AI gives smaller businesses access to capabilities that once required larger teams, specialist departments and significant technology budgets.

That is a genuine opportunity. It does not mean every business should automate every process or build its own AI product.

The advantage comes from combining the technology with things that are harder for competitors to copy:

  • customer knowledge
  • operating experience
  • specialist expertise
  • local context
  • service standards
  • brand language
  • commercial judgement

I was reminded of this at an excellent ModelProp event about the use of AI in property, presented by Mal McCallion.

The demonstrations were practical. The most interesting part, however, was the conversation among the property businesses attending.

People were willing to experiment. They could see the potential, but they were less certain about what they should build and how they should use it.

The discussion covered questions such as:

  • Could it write property blogs?
  • Could it analyse enquiries and recommend the next action?
  • Should an agency wait for its existing software supplier to introduce AI?
  • Would something tailored to the business create a better experience?

There is no universal answer. Each use involves a different balance between speed, cost, control, consistency, risk and customer experience.

The right starting question is not:

Where can we add AI?

It is:

Which customer or operational problem are we trying to solve?

How should a small business use AI?

Start with a recognised problem.

Use AI where it can:

  • reduce avoidable effort
  • make useful knowledge easier to access
  • improve consistency
  • speed up a defined task
  • help someone make a better-informed decision
  • remove friction from a customer journey

Keep human judgement where decisions affect trust, risk, personal data, financial commitments or the customer relationship.

That sounds obvious, but it is easy to reverse the sequence. A business sees a new capability, becomes excited about what it can do and then starts searching for somewhere to use it.

A better sequence is:

  1. Understand the problem.
  2. Define the desired outcome.
  3. Decide what needs to change.
  4. Identify where technology can help.
  5. Agree where human judgement remains essential.
  6. Review whether the change improved the result.

AI should support a useful system. It should not become the reason for creating one.

Why AI gives smaller businesses access to scale

AI changes the cost and pace of producing a first version.

A smaller team can now use it to:

  • analyse large quantities of information
  • compare several versions of a plan
  • create first drafts
  • organise internal knowledge
  • develop simple assistants
  • identify patterns in enquiries
  • improve access to guidance
  • test an idea before making a larger investment

Work that might once have needed an agency, analyst, developer or specialist department can at least be explored by a much smaller business.

The UK Government's June 2026 AI Adoption Plan for professional and business services reported that 43.4% of firms in the sector were using AI in December 2025, up from 31.4% a year earlier. It also identified limited internal expertise, implementation costs and concerns about safety and transparency as continuing barriers.

That combination is understandable. Businesses can see the opportunity. Many are still working out how to turn experimentation into something dependable and operationally useful.

The important distinction is between access to capability and quality of outcome. AI can make the first version faster and cheaper. It does not remove the work required to make that version correct, relevant or useful.

Why generic AI creates generic customer experiences

Most businesses can access similar AI tools. They can use similar prompts, templates and software. If they provide the same limited context, they are likely to receive broadly similar outputs.

That creates familiar patterns:

  • similar articles
  • similar email sequences
  • similar property descriptions
  • similar chatbots
  • similar advice
  • similar claims of personalisation

The technology may be impressive, but the experience becomes less distinctive. This is where business knowledge matters.

The most valuable inputs are often not dramatic secrets. They are the practical things an experienced business understands:

  • what customers repeatedly ask
  • where conversations tend to stall
  • what creates confidence
  • what causes uncertainty
  • which enquiries need an urgent response
  • when a standard answer is enough
  • when somebody needs to pick up the telephone
  • what a good next step looks like

That knowledge is difficult for a competitor to copy because it has been accumulated through experience.

In Your Marketing Isn't Broken. Your Inputs Are, I argued that poor outputs are often caused by unclear inputs across marketing, CRM and customer journeys. The same principle applies to AI: better tools do not compensate for vague decisions, weak context or uncertain priorities.

The AI does not know:

  • what makes your service useful unless you explain it
  • which compromises your team can realistically manage
  • the standards you want to protect
  • what you are unwilling to automate

Those are business decisions.

Should you use an off-the-shelf or bespoke AI product?

