When Everyone Can Build: The New Enterprise Challenge of AI Adoption
By Anant Kabra, VP, Consulting Solutions, Catalyst Solutions
AI is no longer just a tool employees use to work faster. It is quickly becoming a way for employees to build new workflows, automate decisions, and create business applications. This shift creates a major opportunity for productivity and innovation, but it also introduces a new enterprise challenge: how do organizations scale what works without recreating shadow IT at AI speed?
At Catalyst Solutions, our decision to become an AI-native organization started with two familiar questions: What value will AI create, and how do we use it responsibly? Those questions shaped how we evaluated AI from the beginning.
From a pricing and deal modeling perspective, my first instinct was to focus on value and ROI. Many AI initiatives fall short of their expected returns, and usage costs, especially token consumption, can add up quickly. If we invest in AI, we need to show measurable gains in productivity, quality, and business outcomes.
My second concern was security and governance. In healthcare, those considerations are inseparable from the technology itself. Any solution that touches Protected Health Information (PHI) or other sensitive data must be designed with compliance and security in mind, including where AI is used, what information is shared, and how models and platforms handle that information.
Those questions remain critical, but as we began building AI-driven productivity solutions internally, another challenge that we hadn’t initially anticipated quickly emerged: What happens when everyone can build?
AI Has Changed Who Can Build Technology
One of the most significant changes AI has brought to organizations is that many more people can now build technology.
Frontier AI models and AI-assisted development tools have dramatically lowered the technical barrier to creating applications, workflows, and productivity solutions. Someone who could not have written an application a year or two ago can now describe a business need to an AI coding tool and produce a working prototype.
We have seen this firsthand at Catalyst. For example, our Talent Acquisition (TA) team wanted a better way to identify internal expertise across healthcare processes and match that expertise with customer requests. Using Claude Code and Claude Enterprise, the team worked with the AI to create a clickable prototype in just over a week. The prototype was reviewed and validated by the TA team, and then the AI was used to turn it into a deployable application.
The solution used AI selectively, where it added clear value, such as matching skills with job descriptions and providing decision support. Deterministic tools handled steps where AI was unnecessary, and the design prevented Personally Identifiable Information (PII) from being shared with AI engines.
The TA team did not need deep technical expertise to build the initial solution. That expertise was still essential, but in a different place: architecture, code review, deployment, and enterprise integration.
That distinction matters. AI has not eliminated the need for technology expertise but has changed where that expertise is needed. It has also opened the door for virtually anyone in an organization to become a technology builder.
I've Seen This Movie Before
While the possibility of anyone becoming a technology builder is exciting, it feels familiar. I lived through the technology boom of the late 1990s and 2000s, when organizations rapidly adopted new systems and gave individual departments more control over technology decisions.
Departments acquired their own systems, teams created separate databases, groups solved similar problems independently, data lived in multiple places, and every system required support.
One of the most familiar examples was the proliferation of Excel spreadsheets. Different departments would maintain their own versions of the same information, with different formulas, assumptions, and update cycles. There was no single source of truth, and resolving discrepancies became a problem in itself.
Eventually, organizations had to consolidate those systems, standardize how data was defined and organized, and create enterprise data warehouses. Depending on size and complexity, those efforts could cost hundreds of thousands or even millions of dollars.
The lesson learned was that while it is important for different departments to be able to innovate, innovation without common standards eventually creates its own cost.
AI could recreate that problem, but much faster. What we are seeing now may be the IT boom multiplied many times over.
During the IT boom, building a new application generally required technical resources, budgets, project teams, and time. Today, an employee with access to an AI platform can begin building a solution with little or no traditional development expertise.
That creative freedom is extraordinary, but it also means organizations need to think differently about how they manage technology creation.
The Risk of Fragmentation
At Catalyst, AI experimentation is already happening across the organization. Marketing is using AI to develop and refine content and media, Talent Acquisition is exploring process automation, and operations teams are looking at decision-support opportunities.
