Navigating the future of digital

Technology moves quickly. Deciding what deserves your attention is harder than keeping up with the announcements.

AI, spatial computing, cybersecurity and cloud infrastructure are changing how digital products are built and used. Some of those changes are already affecting day-to-day work. Others still need testing before they justify a substantial investment. A useful digital strategy has to make that distinction.

This was the starting point for a series I wrote for Deloitte Digital: understand the shifts, look at what they mean for your organisation, and decide where to act.

AI is changing the work of building products

AI is becoming part of the product itself, as well as the process used to build it. Assistants can help people find information, complete tasks and make sense of unfamiliar systems. Behind the interface, models can support classification, prediction and automation.

Software development is an obvious example. Code generation, test assistance and documentation tools are changing how teams spend their time. GitHub Copilot and similar tools can take on parts of an implementation, but somebody still has to decide whether the result is correct, secure and worth maintaining.

That changes the skills a team needs. Reviewing generated code, describing a problem precisely and recognising a weak assumption become more important as more of the implementation can be produced automatically. Developer experience matters here: a tool is only useful if it fits the way people work and helps them deliver software they can stand behind.

Low-code tools give more people a way to build small applications and automate their work. That can remove a bottleneck. It can also create software that nobody has planned to support. Access, ownership and maintenance deserve attention before an experiment becomes a business dependency.

For organisations, the questions are practical. Which decisions can an AI system make? Where does a person need to remain involved? What happens when the answer is wrong? Those decisions belong in the product design, not in a policy written after deployment.

Spatial computing needs a reason to leave the screen

Extended reality and spatial computing offer different ways to interact with digital information. Instead of putting every task in a browser window, an interface can sit alongside the equipment, room or physical activity it relates to.

Training and field service are useful examples. A person can practise a procedure in a simulated environment, or see instructions in the context of the machine they are working on. In retail, digital information can help connect what somebody sees online with what they find in a physical space.

Through Deloitte Digital’s Unlimited Reality practice, this was already becoming concrete work rather than a discussion about headsets. The value depends on the task: does the additional spatial context help somebody understand, practise or complete it?

A new device does not automatically make an experience better. Comfort, accessibility and the effort of using it can outweigh the benefit. The starting point should be a problem that a conventional interface handles poorly.

Trust has to be designed into the product

AI-generated content and impersonation make it harder to judge what is authentic. Connected services also give organisations more data to protect and more dependencies to understand.

Security decisions shape the experience people have with a product. Authentication, permissions and recovery are part of that experience. So are less visible decisions about what data is collected, how long it is retained and who can use it.

Zero-trust approaches ask systems to verify access rather than assume that somebody inside a network should be trusted. Preparing for post-quantum cryptography raises a longer-term question: which systems and data need protection over a period that may outlast the cryptography they use today?

These are not interchangeable solutions, and they do not all belong in every project. They are reasons to involve security expertise early, understand the risks that apply, and make explicit decisions about them.

People need to be able to trust a product with something that matters to them. Meeting a compliance requirement is part of that work, but it is not the whole measure of success.

Cloud and edge choices should follow the workload

Cloud-native infrastructure gives teams options for deploying and operating services. Edge computing can put processing closer to the place where data is generated or used. AI workloads add their own demands for computing capacity, latency and data handling.

Those options are useful when they solve a specific constraint. A service that needs to respond immediately on a factory floor has different requirements from a reporting system. Sensitive data may need to stay within a particular environment. An industry-specific platform may offer capabilities or controls that a team would otherwise have to build.

The architectural work is deciding which trade-offs are acceptable. Multi-cloud can address a real dependency or regulatory concern; it can also introduce operating costs and complexity that a team did not need.

The aim is an infrastructure the organisation can run and adapt. That means understanding the workload, the people who will support it and the consequences when a dependency fails.

Keep an eye on the hardware

Digital strategy cannot stop at software. AI-specific chips and local computing hardware affect where models can run and how much they cost to operate. Sensors and connected devices give software access to conditions in the physical world. Robotics adds the possibility of acting on that information.

That creates opportunities in places such as warehouses, industrial operations and healthcare. It also brings constraints that a software-only prototype can hide: physical safety, reliability, power and the cost of maintaining equipment.

Quantum computing is worth following for its potential in particular kinds of problems, including materials research and optimisation. It should not be treated as a general promise that difficult computing problems will soon become trivial. The useful question is whether a developing capability could change a problem your organisation actually has.

The same discipline applies to other emerging fields, from biotechnology to new forms of human-computer interaction. A technology can be important without being ready for your next project.

Decide what to test, and what to leave alone

There is no shortage of technology to explore. The difficult part is choosing an experiment that can change a decision.

Start with a real problem, name the assumption you want to test and decide what evidence would justify doing more. Involve the people who would use the result and those who would have to operate it. A prototype that works once is a different proposition from a service somebody depends on.

Some ideas will earn further investment. Others should stay on the watch list. Either outcome is useful if it helps the organisation make a clearer decision about what to build next.