Why This AI Strategy Session Never Became an AI Demo

There was a moment during a recent AI Strategy Session when I realised we weren’t going to get to the AI demonstrations I’d prepared.

And that was a good thing.

I was working with the leadership team of an established professional services business. Five people with different responsibilities, different levels of technical understanding and different perspectives on what AI might mean for their organisation.

The session was originally designed to run through AI strategy, Microsoft 365 Copilot, organisational knowledge, governance, prompt engineering, agents, automation and practical opportunities within the business.

We covered most of it.

But over four hours, something more interesting happened.

The conversation progressively moved away from “What can this AI tool do?”
Towards: “How should our business change because this technology now exists?”
That’s a much more important question.

AI adoption isn’t really a technology problem

One of the ideas I introduced early in the workshop was that AI represents a paradigm shift similar to previous changes in computing.

We moved from mainframes to personal computers.
Then from desktop computing to the internet.
Then from computers to smartphones and cloud platforms.

AI represents another shift because intelligence is progressively becoming embedded into the software, devices and workflows around us.

But there’s an important distinction.

Introducing AI into a business isn’t the same as installing another piece of software.

You’re introducing systems that can interpret information, generate content, analyse documents and increasingly take actions across workflows.

That immediately introduces questions around:

• information
• permissions
• security
• governance
• professional judgement
• organisational knowledge
• accountability
• culture
• change management

The technology quickly becomes the easy part.

Copilot, not autopilot

One principle kept resurfacing throughout the conversation:

AI should be treated as a copilot, not an autopilot.

This became particularly important when discussing professional work.

AI can help analyse documents, prepare information, compare requirements, populate templates, organise files and identify inconsistencies.

But professional judgement still belongs with people.

That means the opportunity isn’t necessarily to remove humans from processes.

It’s to redesign processes so machines handle more of the repetitive cognitive workload while people focus on judgement, relationships, verification and accountability.

Before AI can understand your business, your business needs to understand itself

This was probably the biggest theme to emerge from the session.

Businesses contain enormous amounts of knowledge.
Some of it lives in SharePoint or OneDrive.
Some lives in policies, procedures and templates.
Some exists inside project folders.
Some sits in email.
And a surprising amount exists only inside people’s heads.

That’s fine when an experienced employee knows where everything is and understands how the organisation works.

It’s a problem when you want AI to help.

AI can only work effectively with the context it can access.
So before asking:
“What agent should we build?”
A better question might be:
“What does our organisation actually know and where does that knowledge live?”

That leads into what I think will become an increasingly important part of AI strategy: organisational knowledge architecture.

For example:
Brand Codex
How does the organisation communicate? What does it believe? What tone should it use? What does good communication look like?

Knowledge Codex
What specialised organisational knowledge should people and AI understand?

Procedures
How is work actually performed?

Templates
What does a good output look like?

Prompt Libraries
What instructions consistently produce useful results?

AI doesn’t magically learn your business.
You have to teach it.

Poor processes don’t disappear when you add AI

They get amplified.

This became another important part of the workshop.

If information is inconsistent, AI inherits that inconsistency.
If permissions are poorly configured, AI can expose those problems.
If procedures aren’t documented, automation becomes difficult.
If nobody knows which document is current, giving an AI access to thousands of documents doesn’t necessarily solve anything.

This is why I increasingly see AI readiness as business readiness.

Before building sophisticated agents and automations, organisations should consider their information architecture, permissions, governance and knowledge.

The better organised the business is, the more useful AI becomes.

Governance isn’t there to stop AI

Governance can sound like the boring part of an AI conversation.

It shouldn’t be.

Good AI governance creates the confidence required to experiment.

For organisations operating inside the Microsoft ecosystem, there are technical considerations around permissions, information access, Microsoft Purview and sensitivity labelling.

But governance isn’t only an IT responsibility.

Leadership also needs to consider:

• acceptable AI use
• privacy and confidentiality
• human oversight
• professional accountability
• verification
• risk
• staff education
• communication

These policies can’t simply be written once and forgotten.
AI is evolving too quickly.
Governance needs to become a living organisational capability.

Start with the business problem, not the AI tool

Later in the session we started identifying actual opportunities.
This is where things became particularly interesting.

Instead of asking everyone:
“What could you do with Copilot?”
we looked at repetitive work and friction inside existing processes.

Examples included:

• processing instructions received by email
• ensuring job information is complete
• filing field data and photographs correctly
• checking plans against internal standards
• preparing quotes from existing templates
• managing repetitive administrative processes

None of these began with AI.
They began with a business problem.

We then used an AI Opportunity Canvas to break those problems down:

What happens today?
Where is the friction?
What information is required?
Could AI assist?
Where does a human need to remain involved?
What would success look like?

Only then should you select the technology.
That’s an important reversal.

Problem → Process → Knowledge → Governance → Technology.
Not:
Technology → find something to automate.

Quick wins versus strategic projects

Not every AI idea deserves to become a project.
Some opportunities are relatively simple, low-risk and immediately useful.
Others require integration, organisational knowledge, governance or substantial workflow redesign.

So we mapped opportunities according to business value and implementation effort.

That creates four useful categories:

Quick Wins
High value. Lower effort.
Start here.

Strategic Projects
High value. Higher effort.
Plan them properly.

Nice to Have
Lower value. Lower effort.
Experiment when appropriate.

Avoid for Now
Lower value. Higher effort.
Don’t automate something simply because you can.

This gives leadership teams a much more grounded way to discuss AI investment.

Prepare → Capture → Transform

By the end of the session, the roadmap had become surprisingly simple.

Prepare

• Get the foundations right.
• Understand licensing.
• Review the Microsoft environment.
• Check permissions.
• Establish governance.
• Build awareness.
• Communicate with staff.

Capture

• Identify opportunities.
• Capture organisational knowledge.
• Document procedures.
• Build templates.
• Structure information.
• Create Brand and Knowledge Codexes where appropriate.

Transform

• Pilot agents.
• Automate proven workflows.
• Train people.
• Measure outcomes.
• Improve.
• Scale what works.

That sequence matters.

Jumping directly to transformation can create impressive demonstrations.
It doesn’t necessarily create sustainable organisational change.

The biggest insight? We barely demonstrated the tools

I went into the session prepared to spend more time demonstrating AI.

We barely needed to.

The leadership conversation became far more valuable.

We talked about people.
Knowledge.
Processes.
Risk.
Governance.
Culture.
Information.
Professional judgement.
And how the organisation might operate differently as AI becomes embedded into everyday work.

That was an important lesson for me too.

After delivering several public presentations on AI education and adoption, this was my first opportunity to take the thinking inside an organisation and work directly with its leadership team.

It reinforced something I’ve suspected for a while.
The biggest AI opportunity isn’t teaching businesses how to use AI.
It’s helping businesses understand how they need to change because AI exists.

AI transformation is business transformation

The models will change.
Copilot will change.
ChatGPT will change.
Agents will change.
Licensing will change.
New platforms will appear.
Today’s exciting AI capability will eventually become tomorrow’s standard software feature.

That’s why I’m increasingly less interested in building AI strategies around individual tools.

The more enduring opportunity is helping organisations improve the things they control:

• Their knowledge.
• Their processes.
• Their governance.
• Their information.
• Their people.
• Their culture.

Get those things right and you create an organisation capable of adapting to whatever comes next.

AI isn’t the strategy.
Building a better, more adaptable business is.

Design with intent.
part mind | part machine.

Read more: 9 E-commerce Insights >>

I really do enjoy getting feedback so please let me know your thoughts on this or any of my articles in the comments field or on social media below.