It’s Confirmed: The Microsoft Power Platform Is 2025’s 1990s Access Database
You’ve seen this movie before. In the 1990s, Microsoft Access promised empowerment. Anyone who was close to the work could build what they needed, when they needed it. Departments moved faster than the central IT backlog. Executives applauded. Then reality hit. Network-drive databases multiplied. VBA macros became business logic. Data lived everywhere and nowhere. Years later, corporations quietly paid for the cleanup, migrations to SQL Server, scripted extractions to repair data drift, and security retrofits after the audit call. It wasn’t malicious. It was momentum.
We are in the same place again. Only this time the tooling is bigger, easier, and turbocharged by AI. The Power Platform, powered by Copilot, Power Apps, Power Automate, Power BI, Power Pages, and Dataverse, puts the promise of “build it now” into every business unit’s hands and it works. That’s the point. It works so well that it scales the very behavior we’ve failed to govern for thirty years. If Access was a spark, Power Platform is a forest fire. And the industry is cheering it on as a triumph of “citizen development,” as if the phrase alone absolves leaders of responsibility for the outcomes.
I am not arguing against empowerment. I’m arguing against amnesia. We act surprised when today’s autonomy creates tomorrow’s chaos, even though we’ve been paying that price since the last century.
The Déjà Vu We Don’t Want to See
Access delivered speed. It also delivered thousands of orphaned databases, each embodying a snapshot of how a process worked at one moment in time. Then the processes evolved and the databases didn’t. Database owners left. The logic stayed. Auditors arrived and the scramble began. Who owns this? What does this macro do? Why does this table differ from the finance system by two percent every month?
Power Platform is Access with a distribution network. It ships not just forms and tables, but end-to-end workflows, integrations, dashboards, AI-assisted generation, and one-click connectors to core systems. When a sales leader builds a canvas app on Friday, a flow on Saturday, and a dashboard on Sunday, they are solving a real need. They are also creating an untracked dependency chain across identity, data, permissions, and uptime. Multiply that by hundreds of teams. Then add Copilot prompting non-developers through “good enough” architecture in minutes. The result is speed at the edges that outpaces discipline at the core of successful IT management.
Why Does Speed Keep Winning Over Structure?
We’ve built organizations that reward velocity. Budgets are quarterly. Targets are immediate. The loudest pain gets solved first. A team that ships a Power App to eliminate email spreadsheets will look like heroes compared to a platform team requesting three months to define data contracts, role-based access, and lifecycle management. One path has an immediate and visible impact. The other path prevents uncertain and invisible risk. Guess which one gets funded?
Leaders rationalize the trade-off. “We’ll fix it later.” Later never arrives. By the time the consequences become visible through a failed control, data exposure, or critical workflow outage, the people who approved the quick win have moved on. The team left behind inherits a landscape of brittle flows, duplicated logic, and dashboards that disagree. They didn’t create it, but they are accountable for cleaning up the issues.
The Human Pattern Driving Technical Debt
The problem was never a tooling issue. It’s a behavior issue. Access amplified human behavior in the 1990s. Power Platform is amplifying it now.
We prefer autonomy to alignment because alignment is slow and uncomfortable. We prefer local optimization because it’s empowering and feels like progress, even if the results increase total system complexity. We trust what we can see. We underinvest in what we can’t see. We mistake one team’s speed for the organization’s speed. And we talk about “governance” as if it’s simply a form to complete rather than a set of design decisions that shape how work gets done.
When technology lowers the friction to build, the instinct to build it yourself scales. A manager who wouldn’t submit a ticket to IT will try a Copilot prompt. A team that never received a data model will create its own schema. A department frustrated with integration timelines will authorize a connector with broad permissions. None of these decisions are irrational in isolation. Together, they illustrate how sprawl becomes the default architecture.
The Future Cost You’re Creating Today
Power Platform makes it dangerously easy to hide cost behind success. The app works. The flow runs. The dashboard updates. Everyone sees the value. Almost no one sees the liabilities accruing underneath:
Data Fragmentation - Teams create Dataverse tables and SharePoint lists that mirror core systems “just for now.” Those copies drift. Downstream reports reconcile inconsistently. The business argues, not about the plan, but the numbers. You can’t decentralize decision-making when no one trusts the source of truth.
Unowned Logic - Power Automate flows become the new macros. They codify order routing, credit checks, or policy exceptions. They rarely ship with tests, versioning, or clear ownership. When one fails, the outage isn’t just a task stuck in a queue. It’s a blind spot in how the business operates.
Identity and Permission Sprawl - Connectors make integration easy. They also create over-permissioned pathways from sensitive systems into lightly governed apps. Least privilege gives way to “whatever works.” Incident response becomes triage across dozens of unknown attack surfaces, many of which are unknown until a threat materializes.
Shadow AI - Copilocals, those ad-hoc, AI-assisted automations and apps, proliferate. They’re trained on private documents, scoped by prompts, and validated by anecdotes. When they mislead or expose data, good luck reconstructing the lineage from a chat history and a handful of flows.
