Stop chasing hype, start harnessing hope, and make real headway to succeed with AI

Enterprise leaders are debating whether agentic AI is hype, hope, or something they can scale across their firms. At an HFS roundtable in New York, executives from financial services, pharma, consumer goods, and telco agreed on two points: the buzz has outrun reality, and the real barrier isn’t tech—its process, culture, and incentives that will unlock the power of AI.

Process debt is the real bottleneck, not the models

The consensus at the roundtable was clear: firms can’t scale AI until they re-architect broken workflows. HFS Pulse data confirmed this, citing process debt as the main barrier. Leaders stressed that chasing shiny copilots or proliferating AI platforms without fixing processes simply automates dysfunction. Survey responses from Global 2000 decision-makers (see Exhibit 1) echoed the same sentiment. If the processes are broken, flawed, or not ready for AI, all the work on cleaning data, training people, and investing in tech will not yield the desired results.

Exhibit 1: The real challenge when driving for AI at scale is process debt

Sample: 305 major enterprise decision makers
Source: HFS Research Pulse, 2025

The five practices you adopt now will shape your AI adoption efforts

While many believe AI is overhyped, the development of skills and expertise in using AI to improve outcomes across teams instills hope for larger, scalable solutions. However, to do so, leaders must rethink how work is done. Confining AI to conversations around “which tech do people want to use?” instead of “what processes can we fix now to prove AI’s value?” limits success to point solutions rather than exploring how their organizations can create better experiences through process improvements.

The most effective AI tool strategy emerging from our discussions is to focus on picking two to three AI platforms and committing to them. Choosing a few rather than attempting to adopt the latest breakthrough will yield longer and more sustainable outcomes. Once the AI tools are selected, the real work begins. Five practices stood out in our discussions, grounded in real-world experiences and observations that act as a blueprint for other enterprises to drive their AI programs from hype to reality. These include:

1. Be hopeful but hard-nosed: Leaders in financial services described agentic AI as “hopeful” and “already delivering value in parts” but warned about overpromised autonomy in regulated contexts, noting that investments demand trust, control, and explainability. Successful implementation requires tight oversight. Some firms, for instance, use AI to automate customer-facing tasks and have larger organizational teams to track privacy and fraud.
2. Support pragmatic patterns: A large telco shared that it is experimenting with voice as a master interface to coordinate multiple agents, only to encounter real questions about authorization delegation, agent-to-agent communication, and marketplace versus in-house orchestration. Success depends on documenting both existing and the desired state of information flow across agents and people, revealing the gaps to be filled in a pragmatic manner.
3. Incorporate reality checks into production: BFS practitioners shared examples from payments tracking and code modernization, where staged reasoning and engineered workflows (not a “just add the LLM” button) separated success from slide decks.
4. Realize you’ll have workforce bifurcation: Akkodis, in particular, shared that it sees AI as both a monumental opportunity and a threat to talent markets. It is co-investing with partners like Salesforce, leveraging each firm’s expertise, labor, and agents to deliver projects based on automation, AI agent task automation, and multi-agent orchestration. By tapping into skills across an ecosystem, it reduces the skill and process knowledge gaps for all involved organizations.
5. Stop agentic washing: The group was aligned on layering their approach, starting with which processes benefit from task agents and co-pilots and then moving to process agents, multi-agent systems, and beyond. They cautioned that much of what is labeled “agentic AI” today is merely more innovative RPA or copilot tools that focus on automation and productivity.

Agentic washing, the mislabelling of AI solutions, is stalling success —fix this for any hope of success

Most enterprises are deploying automation, copilots, and purpose-built AI agents but are labeling them agentic AI, which only muddies expectations. Attendees discussed how overpromising autonomy and agency in AI can undermine trust and delay outcomes.

According to HFS’ research on agentic washing, maturity must be staged—from copilots to process agents to multi-agent systems. Enterprises that skip this risk hype, fatigue, and regulatory backlash. Exhibit 2 outlines the five stages of AI, with agentic AI being more complex; so make sure your team knows what AI stage they really need.

Exhibit 2: Be wary of calling everything agentic AI; there are discrete value points across all five levels of AI maturity

Source: HFS Research, 2025

Making headway on your AI programs requires significant change management across people, processes, and technologies

If you’re looking to maximize the benefits of your AI efforts, the panel’s insights are worth incorporating into your strategy. Here are five areas you should focus on:

Exhibit 3: The five areas to focus on to maximize the benefits of your AI efforts

Source: HFS Research, 2025

The Bottom Line: The only way to make real headway with agentic AI is to rethink how the business must evolve and how people work. Accept AI as an operating model transformation, not a technology evolution.

Agentic AI won’t scale based on hype or half-measures. To move from the buzz to business value, enterprises must fix process debt, bake trust into design, and rewire incentives. Treat agentic AI as the later stage in a fundamental operating model shift, not just another IT program. Ignoring this will jeopardize hope, leading to its collapse under fatigue.