AI strategy and subject matter expertise

AI Won’t Replace Subject Matter Experts. It Will Empower Them.

AI will not replace true subject matter experts. It will expose shallow expertise, remove low-value admin and give real experts more leverage, provided organisations design AI around human judgement rather than pretending tools can replace it.

Topic: AI adoption Focus: Subject matter experts Reading time: 10 minutes Author: Steve Wilson

The wrong question keeps getting asked

There is a tired argument that keeps appearing whenever AI is discussed.

Will AI replace people?

It is the wrong question.

A better question is this: which parts of work should AI take over, and which parts should remain firmly in the hands of people who understand the domain?

That distinction matters because not all work is equal.

Some work is repetitive, low-risk and rule-based. AI can help there. Some work is analytical, contextual, ethical, relational or judgement-heavy. AI can support that work, but it should not own it.

Subject matter experts sit in the second category.

They do not simply know facts. They understand context. They know what matters, what is missing, what sounds plausible but is wrong, what risk looks like before it becomes visible, and when a technically correct answer is still not the right answer.

Key point: AI is not a substitute for real expertise. It is a leverage layer for experts who know how to question, verify, interpret and apply its outputs.

What subject matter experts actually do

Subject matter experts are often described as people who know a lot about a specific area.

That is true, but incomplete.

A true SME does more than hold information.

They interpret situations. They notice weak signals. They understand constraints. They know how a decision will land with real users, customers, staff, regulators, suppliers or stakeholders.

They can tell the difference between a standard answer and the answer that fits the context.

That kind of expertise is hard to automate because it is not just stored knowledge. It is judgement built through experience.

In business terms, SMEs help organisations avoid expensive mistakes. They understand why a policy reads well but fails in practice. They know why a workflow looks simple on paper but creates hidden work. They can see when an AI output sounds polished but misses a critical assumption.

This is why Changeable treats AI adoption as a human-led discipline that combines AI strategy, process improvement, AI governance and workflow design.

AI is very good at some parts of expert work

AI can be extremely useful for subject matter experts.

It can reduce the time spent on repetitive, administrative or first-pass work.

Summarise long documentsTurn heavy reading into a faster starting point for expert review.
Draft first versionsCreate report, email or briefing drafts that an expert can refine.
Extract key pointsPull out actions, risks, themes or obligations from messy source material.
Compare informationHelp review policy documents, research notes or draft options side by side.
Generate checklistsCreate practical review lists, planning prompts and quality checks.
Find gapsFlag inconsistencies, missing assumptions or unclear sections for review.

Those tasks matter, but they are not the heart of expertise.

They are preparation tasks.

AI can help experts get to the judgement layer faster.

This is where generative AI and AI agents can create real value. They can act as support layers for research, drafting, triage, summarisation and analysis, while the expert remains responsible for interpretation and final decision-making.

Useful distinction: AI can help produce a first answer. A subject matter expert knows whether that answer is useful, safe, complete and appropriate.

AI exposes shallow expertise

AI will not replace strong subject matter experts.

But it will expose people who have been relying on surface-level expertise.

If someone’s value is mainly summarising generic information, repeating standard advice or producing basic documentation, AI will make that work easier to replicate.

That does not mean the person has no future. It means their value needs to move up the chain.

Ask better questions.
Frame the real problem.
Identify assumptions.
Check sources.
Interpret trade-offs.
Apply judgement.
Communicate clearly with humans.
Use AI outputs without being captured by them.

The World Economic Forum’s Future of Jobs Report 2025 makes this direction clear. It identifies rapid growth in AI, big data and technology skills, while also noting the continued importance of human skills such as analytical thinking, resilience, leadership and collaboration.

That is the real shift.

AI does not make expertise irrelevant.

It raises the standard for what expertise needs to look like.

Why AI needs experts more than people think

AI systems can produce confident outputs from incomplete, outdated or poorly understood information.

That confidence is part of the risk.

When AI is wrong, it often sounds right.

A subject matter expert is needed to check whether the output makes sense in context.

