Hitachi Digital Services builds the bridges to cross the Agentic AI chasm

The IT industry is a fickle world.  A place where the market sentiment oscillates between a gold rush sentiment when ChatGPT came onto the scene and soul-searching after a slew of studies suggested that organizations struggle to scale their AI deployments and that very few of those projects have a tangible impact on the bottom line. A study by MIT, which suggested that only 5% of AI projects scale, sparked much debate. Regardless of the specific methodology behind that study, the concerns about scaling and operationalizing AI are widespread. McKinsey described those challenges as the GenAI Paradox, which states that widespread AI adoption isn’t producing material business gains. If more evidence of the changing market sentiment were needed, Goldman Sachs and the Bank of England, among others, are warning about a substantial correction in company valuations due to inflated expectations regarding the impact of AI. PAC was discussing these issues with Hitachi Digital Services (HDS) in the context of the launch of their HARC agents, a unified agentic operating framework.

Realistic depictions of clients’ challenges are the best way to cut through the market noise

Executives at HDS elaborated on the challenges that clients face when attempting to operationalize AI. The main point to remember is that HDS has been involved in enterprise AI for a long time, predating the emergence of LLMs on the market, given its focus on the intersection of IT and OT. Therefore, it has earned the trust of many of its clients to be their trusted advisor, guiding them through the ever-changing AI landscape. Against this background, they highlight that enterprises want to scale but need ways to deploy Agentic AI faster to capture value before their competitors. As a market, we haven’t seen the proof points yet, but the dynamics will strongly accelerate once best practices have been established. What is missing is enterprise readiness, as most projects are stuck in silos. Moreover, there is a lack of a unified control, security, and observability layer for managing agents across platforms. This leads to organizations reinventing components (such as agents and workflows) instead of reusing or leveraging pre-built agent systems.

To overcome those challenges, HDS is focusing its Agentic AI strategy on four core elements:

R2O2: A governance-centered approach to developing AI that is dependable, ethical, transparent, and efficient.

HARC for AI: A managed services layer that ensures AI systems are always-on, cost-optimized, and production-ready.

Agent Library: A repository of 200+ pre-built agents to accelerate solution development and maximize output.

Agent Management System: a unified control layer for governing, securing, and observing AI agents across diverse platforms.

Minimal Viable Agent as a lever to get comfortable with a non-deterministic approach

In practical terms, HDS is pursuing a strategy it calls the Minimal Viable Agent. As many clients are not yet comfortable with fully embracing a non-deterministic mindset, the approach hardcodes and preconfigures significant parts of the workflow. While an LLM is leveraged to understand intent and generate a response, the task manager, which is central to the notion of agentic orchestration for scheduling and defining tasks, will be deployed only once clients are more comfortable with non-deterministic decision-making. Similarly, learning and feedback mechanisms such as reflection and self-critique, as well as pure reinforced learning using systems of reward, are only added with increased maturity. Lastly, domain compliance will also only be agentized when customers progress on their transformation journey.

What HDS aims to offer its clients is an “Intelligent Architecture” that enables a phased push toward non-deterministic systems, while also being flexible enough to integrate new innovations that emerge on the market every other day. As executives at HDS explained, this Intelligent Architecture is the confluence of rule-based logic that has always existed, along with reason-based logic that Agentic AI provides. Thus, HDS is avoiding leaning on clients with an ambitious North Star. This brings us back to the change in market sentiment as many organizations struggle to capture value from their investments in AI.

To cross the Agentic AI chasm, focus on workforce transformation and operating model reimagination

While the title of Figure 1 is clearly alluding to Geoffrey Moore’s seminal book, which examines the adoption gap that lies between early and mainstream markets, that chasm is widened and deepened by the perception that early deployments struggle to enable value capture. The reasons for that are manifold. Most critically, many deployments are done in isolation rather than at an enterprise level, and retrofitting innovations like AI into traditional operating models means the real value is not being realized. However, the infographic clearly depicts the trajectory of an S-curve, where multi-agent orchestration must be understood as a workforce transformation that is underpinned by the reimagination of operating models. Put another way, organizations must progress from augmentation in the early phase of adoption to transformation.

Figure 1: Crossing the Agentic AI Chasm

Thus, HDS has laid the groundwork and constructed the first bridges across the vast chasm of Agentic AI. Now it must follow with compelling narratives of organizations scaling their deployments and capturing that elusive value. Suffice it to say, HDS is not alone in that regard, but it has provided a much more nuanced narrative depicting a realistic picture of both the challenges but also the immense opportunities of that transformational journey.