Where AI Creates the Most Value in a Small Business

AI for SMEs
Written By thomas czarnecki

The best AI opportunities for an SME are not necessarily the most sophisticated. Start with how your business actually operates: where people are stretched, context gets lost, demand changes, knowledge is concentrated and valuable human time is consumed by work around the work.

A small business is not a miniature enterprise.

That matters when deciding where AI actually makes sense.

In a smaller organisation, one person may perform several roles. Important knowledge may be concentrated in a handful of experienced employees. Part-time, casual or seasonal staff may move through roles. Demand may change significantly during the year.

There may be no data team, AI lead or transformation office.

And the personal interaction that looks inefficient on a process diagram may be one of the reasons customers choose the business.

These characteristics mean an SME should not simply take an enterprise AI strategy and reduce its scale.

A better approach is to start with the business itself.

Ask:

Where does the way our organisation operates create an opportunity for AI to give us meaningful leverage?

The answer may be less dramatic than automating an entire function.

It may also be much more valuable.

Start with leverage, not AI

It is easy to start AI adoption with technology questions:

  • Which AI should we buy?

  • What can we automate?

  • Which new capability should we try?

Those questions matter later.

First identify the operating constraint.

For a smaller organisation, valuable AI leverage might mean:

  • giving an owner several hours back each week;

  • helping a seasonal employee become productive faster;

  • improving continuity between people who share a role;

  • reducing interruptions to the employee everybody depends on;

  • helping a manager resume complex work after being pulled elsewhere;

  • removing administration around an important customer interaction.

None requires the most sophisticated AI available.

The objective is not maximum AI.

It is maximum useful leverage from an appropriate use of AI.

Opportunity 1: People performing several roles

Small organisations frequently depend on people wearing several hats.

An owner might move between sales, finance, recruitment and operations in a single day. A founder might move between product strategy, development, testing and marketing.

The cost is not simply workload.

It is context switching.

When somebody changes roles, they have to reconstruct part of the previous context:

  • What was I working on?

  • What had we decided?

  • Which documents matter?

  • What remains unresolved?

  • What should happen next?

AI can potentially reduce this friction by helping preserve context around a role.

Instead of treating AI only as a generic assistant, consider a role-specific capability that works with the information relevant to that role.

A marketing co-pilot, for example, might work with:

  • the current marketing plan;

  • approved product information;

  • publishing standards;

  • previous decisions;

  • active campaigns;

  • work in progress.

The person can move to another responsibility and return later asking:

What were we up to?

The value is not that AI turns one employee into several specialists.

It is that the organisation may spend less time reconstructing context every time a scarce person changes roles.

For a small business, AI can create value by helping role context persist.

Opportunity 2: Shared, part-time and seasonal roles

The same principle applies when the person changes rather than the role.

Many SMEs rely on:

  • job sharing;

  • part-time employees;

  • casual workers;

  • seasonal staff;

  • rotating responsibilities.

Continuity often depends on handover notes, procedures, conversations and memory.

Consider a simple model:

PERSON → AI-SUPPORTED CONTEXT → ROLE

The person can change while the role continues.

An appropriate AI capability could potentially help preserve:

  • objectives;

  • relevant procedures;

  • current priorities;

  • work in progress;

  • previous decisions;

  • organisational knowledge;

  • unresolved issues.

Someone beginning a Thursday shift could ask:

What happened earlier this week, what's still open and what needs attention today?

A seasonal worker might ask how the organisation normally handles an unusual customer situation.

The important point is that the answer should not merely sound plausible.

Useful organisational AI needs access to the appropriate context for that role and situation.

That can reduce a significant but easily overlooked SME cost:

organisational discontinuity.

Opportunity 3: Capacity that changes with demand

Some businesses operate very differently at different times of the year.

Tourism, hospitality, agriculture, events, retail and seasonal production may move through something like:

QUIET → PREPARE → RAMP UP → PEAK → CONTRACT

Permanent staffing cannot always economically expand and contract at the same speed.

This creates another useful place to investigate AI.

Ask which work becomes a bottleneck when demand rises:

  • customer enquiries;

  • routine documentation;

  • internal questions from new employees;

  • booking support;

  • administrative processing;

  • research;

  • information retrieval;

  • operational coordination.