There are several sensible ways to introduce AI into a business.

Wait for an existing supplier

A CRM, property platform or other technology provider may add AI functionality to a product you already use.

This may offer:

  • easier implementation
  • existing integrations
  • familiar support
  • lower technical risk
  • less internal development

The limitation is that competitors using the same supplier may receive the same capability.

Use a general-purpose AI tool

This is often the quickest way to experiment.

A business can use general-purpose tools to:

  • explore ideas
  • draft content
  • analyse information
  • summarise documents
  • test processes
  • build simple internal guidance

The quality depends heavily on the information, constraints and examples provided.

Buy a specialist product

A defined product may already solve a recognised sector problem.

ModelProp, for example, presents both ready-to-use property AI products and more customised applications for estate and letting agencies.

A specialist product may offer useful sector knowledge without requiring the business to start from scratch.

Configure or build something tailored

A more tailored approach may make sense where the business's own knowledge or customer experience creates meaningful value.

This does not always require a large software development project. It may involve applying existing technology using your own:

  • knowledge base
  • customer language
  • examples
  • business rules
  • service standards
  • escalation points
  • decision criteria

None of these routes is automatically better.

Where would something specific to our business create a better result?

Not every process needs to become a competitive advantage. A standard product may be entirely appropriate for a standard task.

Tailoring matters most where it affects the customer, the commercial decision or the knowledge that makes the business distinctive.

What could this mean for a property business?

Property provides plenty of practical AI use cases.

An estate or letting agency might use AI to support:

  • after-hours call handling
  • first drafts of property particulars
  • local area information
  • applicant matching
  • landlord communication
  • valuation enquiry triage
  • internal knowledge retrieval
  • post-valuation follow-up
  • identifying enquiries without a next action
  • management reporting

The relevant question is not simply whether AI can perform the task. The business should examine how the task fits into the wider customer journey.

Take a valuation enquiry. A useful system may need to:

  1. Capture the enquiry and its source.
  2. Recognise the likely intent and urgency.
  3. Retain the context supplied by the customer.
  4. Notify the right person quickly.
  5. Provide a relevant first response.
  6. Agree and record the next action.
  7. Alert somebody when progress stops.
  8. Move to a human conversation when judgement is required.

AI might support several of those steps. It does not automatically create clear responsibility, a good response or consistent follow-up. Those elements still need to be designed and managed.

Do not automate something you do not understand

Before automating part of a process, inspect what happens now.

Ask:

  • What does the customer need at this point?
  • What information is already available?
  • What information is usually missing?
  • Who is responsible for what happens next?
  • Where does the process slow down?
  • Where is context lost?
  • Which decisions require experience?
  • How will we know whether the change helped?

Otherwise, AI may simply help an unclear process move faster.

Examples include:

  • A generic reply is still generic when it arrives instantly.
  • An enquiry without a clear owner remains unowned after it has been classified.
  • A completed CRM task does not mean the customer has moved forward.
  • A chatbot can add another handoff if the customer must repeat everything to an employee.
  • Faster content production is not useful if nobody has agreed what the content needs to achieve.

Technology can support the customer journey. It cannot define that journey on the business's behalf.

This is why AI adoption is not only a technology question. It is also a question of customer need, process, responsibility, information and management visibility.

How do you get better answers from AI?

Do not always accept the first answer.

AI is designed to provide a response. That does not mean its first response is the right recommendation.

The first answer may be:

  • based on incomplete context
  • technically plausible but operationally unrealistic
  • too generic
  • built on hidden assumptions
  • biased towards producing something rather than questioning whether it should be produced

Challenge it. Useful follow-up questions include:

  • What assumptions have you made?
  • Which part of this recommendation is least certain?
  • What information is missing?
  • What evidence would change your answer?
  • What could go wrong operationally?
  • How might a customer experience this?
  • What are the trade-offs?
  • What should remain a human decision?
  • What would make this difficult for a small team to implement?

Then add more context and ask again.

You can also ask the AI to review an idea from several professional perspectives:

  • What might a Finance Director question?
  • What could a Marketing Director consider unclear?
  • What might an operational manager struggle to implement?
  • What could a customer misunderstand?
  • What might a data protection specialist want checked?
  • What might a subject-matter expert consider too simplistic?

