Software has changed. Quality has to change with it.

AI is changing software faster than most of us predicted. Not because change itself is new. This industry has always moved quickly. What is different is the rate of change.

Software can now be designed, written and changed at a speed that simply wasn't possible when most of today's software development and testing practices were created. That creates an obvious opportunity.

It also creates a problem.

If we can create software faster than we can establish whether it works, quality becomes the bottleneck. That is the problem I think our industry now needs to solve.

It's why today we're launching Touchstone, the standard of proof for software releases. It's built on a simple idea: before you ship software, you should have the evidence to trust it.

We shouldn't use AI to do testing the old way faster

This is where I think we need to be careful. The easy response to AI is to apply it to what we already do. Generate test cases faster. Write automation faster. Execute more tests. Fix broken automation.

All useful. But I don't think that goes far enough.

If AI is fundamentally changing how software is created, we have an opportunity to rethink how software quality works from first principles.

And that starts with a different question.

Not:

How do we automate more testing?

But:

What do we need to prove before we trust this software?

Those are very different things.

A test can pass and still tell you very little. A thousand tests can pass and still miss the thing that matters. What matters is whether the software does what the business intended, whether the important risks have been tested, whether the right assertions have been made and whether there is evidence to support the decision to release it. That thinking has shaped what we have built with Touchstone.

It also changes the role of people

There is another part of this that I feel strongly about.

For years, we have put highly capable people into parts of software testing where humans are fundamentally disadvantaged. Repetitive work. At scale. Requiring absolute consistency.

Writing variations of tests. Maintaining them. Executing them. Checking results. Repeating the same process after every change. Machines are increasingly very good at this. People are good at something else.

Understanding why something matters.

Understanding the customer.

Interpreting business intent.

Identifying risk.

Challenging assumptions.

Making judgements.

And ultimately being accountable for a decision.

That is where I want the human in the system.It doesn't diminish the role of people in quality. It elevates it.

AI proposes. Humans govern.

This has become an important principle for us at Virtuoso.

Touchstone can understand an application and its context. It can turn objectives into requirements, journeys and assertions. It can create tests, execute them, maintain them and analyse the results.

And it can do that at a scale that would be extremely difficult for a human team. But there are points where a person should be involved.

Is this actually the right requirement?

Does this reflect the business objective?

Is this an acceptable risk?

Is the evidence sufficient?

Are we prepared to ship?

Those aren't tasks we should be trying to eliminate. They are the important ones. Our model is therefore deliberately simple:

AI proposes. Humans govern.

The machine provides scale, speed and consistency. The human provides context, judgement and accountability.

The output shouldn't be more tests

This is probably the biggest change in our own thinking. For years, our industry has measured testing through activity.

Number of tests.

Percentage automated.

Execution time.

Pass rates.

Those metrics have value, but none of them answers the question a business actually cares about:

Can I trust this release?

That is why we're increasingly thinking about the output of testing as evidence.

Evidence that the business objective has been tested.

Evidence that the important risks have been considered.

Evidence that the expected outcome occurred.

Evidence that the things that must not happen didn't happen.

Evidence that somebody reviewed what mattered and was prepared to put their name against the decision.

The goal isn't more automation.

It is greater confidence.

This is the opportunity

AI will create an extraordinary amount of software over the 12 months.

It will also create an extraordinary amount of change.

I don't believe enterprises will be able to respond by simply adding more people, writing more tests or running bigger regression packs. The economics won't work and the speed won't work. We need machines to carry far more of the work. But we also need people to become more important at the points where judgement and accountability matter.

That is the future we're building Touchstone for.

AI generates software.

Touchstone generates the evidence to trust it.

And people make the decision to ship.