Legal summariesMay miss a critical clause.
Policy interpretationMay ignore local operating practice.
Customer emailsMay sound polished but create a promise the business cannot keep.
Data analysisMay identify a pattern but miss the reason behind it.
Process mapsMay document the official process but miss hidden work.
Governance recommendationsMay sound sensible but be impossible for the team to apply.

AI can generate outputs.

Experts determine whether those outputs should be trusted, changed, rejected or escalated.

This is why AI governance needs human review points. The goal is not to slow everything down. The goal is to preserve judgement where mistakes would matter.

The best SMEs will become AI-enabled experts

The strongest subject matter experts will not ignore AI.

They will use it well.

AI-enabled experts will be able to move faster because they can delegate low-value cognitive work to tools while retaining control over judgement.

Prepare before meetings.
Review documents faster.
Test assumptions.
Generate options.
Compare alternatives.
Summarise evidence.
Identify risks.
Draft outputs for refinement.

But they will also know when not to use AI.

That is equally important.

An AI-enabled expert understands that some work needs privacy, care, human conversation, professional responsibility, ethical judgement or deep contextual interpretation.

This is the principle behind identity-safe automation. AI should remove the work that drains expertise, not undermine the expert’s value.

AI changes the shape of expertise

Expertise will not disappear, but the shape of it will change.

In many roles, the value will shift away from producing every output manually and toward supervising, interpreting and improving AI-assisted outputs.

That means SMEs will need new habits.

Prompt clearlyWrite effective prompts and instructions.
Review criticallyCheck AI outputs before using them.
Trace claimsConnect outputs back to sources.
Recognise overconfidenceNotice hallucinations, missing context and weak assumptions.
Set review pointsDecide when human review is required.
Build reusable workflowsCreate repeatable patterns that improve over time.

Microsoft’s 2025 Work Trend Index describes a shift toward employees working with AI agents and becoming “agent bosses”, meaning people increasingly delegate to, guide and manage AI tools as part of their roles.

That does not remove expertise.

It changes how expertise is exercised.

The risk of deskilling

There is a real risk that AI can weaken expertise if it is used badly.

If junior staff rely on AI too early, they may not develop the judgement that comes from doing the hard work themselves.

If experts stop checking outputs carefully, their standards may slip.

If organisations automate too much of the thinking process, people may lose the ability to spot when something is wrong.

This is the deskilling risk.

It matters because expertise is not built only by consuming answers. It is built by struggling through problems, testing assumptions, making mistakes, receiving feedback and learning how to distinguish quality from noise.

Research on expert cognition and generative AI has highlighted this tension: AI can reduce cognitive load and support expert work, but systems need to be designed in ways that preserve expert agency, verification and expertise development.

Practical rule: Use AI to reduce unnecessary workload, not to remove the learning experiences that build real expertise.

That is why AI implementation should not simply ask, “What can we automate?”

What expertise must we preserve?
Where do people still need deliberate practice?
Which outputs require expert review?
How will junior staff learn if AI does the first pass?
How do we prevent AI from becoming a crutch?
Where is Minimum Viable Friction needed?

Some deliberate pause is needed to protect judgement and learning.

What this means for organisations

Organisations should stop framing AI as a replacement strategy.

That framing creates fear, resistance and poor implementation.

A better frame is expertise amplification.

Ask where AI can help experts spend less time on low-value work and more time on the work that actually needs them.

Expert roleLess time onMore time on
Policy expertSummarising documentsInterpreting implications
Business analystFormatting notesClarifying requirements
LawyerSearching first-pass materialReviewing risk
AccountantSorting transactionsAdvising on decisions
Healthcare workerDrafting routine notesSpending time with patients
Tourism operatorAnswering repeated questionsHosting visitors
Dairy adviserManually compiling reportsInterpreting farm performance

That is the real productivity opportunity.

Not replacing the expert.

Giving the expert more leverage.

Where AI support works best

AI works best around subject matter experts when the work is clearly bounded.

Good use cases

  • Document summarisation.
  • First-draft briefing notes.
  • Research organisation.
  • Meeting synthesis.
  • Checklist generation.
  • Policy comparison.
  • Knowledge retrieval.
  • Repeated customer enquiry support.
  • Quality review against a standard.
  • Draft communication for expert review.