That experimentation is exactly what we want to see. The risk begins when successful experiments become business-critical solutions without a common approach for how they are built, deployed, and supported.
Several questions emerge quickly:
Are we using consistent platforms and tools across the organization?
Are solutions being developed according to appropriate security, compliance, and data standards?
Can we deploy these solutions into an enterprise environment?
Who is responsible for maintaining and supporting them once they're in production?
That last question may be the hardest. Building a solution in a weekend or a week is one thing; supporting dozens, or eventually hundreds, of solutions is something entirely different.
Demand for AI-enabled productivity improvements will likely grow faster than technical support capacity. Employees will keep identifying opportunities and building prototypes, and the business will increasingly want to put the best of them to work. Organizations need to be ready for that volume.
Governance Should Enable Innovation, Not Stop It
This is where platforms, standards, and governance become essential. But governance cannot simply mean creating another approval process that slows everyone down.
The opportunity with AI is precisely the ability to move quickly. If organizations respond by forcing every idea through a traditional technology-development process, they risk eliminating much of the productivity benefit that made AI attractive in the first place. The challenge is finding the balance between speed and control.
At Catalyst, we are working through that balance now. Our first step has been establishing an AI governance team and a common process for people across the organization to bring forward ideas. The team includes people who are further along in their AI journeys and understand both the capabilities and the rapidly evolving options in the market.
Rather than dictate a permanent set of tools or standards, the goal is to create a common direction. That includes defining standards for how AI solutions are developed, deployed, and governed; building a Center of Excellence; evaluating tooling options; applying traditional software engineering practices to shared infrastructure and data; and developing security, compliance, testing, auditing, and validation practices.
The model will continue to evolve. That's part of the reality of working with AI.
The New Enterprise Challenge: Supporting What Everyone Builds
Employees are already using AI. The question is whether your organization is prepared for what happens when they start using it to build.
AI adoption will move beyond presentations, content refinement, and information summaries. Employees will automate processes, augment decisions, and create applications that solve specific business problems.
Some of those solutions will have limited value while others will be valuable enough that the organization will want to move them into production. This is where the operating model matters. Organizations need to think about the full lifecycle of an AI solution:
Idea → Prototype → Validation → Deployment → Production → Support → Continuous Improvement
Without a clear path through that lifecycle, organizations risk creating a new version of shadow IT, only this time with a much faster pace of proliferation.
The challenge is building the people, processes, platforms, and governance required to support what gets built.
Start by Enabling the People Already Leading the Way
For organizations beginning their internal AI adoption journey, my advice is straightforward: identify the people already leading the way and enable them. They will be some of your most important resources.
Give them a forum to share what they are learning, access to the right tools, and a process for evaluating and advancing promising ideas. Establish standards that protect the organization without unnecessarily restricting experimentation.
And most importantly, plan for the flood.
The number of employees capable of creating AI-enabled solutions will only increase. The productivity gains could be significant, but only if organizations build the infrastructure and support model to capture them.
The opportunity is to create an environment where everyone can innovate, while the enterprise can safely scale what works.
That is the difference between simply adopting AI tools and becoming an AI-native organization. The next question for every AI-native organization is not whether employees will build with AI. They will. The question is whether leaders can create the discipline to know what is working, what is safe to scale, and what is truly delivering value.
-------------------------------------------------------------------------------------------------------------
Coming Soon: The Internal AI Adoption Series
This is just the beginning of the conversation. In the next two articles, we’ll explore two questions every organization pursuing AI adoption will need to answer:
Part 2: Proving AI’s Value: How to Measure ROI and Impact
How do you know AI is actually working? We’ll look at how to establish meaningful baselines, measure productivity and quality improvements, and determine whether AI is delivering greater value than traditional automation or decision-support approaches.
Part 3: Scaling AI Adoption: Building the Capability to Support What Everyone Builds
What happens when AI adoption moves from experimentation to enterprise scale? We’ll examine the operating model, governance, talent, and technical capabilities organizations need to deploy, support, and continuously improve the growing number of AI solutions being developed across the enterprise.
