When the mainstream market responds cautiously to the arrival of new technologies it is a well-rehearsed vendor gambit to showcase itself as its own “customer zero” for the new approach. 

In this vein, Mike Ferris, Red Hat’s Chief Strategy and Operations Officer, recently briefed analysts on how Red Hat is deploying AI to transform the business operations that actually run the company. Getting software engineers on board is one thing, but getting other professionals on board can be more difficult, so how did Red Hat motivate its functional staff?

Turning functional staff into product developers

Like many organizations, Red Hat has been experimenting with individual tools and individual pilots but is now moving to a new phase to operationalise the use of AI within the business. The main driver for the company’s adoption of AI is to accelerate Red Hat’s own growth by turning its use of the technology into capabilities that go into offerings for customers. So, a basic question to all staff – or associates as Red Hat calls them – is how do you contribute to what the firm sells by turning what it does into product?

Employees needed to build their skills and gain confidence in using AI tools. As Ferris explains:

We started from the ground up from the engineering teams, but on the operation’s side there is a lot more scope for engagement with associates. With these functional experts, we put the technology into their hands to create something new because we felt in this way we could evolve both our products and the operations themselves.

Red Hat is building an AI-enabled workforce by embedding in-the-flow copilots to eliminate repetitive work and speed up the execution of routine daily tasks. The company is deploying Gemini for workspace as well as Cursor and Claude to create internal AI agents. Continuous up-skilling is supported via its Train the Trainer program.

The approach is enabled by the connected intelligence available in Red Hat’s data lake. Ferris expands: 

We are not just connecting data but allowing associates to see far more than they ever have before, while protecting private personal information. Now we can immediately find the right people, processes and data using AI, rather than me having to physically walk the hallways trying to make the right connections to get cross-functional things done. 

Once we started getting all this data, we began applying engineering principles to the business operations side of the house, applying it to things that it couldn’t have been applied to before. We are treating business operations as software code under the term  ‘business-as-code’. This is generating a new momentum around efficiency and more associates are able to create new value for the company.

A team was set up working across the company to identify individuals that understood the workflows and could then put this knowledge in mark down language. This unstructured information went into a data lake that meant it was accessible to insights tooling and could then be queried, prompted and transacted against by generative and agentic AI. According to Ferris said:

The evolution of workflow and business processes in this environment is similar to the impact using Claude has on developing code.

Business-as-code, a unified purpose

Red Hat has created a business-as-code foundation of digital controls, policies and processes to ensure the company remains AI-ready, secure and compliant. This foundation is translating the company’s business logic, processes and workflows into machine–readable code and assets for agentic execution and evolution.

The firm is now running AI-augmented operations as it transitions from manual process steps to orchestrated business lifecycles where systems are digitally operated by agentic AI. To date, Ferris reports that Red Hat has re-designed or optimized across more than 540 in-flight AI projects across processes such as go-to-market, talent lifecycle and quote-to-cash.

The business-as-code approach comes to life in the form of a connected network of agents that surface real-time natural-language intelligence. This network currently has around 70 agents in Legal, Finance and HR that “talk” to each other, using un-structured knowledge curated by subject matter experts. Larger reasoning models orchestrate the agents while small specialized models execute tasks. Ferris explains:

Across the board, we are running evals, building this up and then can measure results as they come out the other end. Over two quarters, this approach has generated around 3.4 times ROI. For example, we launched the Lightwell product in eight weeks, which would have taken much longer without the agent. We can also understand the impact of the launch on the rest of our portfolio.

Of course, there were challenges, especially around balancing innovation and standardisation. With the widespread use of AI agents, there were dis-connected pilots and duplicated efforts as teams would rebuild rather than re-use. Red Hat has set up a central group to help with this by getting communities to work together to build common libraries of skills.

Another issue lies with content sprawl as inconsistent formats limit AI readiness and so the most important sets of documents need to be moved to a common format. Ferris says:

These challenges are more easily tackled with company buy-in to business-as-code, which is promoted as the ultimate goal. This has clarified purpose for the entire workforce, as everyone is responsible for their own functions contributing to the efficiency of the company and the value they can bring to customers.

My take

The big lesson here is that the breakthrough for operationalizing AI at scale in Red Hat was deploying tools to operations teams as well as to software engineers. We have already accepted that the technology is transforming software development but the chasm-crossing point for sectors beyond tech is taking longer to achieve. Here is a tech vendor candidly explaining the impact that this technology is having in its functional business areas, and the key to unlocking this value was ensuring everyone got involved, not just the engineers. 

Now it may prove easier for a company in the tech sector to get this type of operational AI initiative going across the whole business, but that notwithstanding, it offers a good example for many other types of enterprises to study.