Operational Fragility - The more the business relies on citizen-built workflows, the more uptime, alerting, recovery point objectives, and disaster recovery become real requirements. Most of these creations don’t have them. Your RTO isn’t “whenever someone notices.”
Audit Pain - Once regulators or internal auditors start following the threads, you discover processes that matter deeply to financial reporting or customer privacy running on platforms with inconsistent controls. The fixes don’t just cost money, they also cost the organization time. Resolution means losing control over your business timeline.
AI as an Accelerant and Force Multiplier
Access required at least a little persistence and some knowledge of VBA. Power Platform plus AI requires curiosity and a sentence. That’s the inflection. We’ve moved from “I can build if I try” to “I can build because I asked.” Purpose and process still matter. Architecture still matters. But the barrier to output has collapsed. The barrier to outcome has not.
AI lowers the bar to generate artifacts. It does not lower the bar to operate responsibly. It will help you connect the wrong systems with the right syntax. It will help you summarize the wrong data with a confident tone. It will help you automate the wrong workflow faster than you can evaluate the consequences. And because AI “assists,” it gives users a sense that the result is safe enough. That confidence is a liability when the underlying design lacks constraints.
We are not facing a small wave of new apps. We are facing a tide of automated decisions built by people who understand the business problem better than the enterprise system and the minimum necessary requirements for data integrity, security, resilience, and change management. For some unlucky organizations, the combination will cause havoc without warning.
Empowerment Without Stewardship is Abandonment
The promise of business-led technology is real. People closest to the work see opportunities that centralized teams miss. They deserve tools. They deserve autonomy. But autonomy without stewardship is just abandonment dressed up as innovation. Leaders who celebrate “citizen development” while underfunding platform engineering, data governance, and shared services are not empowering the business. They are outsourcing accountability to the least prepared part of the organization.
Stewardship isn’t a ceremony. It’s a set of constraints that make good choices the easiest choices. It’s the whole organization knowing paved roads are the faster route to success than going off-road through a vacant lot filled with unknwon debris, potholes, and poison ivy. It’s an environment strategy that makes “dev–test–prod” default, not aspirational. It’s access control patterns that make least privilege simpler than over-permissioning. It’s data contracts that put reconciliation into the design, not the month-end crisis. It’s platform product management that measures adoption and debt with the same intensity. It’s sharing this knowledge across the entire organization so business leaders also endorse and adopt the messages.
If that sounds slower, remember what you’re comparing it to. You’re not comparing it to a mythical Agile IT team that ships the perfect solution tomorrow; your solution team is a combination of experienced and inexperienced members who are learning as they build. You’re comparing it to an organization that ships hundreds of small solutions today and spends the next three years repairing them in fragments, outages, and re-platforming. The goal is maximizing learning during build and minimizing repair.
What Disciplined Empowerment Looks Like in Practice
You can’t prevent sprawl with policy alone. You shape it with defaults and incentives. In every enterprise where low-code has worked at scale, the pattern is similar.
The platform is treated as a product. That means a roadmap, SLOs, and a team accountable for enablement and reliability. It means publishing reference architectures as living code. It means curating connectors and patterns so the safe path is also the fastest. When teams choose the platform, they inherit good decisions. When they go off-road, it’s an explicit exception with a refactoring plan.
Data is designed, not discovered. Authoritative sources are defined. Contracts are versioned. Access is delegated through roles mapped to business responsibilities, not ad-hoc approvals. If a team needs a new attribute, they know where to request it and how it propagates. The platform measures drift and flags copies before they become systems of record by accident.
Ownership is explicit and durable. Every app and flow has an owner and a sponsor. Ownership survives job changes. Orphans are not tolerated. Decommissioning is a success metric, not an afterthought. The platform maintains an inventory that can answer a simple question within minutes: What breaks if this table changes?
AI is bounded by design. Guardrails exist for prompt security, data loss prevention, and model selection. Business users can create with AI, but they should create inside fences. If a Copilot builds a flow that touches sensitive systems, that flow goes through an approval that adds value:
Threat modeling
Test harnesses
Monitoring
Approval is not a checkbox. It’s a design review that teaches.
Funding recognizes the whole lifecycle. If you fund only the build, you get a garden of annual flowers, pretty for a season, dead by winter. If you fund operations, governance, and continuous improvement, you get perennials, systems that return value year after year. CFOs will pay for the cleanup later. They always do. They rarely get the ROI they thought they bought.
“Citizen developers” are not a category. They’re colleagues who solve problems. Treat them with respect. Train them in fundamentals. Make it easy to do the right thing. Celebrate solutions that align with the platform’s standards. Retire those that don’t. People respond to clarity and support more than slogans.
The Governance Gap Leaders Create
The uncomfortable truth: most governance gaps are leadership gaps in disguise. We blame shadow IT while designing a budget process that makes shadows rational. We demand standardization while rewarding the teams that bypass it to hit short-term goals. We posture about risk while quietly accepting it in the name of speed.