AI may provide additional capacity around some of that work.

That does not necessarily mean replacing seasonal employees.

Often the stronger opportunity is to help temporary employees become useful faster while reducing the amount of repetitive support required from the permanent team.

A useful question is:

Where does demand vary faster than our permanent capacity can economically vary with it?

Opportunity 4: Knowledge concentrated in a few people

Most smaller businesses have somebody who knows things that are difficult to find anywhere else.

They know:

  • which supplier to call;

  • why a customer arrangement exists;

  • how an unusual process works;

  • what happened last time;

  • which exception was approved;

  • what the organisation normally does when the documented procedure does not quite fit.

Ask what would happen if that employee were unavailable for a month.

The answer can reveal valuable AI opportunities.

AI may help make appropriate organisational knowledge easier to retrieve, explain and transfer.

But there is an important qualification.

Making knowledge easier to access should not mean making all knowledge accessible to everyone.

A casual employee may need operating procedures and supplier information. They probably do not need payroll information, employee records or confidential strategy.

So the opportunity is not simply:

Put everything into AI.

It is:

Can the right person obtain the right organisational knowledge for the role and situation they are in?

That distinction becomes important as organisations build AI around their internal knowledge.

Opportunity 5: Protect valuable human interaction

Not every expensive human activity is an inefficiency.

This is particularly important for smaller professional, regional, rural and relationship-led businesses.

A local accountant may understand the history behind a family business.

An accommodation owner may remember returning guests.

A rural supplier may understand the customer's operation.

A professional adviser may be valuable because the client trusts their judgement.

Viewed purely as labour, these interactions may appear expensive.

But the human relationship may be part of the product.

That gives SMEs a useful test:

Is the human interaction a cost around the service, or part of the value of the service?

If it is part of the value, automating it may make the business more efficient while simultaneously making the service less valuable.

AI can still help.

It might:

  • prepare the employee for the conversation;

  • retrieve relevant history;

  • identify unresolved issues;

  • prepare options;

  • summarise information;

  • draft follow-up;

  • handle administration around the interaction.

Instead of asking:

Can AI automate this customer interaction?

ask:

Can AI remove work around this interaction so our people have more time for the part customers actually value?

Opportunity 6: Context that repeatedly gets lost

AI adoption discussions naturally focus on tasks.

What can we automate?

What can we generate?

What can we make faster?

Those are useful questions.

But SMEs should also look for lost context.

Look at:

  • handovers;

  • role switching;

  • seasonal onboarding;

  • work interrupted by urgent problems;

  • the employee everybody has to ask;

  • job-share arrangements;

  • people returning to work after several days away;

  • staff searching repeatedly for the same organisational information.

The opportunity may not be for AI to perform the work itself.

It may be to help the context of the work survive changes in people, shifts and responsibilities.

That is a different way to think about AI productivity.

Look for these operating signals

Rather than beginning with a catalogue of AI products, examine your organisation for signals such as:

  • scarce people repeatedly switching roles;

  • important context constantly being reconstructed;

  • roles shared between several people;

  • temporary employees who need to become productive quickly;

  • demand changing faster than permanent staffing capacity;

  • organisational knowledge concentrated in a few people;

  • repetitive knowledge work consuming expensive human time;

  • employees spending significant time finding and preparing information;

  • valuable customer-facing people burdened by administration;

  • work stopping because the person with the answer is unavailable.

These are not automatically AI use cases.

They are places worth investigating.

The next question is whether AI can improve the situation in a way that is valuable, understandable and controllable.

Assistance, influence and action require different thinking

Not every AI use gives the system the same degree of responsibility.

A useful distinction is:

ASSIST → INFLUENCE → ACT

Assist

AI drafts, summarises, retrieves, organises, compares or prepares information.

A person remains responsible for interpreting the output and taking the action.

Influence

AI analysis or recommendations materially affect a human decision.

A person may still make the decision, but the AI has become part of the decision process.

Act

AI performs an action: updating a system, sending a communication, triggering a workflow, making a booking or carrying out another operation.

