How Are Companies Really Using AI?

A few years ago, AI still felt distant to most businesses – talked about often but applied under very specific circumstances. That changed at the enterprise level as the technology became more accessible, more widely distributed, and able to handle a broader set of challenges with higher quality.

Now, its footprint is significant. McKinsey’s most recent State of AI Global Survey shows that 78% of companies use AI in at least one business function, up from 55% in 2023. But the more useful question is where it’s being used – and what that tells us about what businesses are trying to solve.

AI in the SDLC

Among all business functions, software development may offer the most immediate and compounding return on AI investment. The logic is simple: Every digital product, service, and internal system depends on software. If you improve the process that builds and maintains that software, you improve everything else downstream. Development can often be a bottleneck to unlocking business value – and improving developer productivity can reduce time to market by as much as 30% to 40%.

This is why we see AI adoption accelerating in the SDLC, especially in complex, regulated sectors like insurance and financial services. These industries face a unique combination of scale, compliance pressure, and legacy systems. AI helps manage all three.

At the code level, generative tools now assist with boilerplate generation, documentation, pattern recognition, and more. They reduce time spent on low-differentiation tasks and free developers to focus on architecture and logic. On the testing side, AI is closing a longstanding coverage gap. Synechron Verifai, a tool used by a leading U.S. bank, increased test coverage from 3% to 55%.

Another shift, though, is in how teams work. AI acts as a connective layer – linking business requirements to code, QA to development, and delivery to operations. That reduces handoff friction and creates tighter feedback loops across the pipeline.

For companies looking to scale AI, this is a critical point. When AI is embedded in the SDLC, its impact touches every part of the business that software enables.

AI for Research and Knowledge Management

Every organization needs access to knowledge, but as data grows, finding and using that knowledge becomes more difficult. Information is often stored in separate silos, which means that retrieving it depends on what people remember. Consequently, accuracy tends to be replaced by rough estimates.

This is where AI creates structural value. Large language models (LLMs) can retrieve answers from across disconnected sources. They can query a system directly, in plain language, and get context-rich, specific responses drawn from the company’s own materials.

This is especially valuable for professional services organizations, where delivery depends on speed and reapplication of expertise. AI accelerates research, reduces duplication, and strengthens continuity, therefore reducing reliance on individual knowledge holders and enabling faster ramp-up for new team members or cross-functional collaboration.

In practice, this allows firms to make better use of what they already know, and most importantly, apply it at scale.

Model Validation

Using AI in production requires a framework to validate them consistently, transparently, and at scale. Without this, organizations may deploy systems that are poorly calibrated, biased, or difficult to audit.

Model validation is what separates experimentation from enterprise-grade deployment. It tests whether models behave as expected, under a range of conditions. It establishes thresholds, tracks drift, and sets guardrails for change. In regulated sectors, like financial services and insurance, this is an expectation often subject to external scrutiny.

A strong validation framework covers input data quality, model interpretability, risk classification, and performance monitoring over time. It ensures that models don't just perform well in ideal conditions, but remain stable in production, too.

Organizations that treat model validation as core infrastructure will be better positioned to scale AI responsibly. If they build validation and testing earlier into the development process, they'll be able to avoid the trap of fast deployment followed by slow, costly rework. That approach also helps earn the trust of internal stakeholders regulators, and customers.

Conclusion

AI is already embedded in how leading companies operate. What sets these apart is focused implementation – they’re choosing use cases where the value is clear and the outcomes are measurable .

The examples in this article show what that looks like in practice: Targeted tools in the SDLC, faster access to institutional knowledge, and rigorous validation frameworks that support safe, repeatable use.

The question is no longer whether to use AI – most companies are – but how to apply it with intent.