AI Is Easy to Experiment With. Enterprise Adoption Is Different.

By David Mills – TasICT President

This is my perspective from having spent my career working in medium-to-large organisations, largely in manufacturing and financial services, and seeing first-hand the challenges involved in turning new technology from an exciting idea into something that can be adopted safely and at scale.

A question I hear regularly is:

“AI is moving so quickly. Why aren’t governments and large organisations moving faster?”

There are plenty of examples of individuals and small businesses experimenting with LLMs and getting impressive results within hours.

But moving from experimentation to enterprise-scale, sustainable adoption is a very different proposition.

Here are some of the reasons.

1. Experimentation Is Cheap. Getting It Wrong Isn’t.

A small business can experiment with an LLM relatively easily.

A large organisation has to consider privacy, security, regulatory, legal, reputational and ethical consequences if an AI system makes a bad decision.

That doesn’t mean organisations shouldn’t experiment. It means the risk framework around experimentation needs to be different.

2. The Legacy Stack Still Matters

Large organisations often have decades of systems, databases, integrations and business processes behind the front door.

Connecting modern AI safely and effectively into those environments is rarely as simple as plugging in an API.

The AI may be the easy bit. Getting it to work reliably with everything around it can be the harder bit.

3. AI Can’t Fix Poor Data

Reliable AI depends on reliable data.

Many organisations are still working through fragmented data architectures, inconsistent definitions, data quality issues and information spread across multiple platforms.

AI doesn’t magically make those problems disappear. In some cases, it makes them more visible.

4. Who Is Accountable When AI Gets It Wrong?

Organisations are still working through some fundamental questions:

What should we allow? What shouldn’t we allow? What are the risks associated with particular models? What data can be provided to them? And who is accountable when something goes wrong?

For executives and agency heads, these are serious accountabilities with significant regulatory, financial and reputational consequences.

5. Cool Technology Isn’t the Same as Business Value

Everyone can see that AI is impressive.

It’s much harder to demonstrate that a particular AI investment will deliver a measurable and sustainable return.

A successful pilot doesn’t necessarily translate into sustainable productivity improvement.

And the cost isn’t just token consumption. There can be significant infrastructure, integration, security, governance, change and people costs associated with getting AI into production.

6. Giving People AI Is Easy. Changing How They Work Isn’t.

Giving thousands of employees access to Copilot is relatively easy.

Changing the way thousands of people actually work is much harder.

AI can challenge established processes, roles and organisational structures. That makes adoption as much a change-management challenge as a technology challenge.

Employee representatives and unions can also have an important role in that conversation.

7. Don’t Turn Today’s AI Experiment Into Tomorrow’s Legacy

Technology teams have been dealing with shadow IT for decades.

AI makes it incredibly easy for business areas to develop their own solutions. That’s great for experimentation, but once these solutions become embedded in business processes, questions around security, support, resilience, integration and ownership become increasingly important.

Otherwise, today’s clever AI experiment can become tomorrow’s technology debt.

8. AI Moves Faster Than Procurement

Large organisations have established procurement frameworks for good reason.

Competitive tendering, security assessments, legal reviews, vendor due diligence and contractual requirements can take time.

Meanwhile, an AI startup can go from an idea to a compelling product in weeks.

The two worlds don’t always move at the same speed.

So, should we accept that large organisations will be slower?

Absolutely not.

These challenges explain why adoption can be harder. They shouldn’t become an excuse for not doing it.

They make it even more important for large organisations to change the way they approach technology adoption.

The organisations that succeed won’t necessarily be the ones with the biggest AI budgets.

They’ll be the ones that can move quickly while managing risk and turn experimentation into real, sustainable business value.

That is the challenge for leaders now.

The views expressed are my own and are shared in my capacity as President of TasICT.