AI for SMEs: The Leadership Decisions That Matter Most

AI is not difficult to buy.

Any business owner can approve another subscription, give the team access to a generative AI tool or ask a manager to investigate automation.

The difficult part is turning that activity into measurable business value.

AI use among UK businesses with ten or more employees increased from approximately 12% in late 2023 to approximately 35% in 2026. However, adoption remains relatively shallow: the average number of AI technologies used by adopting businesses has increased only modestly.

The technology is moving quickly. Organisational capability is moving more slowly.

That is why AI adoption for SMEs is fundamentally a leadership challenge.

The businesses that benefit will not necessarily be those with the largest technology budgets. They will be the ones that make clear decisions about where AI belongs, who owns the result, how performance will be measured and where human judgement remains essential.

Why AI adoption stalls inside SMEs

Most unsuccessful AI initiatives do not begin with a poor tool.

They begin with an unclear objective.

A business owner hears that competitors are using AI. A director attends a demonstration. A few employees start experimenting with different platforms. Marketing uses one tool, sales tries another and operations begins a separate automation project.

There is activity everywhere but ownership nowhere.

The Chartered Management Institute found that 68% of managers’ organisations were still experimenting with AI or running early-stage pilots. Although 70% reported some productivity improvement, only 5% described the gains as transformational.

This is the gap between using AI and implementing it.

Using AI means individual employees find ways to complete certain tasks more quickly.

Implementing AI means the business changes a defined process, sets a standard, assigns accountability and measures whether the change produces a worthwhile result.

That requires leadership.

Start with the business problem, not the AI tool

The wrong starting question is: Which AI tool should we buy?

The better question is: Which recurring business problem is expensive, slow, inconsistent or overly dependent on one person?

An effective AI strategy for small businesses begins with operational friction.

Look for:

• Repetitive administrative work
• Delays caused by manual information transfer
• Data that is collected but rarely analysed
• Routine customer enquiries
• Inconsistent document preparation
• Reporting that consumes management time
• Processes that depend on one employee’s knowledge
• Rework caused by missing or inaccurate information

Once the problem is clear, the business can decide whether AI is an appropriate part of the solution.

Sometimes it will be. Sometimes a clearer process, better training or more disciplined management will produce a stronger result.

Technology should not be used to automate confusion. If a process is poorly defined, adding AI may simply make the confusion move faster.

Leadership decision 1: What outcome must improve?

Every AI initiative should begin with a measurable commercial or operational outcome.

“Use more AI” is not an outcome.

Neither is “improve productivity” unless productivity has been defined.

A useful outcome might be:

• Reduce the time required to prepare a proposal
• Shorten the response time for customer enquiries
• Improve the consistency of management reports
• Reduce manual data entry
• Increase the proportion of sales calls reviewed
• Reduce errors in standard documentation
• Give managers faster access to relevant operational information

Set a baseline before introducing the tool.

If preparing a weekly report currently takes four hours, record that figure. If customer enquiries take an average of two working days to answer, measure it. If a process regularly generates errors, identify the current error rate.

Without a baseline, a business cannot distinguish genuine improvement from enthusiasm.

Where is your business losing time?

Identify one recurring process that is slow, inconsistent or dependent on the owner. That is a stronger starting point for AI than a list of popular tools.

Leadership decision 2: Who owns the result?

AI initiatives often sit between departments.

The technology may be purchased by IT, introduced by the owner, used by marketing and expected to improve a company-wide result.

That makes accountability unclear.

Every initiative needs one named owner. That person should be accountable for:

• Defining the use case
• Selecting appropriate users
• Setting the operating standard
• Coordinating any required training
• Monitoring risk and quality
• Measuring the agreed outcome
• Recommending whether to expand, change or stop the initiative

The owner does not need to be the most technically knowledgeable person in the company.

They do need sufficient authority to make decisions and sufficient commercial understanding to judge whether the initiative is working.

If nobody owns the number, the AI tool will become another subscription rather than a business capability.

Leadership decision 3: What decisions can AI support?

AI can produce information, identify patterns, draft material and recommend actions.

That does not mean it should make every decision.

Leaders must define where AI sits in the decision-making process.

For each use case, establish whether AI will:

• Generate an initial draft
• Summarise existing information
• Identify options
• Flag an exception
• Recommend an action
• Complete an action after approval
• Complete a low-risk action automatically

The level of human oversight should reflect the consequences of being wrong.

A first draft of an internal meeting summary carries a different level of risk from customer pricing, recruitment decisions, financial projections or advice that could create legal obligations.

“Human in the loop” is not a complete policy. The business must specify which human, at which stage and against which standard.

Leadership decision 4: What information can employees use?

AI adoption frequently begins before a formal company decision is made.

Employees use publicly available tools to summarise meetings, review documents, draft emails or analyse information because the tools are convenient.

This creates an immediate leadership question: What information is permitted to enter an AI system?

An SME should establish practical rules covering:

• Personal and employee information
• Customer records
• Commercially sensitive data
• Pricing and margin information
• Contracts and legal documents
• Intellectual property
• Passwords and access credentials
• Confidential client material
• Information covered by non-disclosure agreements

The policy must be understandable to an employee completing real work. A vague instruction to “use AI responsibly” is not enough.

