One of the challenges of IT operations at a global consulting firm like Accenture is it runs a huge and diverse portfolio of technologies to drive its digital core. Maintaining and operating that core requires continuously developing deep skillsets that are becoming rarer as every industry competes within a limited talent pool.

Accenture’s answer is the Accenture Advanced Technology Agent (AATA), an agentic AI-powered integration platform that sits between the company’s human workforce and its technology platforms, and something that’s earned the company a .

“IT operations largely run through automation, either RPA, scripts, or CICD pipelines,” says Steven Courtney,  Accenture’s MD of global IT, technology vision, and strategy. “Wouldn’t it be great if we could add an orchestration layer above that, as an aggregation of AI agents, that we could apply to all our IT operations, one-by-one, as agents become available.”

Core means to efficiency

Like many large enterprises, Accenture has a highly complex digital core that contains a variety of IT solutions ranging from office architectures and multi-cloud, to data fabric and security platforms. The digital core is essential to run the business and enable quick onboarding of emerging technology, but the complex service chains and multiple technology stacks require a significant amount of human support. That, in turn, has made the rapid adoption of new technology difficult to integrate throughout the business.

To smooth that process, AATA is an agnostic, open system with agentic architecture that enables IT teams and end users to interact with the core to resolve issues or deliver customer solutions without having to connect to an agent or file a single ticket.

Accenture started working on AATA in 2023, starting with a conceptual architectural model. “Then we started to think about how people would consume the technology to derive value, so we wanted it to be largely democratized,” Courtney says. “You could have an open standards orchestration layer and a consistent prompt layer, but then enable teams to be able to either consume natively available agents, plug in existing automation or gen AI processes, or actually create custom agents if that was what required.”

Accenture

Some specific focus areas for AATA included variations in AI maturity. Recognizing not every organization will be in the same place when it comes to AI, Accenture built AATA to have open architecture with the flexibility to swap models in and out, allowing it to stay model agnostic and keep pace with the market.

There was also the importance of data availability and accuracy. AATA had to not only crawl Accenture’s internal content resources and repositories, but execute tasks based on the information in the firm’s living systems. That required secure access to the right data. Accenture’s global IT had already built a platform to bring all the company’s enterprise data together into a fabric, so AATA was able to capitalize on that work with common log lakes of critical data needed for operational decisions.

AATA also supports audio prompts and text chat embedded in Accenture’s collaboration tools, with a cloud-based workflow automation platform sitting behind the user interface. That platform acts as an orchestrator across various technologies and copilot solutions to collect data, create code, execute pipelines, or provide status based on the operational needs of the user.

Striving for continuous improvement

Building AATA wasn’t without problem solving. Upskilling was one focus area because consuming gen AI requires Accenture’s teams to behave and operate differently. But Courtney notes that the culture of continuous innovation at the firm meant that upskilling was part of the regular cycle of technology operations. Another area of concern was data.

“One of the biggest challenges we found in terms of making this happen on the ground was data and data accuracy,” Courtney says. “It’s nothing new, but it’s the fuel behind all the issues you find with gen AI and agentic architectures.”

He notes the framework calls on enormous amounts of telemetry.

“Every piece of telemetry that’s delivered from all of our infrastructure platforms, and some of our software platforms, we had to be comfortable that it was good, accurate, compliant, and that we were using it in a responsible and secure way,” Courtney adds.

The trick, adds Rajendra Prasad, Accenture’s chief information and asset engineering officer, is to not implement technology for technology’s sake.

Accenture

“You need to get your business processes reinvented before you infuse technology,” he says. “If you have a very inefficient business process, adding very high-powered agentic tech to that inefficient process just makes your inefficiency run more efficiently.”

Prasad also says that CIOs should use agentic AI as an innovative lens to reinvent and redefine processes.

Getting accustomed to the pace

Today, AATA has more than 100 active agents, and some results include being able to facilitate the configuration of VPN connections with much greater speed and accuracy. Configuring these connections used to require a highly skilled engineer 30 minutes, but AATA can now do it in two minutes, which is more than a 93% reduction with a lower risk of error. The platform also assists Accenture’s roughly 800,000 employees around the world with troubleshooting. Users can chat with AATA and receive a near-immediate response, facilitating auto-remediation of issues to close tickets faster. And cloud engineers also use AATA as an embedded partner.

“Adoption has been fantastic,” Courtney says. “We’re seeing about an 80% increase in terms of speed to provision and change, and the rate of agent growth about 15% week-on-week at the moment. We’re also finding that as people become more used to building an agent rather than using legacy techniques, things just accelerate.”