection, financial or regulatory advice. AI systems, platform features and associated rules are changing quickly. Businesses should check the latest platform documentation and obtain appropriate professional advice before using AI for regulated, sensitive or high-risk customer interactions.

About STORY22

STORY22 helps organisations make their marketing clearer and their customer experience more coherent.

As certified StoryBrand guides, we help businesses develop a customer-focused message and apply it across websites, campaigns, sales collateral and customer conversations. Through our customer-experience work, we also help teams design thoughtful, consistent interactions that turn the brand promise into something customers can genuinely feel.

The combination matters because customers should receive the same clear message whether they speak to a salesperson, read a proposal, visit the website or interact with an AI agent.

Schedule a free 30-minute strategy call and make sure every representative of your business – human or otherwise – knows what good looks like.