Moving from assistance towards action usually increases the importance of:

  • authority;

  • information access;

  • reliability;

  • human oversight;

  • consequences of error;

  • evidence about what occurred.

The same distinction applies to information.

AI drafting ideas from public material is different from AI working with identifiable customer records.

AI suggesting available appointments is different from cancelling bookings.

AI preparing a recommendation is different from making a consequential decision.

The relevant question is therefore not simply:

Is AI safe?

It is:

Is this particular use of AI—with this information, purpose, consequence and level of authority—appropriate for our organisation?

A 10-question test for an SME AI use case

Before committing substantial time or money, work through these questions.

1. Value
If the use case succeeds, will it materially improve something the business cares about?

2. Frequency or variability
Does the work happen frequently enough to matter, or become a serious bottleneck during peaks?

3. Capacity
Does it consume time from people who are scarce, overloaded or performing several roles?

4. Context
Does work lose momentum because people repeatedly reconstruct history, hand over work or search for organisational knowledge?

5. Boundaries
Can you clearly describe what the AI should and should not do?

6. Verifiability
Can a competent person reasonably determine whether the output or outcome is correct?

7. Information
What customer, employee, confidential or proprietary information would the AI need?

8. Consequence
What happens when the AI is wrong?

9. Authority
Is the AI assisting, influencing a decision or taking action?

10. Human value
Is the human activity being considered for automation actually part of what customers, employees or owners value?

There is no universal score that turns these questions into an automatic decision.

Their purpose is to make the trade-offs visible before investment and implementation.

Where SMEs should be cautious about starting

Some use cases make poor first experiments.

Be cautious where:

  • nobody can define good performance;

  • errors could create serious consequences;

  • outputs are difficult for a competent person to verify;

  • broad access to sensitive information is required before controls are understood;

  • significant autonomous authority is required immediately;

  • the underlying business process is poorly understood;

  • the activity occurs too rarely to justify the implementation effort;

  • automation would remove a human relationship or judgement that customers genuinely value.

An organisation does not need its most ambitious AI use case to learn something valuable.

A bounded first use case can teach you:

  • whether employees find AI useful;

  • where information issues appear;

  • what human review is required;

  • which assumptions were wrong;

  • how much value is actually created;

  • what governance the organisation needs before expanding.

What good SME AI adoption looks like

The objective is not to maximise the number of AI systems inside the business.

It is to find places where AI creates meaningful operating leverage without introducing disproportionate complexity or risk.

For an SME, that often means starting where:

**VALUE IS MATERIAL

  • WORK IS UNDERSTANDABLE

  • BOUNDARIES ARE CLEAR

  • OUTCOMES CAN BE CHECKED

  • AUTHORITY IS APPROPRIATE**

Then learn before expanding.

This is also consistent with Agorik's approach to governed intelligence.

Agorik starts from the premise that useful AI depends on understanding the organisation in which it operates: its people, roles, information, systems, responsibilities and governance.

The question is not simply which AI capability exists.

It is whether that capability fits the organisation, the work and the authority it has been given.

That becomes increasingly important as AI moves from helping people create information towards influencing decisions and taking actions.

Start with the business you actually have

SMEs do not need an enterprise AI transformation programme before they can benefit from AI.

They do not need to automate everything.

And they do not need to introduce AI into every customer interaction simply because the technology makes it possible.

Start with the characteristics of your organisation.

Where are scarce people stretched across several roles?

Where does context disappear?

Where does seasonal demand create pressure?

Where is important knowledge concentrated?

Where does administration consume time that could be spent with customers?

Where could AI strengthen a valuable human interaction rather than replace it?

Where is the work bounded enough that you can tell whether AI is helping?

Those questions are less exciting than asking what the latest AI can do.

They are considerably more useful.

A small business is not a miniature enterprise. Its AI strategy should not be a miniature enterprise AI strategy either.

Next step

Use the 10-question test above to assess one AI opportunity in your organisation before evaluating products or vendors.

Explore Agorik's approach to governed organisational intelligence, or subscribe to Governed Intelligence for practical perspectives on how SMEs can understand, adopt and govern AI.

thomas czarnecki
Previous

What Happens to Your Business Information When You Use AI?
Next

Why AI Governance Needs to Work Differently