The Five-Step Enterprise AI Journey

As we come to the end of 2025, it’s safe to say that no technology innovation has been as omnipresent and the topic of as much discussion and debate as Artificial Intelligence (AI).

Since bursting onto the scene in November 2022 with the launch of OpenAI’s ChatGPT, companies in practically every market and industry have been grappling, experimenting, and evaluating AI tools and applications across their enterprises. Many of these organizations have been drawn to AI’s siren song of increased operational efficiency and lower costs, driven by greater automation and autonomous processes.

This is especially true in the financial services and insurance industries, where AI solutions are being leveraged to expedite insurance claims, interact with account holders and policyholders, and help companies make better, more informed decisions on everything from policy pricing to setting competitive mortgage rates.

Today, this demand and desire for AI, and the benefits it enables, have made those two little letters magical drivers of everything from economic growth in the technology and microprocessor industries to the accelerated construction of data centers across the country. However, while the fervor and excitement over AI are certainly warranted, they’re not new.

Practically every disruptive technology that has entered the marketplace in the past fifty years has been met with similar enthusiasm and interest. I remember the quantity of articles and magazine covers dedicated to the personal computer, the world wide web, and “The Cloud” being similar to those devoted to AI.

I mention the Internet and cloud computing for a very good reason – similar to AI, these technologies started with individual consumers, users and developers while organizations took time to figure out if and how to take advantage. But eventually, these innovative technologies fundamentally changed the way business was conducted and opened the door to new growth and revenue opportunities.

It’s important to note before we start looking at what I would consider a relatively universal AI maturity model that no two companies will have the exact same path. Each company will adopt AI at different rates and follow different steps. It’s also important to note that different departments and organizations within a company could adopt AI more quickly than others and could develop and utilize more advanced AI applications before others even get out of the AI adoption starting blocks.

However, based on my experience with past disruptive technologies – including cloud computing – here is how I see the vast majority of AI adoption journeys playing out. It’s a journey that can be viewed in five stages or levels of maturity:

Step One: Chat and Generative AI

Every journey begins with a single step, and this is the step where most organizations likely find themselves.

The first widely available GenAI solution was launched just about three years ago, and these “chat” AI agents – where users ask AI simple questions or ask AI to generate content for them – are widely being used in people’s day-to-day lives.

Considering the wide adoption of AI in their personal lives (Even my 85-year-old mom is using ChatGPT!), it should come as no surprise that people are also starting to use these agents in their professional lives.

Encouraged or sanctioned by their organizations or not, employees are most likely using these agents to generate job descriptions to aid the hiring process, write blog posts for marketing, and even vibe code their own applications. Ultimately, this is where the AI adoption process starts within many organizations.

This AI use may be driven by employees – individuals who want to use the same AI agents and tools in their jobs that they’re using at home for vacation and meal planning – but it can still have ramifications on the larger organization, as a whole. That’s why it’s essential that organizations begin to establish AI governance and guardrails at this maturity level or even earlier.

The risks of AI are well documented at this point. But data security remains one of the largest. It’s imperative at this stage for organizations to begin answering difficult AI questions about what data can be used within certain AI models, what intellectual property or personally identifiable information is appropriate to share with an AI agent, and establish other baseline guardrails. This guidance will be essential throughout an organization’s entire AI journey.

Step Two: Agentic Workflows

When enough employees across an organization use “chat” or GenAI, there begins to be commonality in their usage. Users begin asking AI the same questions, generating the same content, and performing the same task. This opens the door to move beyond individual or small team productivity to a universe of more advanced AI automation – agentic AI.

AI agents are reusable automations that effectively execute a repeatable task. Instead of simply answering a question or generating content, they will actually do the work to fulfill an employee’s request. For example, let’s look at the hiring process. While “chat” or GenAI will create a job description for an employee, it stops there. An AI agent could generate a job description, upload it to an HR system for approval, and then – if approved – post on job boards and hiring Websites.

Hiring is just one example of a process or task that AI agents could handle. Financial services companies could also use them to automate government compliance and regulatory reporting, as well as credit risk and underwriting processes.

Step Three: AI Operations at Scale

Once the genie is out of the proverbial bottle and employees are leveraging AI across the enterprise, organizations can begin to evolve AI usage from individual pockets of maturity to organization-wide access governed by established best practices. In this stage, organizations move from individual power users making one-off agents to having the established processes, systems, and repositories needed to manage “chat” interactions and agents at scale for broad positive business impact.

Much like what we witnessed with the cloud – where individual developers were deploying and managing their own cloud resources – for organizations to deploy AI effectively at scale, they need to stop treating AI prompts and agents as pets and start thinking of them as cattle. This means establishing repositories for storing, managing, and centrally deploying effective AI agents.

Also, as prompts and agents become integral to an organization’s operations, safeguards need to be put in place to ensure they’re continually trained, upgraded, enhanced, and maintained. This means ensuring that no single employee is responsible for the care and feeding of an AI agent and ensuring that the task of maintaining what is effectively a fleet of AI solutions is assigned to a team of capable AI professionals.

Step Four: Developing AI Applications

Once the organization has adopted AI agents and agentic workflows and is stitching them together to automate processes with a level of proficiency and AI literacy – all with governance and guardrails in place – it’s time to take AI to the next level. In this step, organizations leverage baseline employee AI literacy and knowledge along with the proprietary data, expertise and knowledge within their own four walls to build THE KILLER AI application.

This could involve creating a unique, proprietary large language model (LLM) leveraging data exclusive to the organization. For example, an insurance company could create an application that does personalized risk modeling or “no touch” claims processing using that company’s decades of unique data.

Ultimately, this step is where the real value and benefit of AI is unlocked. If AI is only ever used to increase individual and team productivity, it will never justify the level of investment markets are making in it.

However, when AI is used to create new revenue streams and business opportunities, and to unlock new markets, it will be worth the investment.

Step Five: The Future (AGI)

Then there is the theoretical concept of Artificial General Intelligence (AGI), where AI eventually gains processing and cognitive abilities on par with human intelligence. This is where the AI industry is looking to go and is the end goal of the trillions of dollars of investment being made in AI research and development.

The industry is setting the table for AI that will have the broad understanding, common sense, and ability to reason that humans have – all of which are characteristics of AGI.

While we’re not there yet, organizations need to put in the work and advance through this adoption journey if they’re going to be ready for AGI when it becomes a reality. This means increasing the AI literacy of employees, having established AI governance and guardrails, developing organizational AI best practices, and establishing AI systems, processes, and repositories that enable the operational use of AI at scale.

Walking the Path

As I discussed, not every organization will follow this maturity curve or linear adoption journey. Not all of them will advance at the same pace. And many will have different divisions and departments within their organization at different stages at different times – some will inevitably leap ahead. This is inherent in the adoption of all disruptive technologies. And, much like all other disruptive technologies, there will be challenges that arise along an organization’s adoption path.

In a future article, my associate, David Kang, will explore some of the pitfalls and problems that arise as organizations move forward in their AI journey, and how innovative new solutions are being introduced to help make their progress easier.

The author, Peter Scott, is the President and Chief Growth Officer at Holokai. He was previously part of the team that launched DivvyCloud, a solution focused on multi-cloud security for commercial enterprises.

Originally published on Financial Technology Today on January 13, 2026.