If you are an executive sponsoring Power Platform adoption without funding platform engineering, you are buying a lottery ticket with someone else’s budget. If you are a CIO who delights in adoption numbers without asking about ownership, telemetry, and decommissioning, you are measuring the wrong thing. If you are an architect who dismisses low-code as beneath you, you are allowing the architecture to happen without you.
The past is not a prophecy. It’s a pattern. We can break it. But only if we admit we keep choosing it.
The Cost of Ignoring the Warning
What happens if we keep going? It won’t be one catastrophic failure. It will be a thousand tiny ones. A silent flow stops after an API limit change. A misconfigured connector exposes sensitive data to a contractor’s personal device. A dashboard drives a decision off a stale field. A compliance audit turns into a treasure hunt through undocumented environments. Morale drops as central teams spend their time in endless “stabilization” projects instead of building the future.
Then come the big moves. A “strategic re-platforming” to consolidate and control. A “data trust initiative” to reconcile source-of-truth wars. A “process modernization” to unwind local optimizations. You will complete these programs. You will declare victory. And within two years, the new tooling will be in the same position if you haven’t changed the incentive structure that created the sprawl.
AI ups the ante. Today’s Copilots make it simple to generate flows and apps. Tomorrow’s agents will chain decisions and act autonomously across systems. If you’re not designing for provenance, reversibility, and bounded action now, you will learn those terms after the fact. The Armageddon of cleanup is not melodrama. It is a steady accumulation of ungoverned automation meeting the real world of audit, regulation, and customer trust.
Denial, Not Speed, is the Enemy
We need speed. Markets punish indecision. Customers reward responsiveness. Teams do better work when they can shape their tools. None of that requires pretending that the past didn’t happen. None of it requires we celebrate results without owning consequences.
The mature posture is simple. Move fast where it’s safe. Move deliberately where it isn’t. Invest in the boring parts because they make the exciting parts sustainable. Hold leaders accountable for downstream effects, not just upfront demos. Treat platform and governance as enablers whose job is to make good behavior the shortest path to impact.
The Power Platform is powerful. That’s why the comparison to Access lands. Empowerment is not the risk. Unaccountable empowerment is. If we repeat the 1990s with AI in the mix, the scale and speed of the sprawl will dwarf what came before. It won’t be a few network drives and a shelf of binders. It will be an organization whose core processes depend on automations no one can see, data no one trusts, and agents no one can fully control.
A Pragmatic Path Forward
Start by telling the truth about where you are. Inventory what exists. Not just the sanctioned apps, but the flows and lists that run the business more than anyone admits. Map ownership. Validate who has access to what. Identify the processes that matter to financials, customer experience, and regulatory obligations. Those become your Tier 1. Treat them like product, not prototypes.
Design your paved road. Pick the patterns you want replicated. Deliver them as templates with defaults set to safe. Bake in logging, alerts, and testing. Make environments free, fast, and tagged with cost centers. Instrument everything. The telemetry is not surveillance. It’s how you prevent the “mystery outage” from ruining your quarter.
Establish a principle you are willing to defend: if an app or flow touches Tier 1 processes or sensitive data, it must be onboarded to the paved road. No exceptions. If a team wants to go faster on the edges for low-risk cases, give them latitude and support. But every step toward the center increases the requirement for discipline. That’s not bureaucracy. That’s physics.
Fund the whole thing. Not a workshop. Not a slide deck. A platform product with engineers, product managers, SREs, and data stewards. Measure success in reduced time-to-value for compliant builds and reduced incidents from non-compliant ones. Publish the numbers. Celebrate the migrations and the retirements.
Teach. Not just tool tips. Mental models. Why identity primitives matter. Why data contracts are the antidote to reconciliation theater. Why automation without observability is a liability. Make it accessible. Make it practical. People rise to clear expectations.
And, yes, say no. If someone wants to wire a sensitive workflow through a personal connector with broad permissions, decline. Offer a better path. Back your platform team when they enforce the boundary. If your culture punishes the people who keep you safe, you will get the safety you deserve.
The Uncomfortable Question
Are you measuring adoption because it makes you feel modern, or are you designing stewardship because you intend to be durable? The first is marketing. The second is leadership. One earns likes on internal chat. The other wins when the audit committee calls.
We can break the pattern. We can have empowerment and integrity. We can move quickly and build structures that survive reorgs, audits, and turnover. But we have to choose it. We have to fund it. We have to demand it from ourselves before we demand it from our teams.
The Verdict
It’s confirmed. The Microsoft Power Platform is 2025’s 1990s Access database. Not because the technology is flawed, but because our instincts are unchanged. Left alone, we will repeat the same behaviors with better tools and worse consequences. Led with intent, we can harness the power without inheriting the pain.
Leaders decide which story we tell five years from now: a proud case study of durable transformation or another quiet budget line for cleanup. Choose stewardship. Choose paved roads. Choose to be measured by what still works after the people who built it have moved on.
We do not need another decade of rediscovering why governance matters. We need a decade of building it in from the start. Are we building the future or repeating the past?