This does not replace those specialists. It provides a light-touch way to identify questions before asking people to invest time, money or customer trust.

The AI can help you examine an idea. You still decide which criticism is valid and what should happen next.

Where should human judgement remain?

Human review is already a normal part of business AI use.

Government AI Adoption Research published in 2026 found that 84% of businesses using AI applied at least some human input or checking to its outputs. Around two-thirds reported significant checking.

The level of review should reflect the importance of the task.

Greater care is likely to be required where an AI system affects:

  • customer trust
  • personal data
  • pricing
  • financial commitments
  • legal rights
  • employment
  • reputation
  • regulated advice
  • significant commercial decisions

Where personal data is processed, businesses also need to understand how data protection requirements apply. The ICO's guidance on AI and data protection covers both legal interpretation and good practice for AI systems using personal data.

Human involvement should not be treated as an admission that the technology has failed. It is part of responsible system design.

The aim is to be clear about:

  • what the system is allowed to do
  • what it must not do
  • when somebody needs to intervene
  • who remains accountable
  • how mistakes will be identified and corrected

Seven questions to ask before introducing AI

1. What recognised problem are we solving?

Avoid beginning with the tool. Define what is currently slow, inconsistent, expensive, unclear or frustrating.

2. Who benefits?

Is the primary benefit for the customer, employee, manager or supplier? A useful change may benefit several groups, but the intended outcome should be clear.

3. What does our business know that the tool does not?

Identify the customer knowledge, examples, rules, standards and experience the system will need.

4. What should remain a human decision?

Be clear about where judgement, empathy, accountability or professional expertise is required.

5. What could go wrong?

Consider inaccurate information, inappropriate tone, lost customer context, weak handoffs, privacy, bias, overconfidence and actions taken without sufficient review.

6. Who owns the result?

Someone must remain responsible for the customer experience and commercial outcome. "We used AI" is not an ownership model.

7. How will we know it is helping?

Measure something more useful than the amount of content produced or the number of tasks automated.

Useful measures might include:

  • Have response times improved?
  • Are customers receiving clearer answers?
  • Has avoidable work reduced?
  • Are employees saving useful time?
  • Are more enquiries progressing?
  • Are fewer opportunities disappearing without a next action?

The opportunity is not more output

AI gives smaller businesses access to capabilities that were previously expensive or difficult to obtain.

That is exciting. The lasting advantage, however, will not come from producing the most content or automating the largest number of tasks.

It will come from knowing:

  • where technology genuinely improves the experience
  • where your knowledge makes the result more relevant
  • where a standard product is sufficient
  • where something more tailored creates value
  • where human judgement remains essential

AI supplies pace and productive capacity. Your business supplies the context. Judgement turns the two into something useful.

A practical next step

Choose one customer journey or internal process where AI is being considered.

Do not begin by listing tools. Map what happens now:

  • What triggers the process?
  • What does the customer or employee need?
  • Where does it slow down?
  • Where is information lost?
  • Who is responsible?
  • What requires judgement?
  • What would a better outcome look like?

Then decide whether AI is the right way to improve it. That is more useful than introducing technology first and looking for a problem afterwards.

References

FAQs

How can a small business start using AI?

Choose one recognised problem rather than trying to introduce AI across the whole business. Define what happens now, what a better result looks like and where human review will remain.

Should a small business build a bespoke AI tool?

Only where business-specific knowledge or customer experience creates meaningful value. An off-the-shelf product may be the better choice for a standard process.

How can a business get better results from AI?

Provide clear context, relevant examples and realistic constraints. Challenge the first answer, ask what assumptions have been made and review the recommendation from several perspectives.

What should not be automated using AI?

Be cautious where decisions affect customer trust, personal data, legal rights, financial commitments or reputation. These uses may require stronger controls, specialist advice and meaningful human oversight.

Will using the same AI tools as competitors make a business less distinctive?

Not automatically. The difference comes from how the tools are applied. Customer knowledge, specialist experience, brand language and service standards can make a common technology produce a more relevant experience.