Poor use cases

  • Final decisions without human review.
  • High-risk advice without expert oversight.
  • Use of sensitive data in unapproved tools.
  • Automation of unclear processes.
  • Replacing professional judgement with AI confidence.
  • Using AI outputs without source checking.

This is why AI use case discovery matters. Organisations need to identify where AI is useful, where it is risky and where the work should remain human-led.

AI and professional judgement

Professional judgement is not just choosing between options.

It includes understanding consequences.

ContextWhat is happening around the decision?
ImpactWho is affected?
AssumptionsWhat assumptions are being made?
EvidenceWhat information is missing?
RiskWhat level of risk is acceptable?
AccountabilityWhat are we responsible for?

AI can support those questions, but it cannot be accountable for the answer.

This is especially important in areas involving people, privacy, public trust, employment, legal obligations, service eligibility, safety or financial decisions.

For New Zealand organisations, the Privacy Act 2020 and Information Privacy Principles are an important consideration whenever AI tools process personal information.

The human expert remains essential because someone must own the judgement.

AI should sit beside the expert, not between the expert and the work

A useful way to think about AI is as a support layer beside the SME.

AI can prepare material, suggest structures, surface patterns and draft options.

The expert then reviews, changes, rejects, approves or applies the output.

This keeps expertise active.

The danger comes when AI sits between the expert and the work.

That happens when the system starts making decisions, shaping conclusions or filtering reality before the expert has a chance to apply judgement.

In those cases, the expert can become a rubber stamp.

That is bad design.

Good AI design keeps experts engaged where judgement matters.

This is why AI agents need clear boundaries. An agent may support research, drafting or triage, but its authority should be limited and its outputs should be reviewed where risk or judgement is involved.

The role of AI governance

AI governance protects subject matter expertise by defining where AI can help and where humans must remain in control.

A practical governance model should answer:

Which AI tools are approved?
What information can be used?
Which outputs require SME review?
Who is accountable for final decisions?
How are errors corrected?
How are prompts, templates and agents maintained?
What work should not be automated?
How will staff learn and improve over time?

Governance should not be a brake on expertise.

It should create the conditions where experts can use AI safely, confidently and productively.

This is especially important for professional services, public sector organisations, councils, health-adjacent services, education, financial services and any organisation using AI to support decisions that affect people.

How to empower SMEs with AI

Organisations can take a practical approach.

1

Start with the expert’s pain points

Ask SMEs where their time is being wasted. Look for repeated admin, manual review, searching, formatting, summarising, copying, checking and report preparation. These are often strong AI support opportunities.

2

Map the workflow

Do not add AI until the workflow is understood. Identify where the expert applies judgement, where the work is repetitive and where handoffs create friction. This is where process improvement should come before automation.

3

Define the AI role

Give AI a clear job. For example, summarise this document for expert review, draft a first version of this response, extract obligations from this contract, compare these policies and flag differences, or find repeated themes in this feedback.

Do not give AI vague authority over an entire professional process.

4

Keep review points explicit

Define where the SME must review, approve or override the AI output. This protects quality and accountability.

5

Train SMEs on verification, not just prompting

Prompting matters, but verification matters more. SMEs need to know how to check outputs, trace claims, identify gaps, recognise uncertainty and decide when a tool should not be used.

6

Measure value properly

Do not measure AI success by usage alone. Measure whether experts have more time for higher-value work, whether quality improves, whether rework reduces and whether decisions become clearer.

This is where reflection as an operating system helps turn AI use into measurable improvement.

What SMEs should do personally

Subject matter experts should not wait for perfect organisational AI strategy before learning.

They can start building capability in low-risk ways.

A good personal AI practice includes:

Use AI for first drafts, not final advice.
Ask AI to challenge assumptions.
Use AI to summarise, then check against the source.
Create reusable prompts for repeated tasks.
Keep a record of where AI helps and where it fails.
Do not enter sensitive information into unapproved tools.
Use AI to prepare, not to avoid thinking.
Keep strengthening your core expertise.