Different tools, accounts and service agreements may handle data differently. Obtain appropriate technical, data-protection or legal guidance where the use case involves sensitive information.

Leadership’s responsibility is to make the boundary clear before an avoidable mistake occurs.

Leadership decision 5: What does acceptable quality look like?

AI can produce a polished answer that is incomplete, inaccurate or unsuitable for the situation.

That makes output standards essential.

Before employees use an AI-generated output, they should know:

• Which facts must be verified
• Which sources are acceptable
• Which brand or technical standards apply
• Who is authorised to approve the result
• When specialist review is required
• What must never be sent without human checking

The standard should match the work.

A customer-facing proposal requires different checks from an internal brainstorm. A financial analysis requires different controls from social-media copy.

Do not leave quality to individual judgement when the consequences affect customers, employees or commercial decisions.

The AI tool may generate the work. Accountability for the work remains with the business.

Leadership decision 6: How will the work itself change?

AI is often added on top of an existing process.

Employees use the tool, but every old step remains. They generate a faster first draft and then move it through the same approvals, meetings and handovers as before.

The business saves minutes rather than changing performance.

Meaningful AI adoption requires leaders to redesign the workflow.

Ask:

• Which steps can be removed?
• Which handovers are no longer necessary?
• Where is information being entered twice?
• Which approvals genuinely manage risk?
• Can exceptions be reviewed instead of every case?
• Does a role need to spend less time producing information and more time making decisions?

The objective is not merely to complete the same work slightly faster.

It is to create a better way of working.

Leadership decision 7: Are managers capable of leading the change?

The success of AI for SMEs will depend heavily on the managers between the owner and the employees using the technology.

Managers must be able to:

• Explain why the change is being made
• Translate a broad goal into a working process
• Set expectations for responsible use
• Review AI-supported work intelligently
• Coach employees through new responsibilities
• Identify risks and exceptions
• Measure whether performance has improved

This cannot be delegated entirely to a technology supplier.

CMI research found that only 12% of managers felt very confident leading AI-enabled teams, while more than 80% believed better understanding of how to manage AI would improve their own and their team’s performance.

The capability gap is not simply about writing better prompts.

It is about leading change, setting standards and remaining accountable when technology becomes part of the team’s workflow.

Leadership decision 8: When will you stop the pilot?

Experiments need a conclusion.

Without a decision date, an AI pilot can continue indefinitely. Employees keep testing it, the subscription remains active and nobody determines whether it deserves further investment.

Agree the success criteria before the pilot begins.

At the end of the test, choose one of four actions:

1. Scale it: The outcome improved and the risks are controlled.
2. Refine it: The use case is valuable, but the process or training needs work.
3. Replace it: The problem is valid, but the selected tool is unsuitable.
4. Stop it: The benefit does not justify the time, cost or risk.

Stopping a weak initiative is not a failure.

Allowing an unmeasured initiative to consume attention indefinitely is.

A practical 90-day AI adoption plan for SMEs

An SME does not need to transform the whole business at once.

A disciplined first 90 days can establish whether the company is capable of turning AI into measurable value.

Days 1–15: Select the problem

• Identify recurring sources of delay, cost, inconsistency or rework.
• Select one use case with a clear operational owner.
• Record current performance.
• Define the expected improvement.
• Review the relevant data and commercial risks.

Days 16–30: Design the pilot

• Select the tool only after defining the use case.
• Choose a small group of users.
• Set rules for data, quality and approval.
• Train users on the process, not only the software.
• Agree how results will be recorded.

Days 31–60: Run and review

• Monitor the agreed measure.
• Record errors, exceptions and unexpected work.
• Gather feedback from employees and customers where appropriate.
• Review whether managers are applying the agreed standard.
• Adjust the workflow when evidence supports it.

Days 61–90: Decide

• Compare performance with the baseline.
• Calculate the real time and cost involved.
• Review output quality and risk.
• Decide whether to scale, refine, replace or stop.
• Assign ownership for the next stage.

This approach may appear slower than giving everyone access to the latest platform.

It is considerably faster than spending a year accumulating disconnected tools and unfinished pilots.

AI will expose the quality of your leadership

AI does not remove the need for leadership.

It makes the absence of leadership more visible.

If responsibilities are unclear, AI adoption will be unclear. If managers cannot hold people accountable, adding technology will not fix it. If the business does not measure performance, it will struggle to prove that AI has created value.

Official UK evidence supports the connection. ONS analysis found that firms with stronger management practices were more likely to follow through on planned AI adoption. Difficulty identifying appropriate business activities and use cases was also the most commonly reported barrier.

The leadership decisions come first:

• What problem are we solving?
• What result must improve?
• Who owns that result?
• What can the technology do?
• Where is human judgement required?
• What information can be used?
• How will quality be controlled?
• When will we decide whether to continue?

Answer those questions well and choosing the tool becomes easier.

Ignore them and even excellent technology will struggle to produce a meaningful return.

Do not start with another tool. Start with ownership.

HPA helps SME founders develop leaders who own measurable outcomes, make sound decisions and lead change without sending every issue back to the owner.

If AI adoption currently depends on scattered experiments, individual enthusiasm or your personal involvement, the leadership structure needs attention before the technology can scale.