For people still building confidence, Zero to AI is designed to help make AI practical and less intimidating, especially for people who want to learn by doing rather than by watching hype unfold from the sidelines.

What leaders should stop saying

Leaders can accidentally undermine SMEs by talking about AI badly.

Statements like “AI will do this work now” or “we should not need experts for this anymore” create resistance and fear.

Better language is more accurate:

“We want AI to remove low-value work so experts can focus on judgement.”

AI assistsHumans remain accountable.
Review is definedSME review points are clear.
Consistency improvesAI strengthens quality without erasing expertise.
Workflows are co-designedThe people who understand the work shape the system.

That language matters because adoption is not only technical.

It is social.

What this means for Changeable’s work

This is the model Changeable is built around.

AI does not replace expert judgement.

It extends it.

Changeable uses AI agents, structured prompts, governance models, data structures and workflow automation to support better consulting work, faster analysis and clearer implementation.

But the accountable human remains central.

This is also how Changeable helps clients: not by pushing AI into every corner of the business, but by identifying where AI can safely amplify the people who already understand the work.

That is the sensible future. Not humans versus AI. Experts with leverage.

What Changeable helps with

Changeable helps New Zealand organisations empower subject matter experts with practical, governed AI support.

AI strategyIdentify where AI can amplify expertise and where human judgement must remain central.
AI use case discoveryTest whether an SME-support idea is viable before investing.
AI agentsSupport research, triage, summarisation, document support and knowledge retrieval.
Generative AI systemsSupport drafting, communication, analysis and content workflows.
Process improvementIdentify where expert time is being lost to poor workflow design.
Workflow automationReduce repeated manual work around expert tasks.
Data modelsStructure the information experts need to work reliably.
AI governanceDefine review points, privacy rules, accountability and safe use.
AI maturity and readiness assessmentIdentify capability gaps before scaling AI across expert teams.
Fractional AI leadershipProvide senior AI guidance without a full-time AI lead.

Start with a Decision Clarity Session

A Decision Clarity Session is a no-obligation conversation where we listen to what you are trying to achieve, what is getting in the way and whether AI strategy, SME empowerment, AI agents, workflow automation or governance is the right next step.

Book a free Decision Clarity Session →

Frequently asked questions

Will AI replace subject matter experts?

AI may replace some repetitive or low-value tasks, but it does not replace true subject matter expertise. Experts are still needed for judgement, context, interpretation, accountability and decisions where consequences matter.

How can AI empower subject matter experts?

AI can help SMEs summarise documents, organise research, draft first versions, identify patterns, compare information, prepare options and reduce manual work so they can spend more time on higher-value judgement.

What tasks should SMEs not delegate fully to AI?

SMEs should not fully delegate high-risk advice, final decisions, sensitive communication, legal or compliance interpretation, people-impacting decisions or work that requires deep context and accountability.

What is the biggest risk of AI for experts?

The biggest risk is not replacement. It is over-reliance. If experts stop verifying outputs, stop developing judgement or allow AI to sit between them and the work, quality and capability can decline.

How should organisations introduce AI to expert teams?

Start with expert pain points, map the workflow, identify low-value tasks, define AI’s role, keep human review points explicit, train for verification and measure whether expert time is being used better.

Why does AI governance matter for SMEs?

AI governance defines which tools are approved, what information can be used, where human review is required, who is accountable and what work should not be automated.

How can Changeable help?

Changeable can help organisations identify practical AI use cases, design AI agents, improve workflows, create governance and support subject matter experts so AI strengthens expertise rather than undermining it.

About the author: Steve Wilson is the founder of Changeable and Ministry of Insights, providing AI strategy, governance and automation consulting for organisations navigating the gap between AI ambition and operational reality.

For people and teams still building confidence with AI before implementation, visit Zero to AI.

Give your experts more leverage without handing judgement to AI.

Changeable helps organisations design practical, governed AI support around the people who understand the work. Start with the workflow, protect the judgement, and use AI where it genuinely improves the way expert teams operate.