blog

Smart People, Dumb Data: The Hidden System That Keeps You Guessing

Written by Tony Wood | Apr 22, 2026 12:57:52 PM

I’m writing this because most organisations say they want to be data-driven.

They recruit smart people, invest in analytics, and talk about evidence-based decision making. Yet when I walk into a large company, I often see the same pattern.

Talented people struggle with basic data tasks. They cannot answer fundamental questions about their own business. Why is this still happening?

Here’s the thing. It’s not a lack of intelligence or motivation. The real issue is structural.

Over the years, most large organisations, and the software as a service (SaaS) vendors they rely on, have kept people away from raw data. Everything is hidden behind polished front-end interfaces.

Dashboards, forms, and reports create the illusion of access. But they rarely allow users to see or shape the underlying data.

If you grew up using SaaS tools, you probably never learned SQL (Structured Query Language), database design, or how data actually flows. You were handed a set of features and told to get results.

But if you never saw the data, you were never taught how to think about it, question it, or challenge the assumptions baked into those tools.

This is not an accident. The business model of many SaaS providers depends on keeping customers inside their environment.

True data ownership is discouraged. Exporting or connecting to the raw database is made hard, expensive, or sometimes impossible.

As a result, the skills needed to work directly with data do not develop. People become expert at navigating menus, not at understanding what is underneath.

It’s common to blame users for not being data literate. In reality, the systems have been designed to suppress that literacy.

If you have never had the opportunity to work with the data layer, how could you develop the right instincts?

There is another way.

Agentic-first tools and MCP (Multi-Component Platform) first environments, like Recall or Agent.io, take a different approach. They empower users, or their AI agents, to work directly with the data.

Instead of hiding everything behind a front-end, these platforms facilitate direct access, exploration, and automation at the data layer. You can build low-friction automation, iterate on workflows, and actually see how your data is structured and moves.

This shift matters.

When people, or agents, can interact directly with data, they develop intuition, ask better questions, and spot errors earlier. Data literacy becomes a natural outcome of working with the right tools, not something that must be taught separately.

So what does this mean for leaders?

Stop blaming your teams for not understanding data. Look at the systems and vendors you have chosen.

Ask yourself:

  • Are we empowering people to work with data, or hiding it from them?
  • Do our tools facilitate agentic workflow, or keep us locked into rigid interfaces?
  • Are we building true community-led innovation, or simply buying another dashboard?

If you want data-literate teams, you need to create the conditions for data literacy to flourish.

That means choosing tools and practices that expose, not hide, the data layer. It means rewarding curiosity, experimentation, and hands-on learning.

Truth is, this stuff is genuinely hard. But the first step is to stop blaming the user and start fixing the system.

The next generation of agentic-first tools is already making this possible. The rest is up to us.

Agentic (AI created research and content)

This section adds context using only the validated quotes provided, so you can see how the market is shifting around the same problem: data access, lock-in, and the economics of SaaS.

What Changed In 2026 (And Why Leaders Should Care)

If your organisation still treats “data literacy” as a training problem, you may be solving the wrong thing.

One reason this is surfacing now is that AI agents change how software gets consumed. The old model assumed more employees meant more seats and more licences. That shaped product design, pricing, and how much access you got to your own data.

A recent LinkedIn post captured the shift bluntly:

“In February 2026, the software sector lost roughly $2 trillion in market cap in less than 30 days. No recession. No rate shock. No accounting scandal. The market simply decided that selling software by the seat — the model that defined two decades of tech — is structurally broken. Here's the core of what's happening: SaaS was built on a linear equation: more employees = more licenses = more revenue for vendors. It was elegant, predictable, and produced gross margins of 70–85% that no other industry could touch. AI agents break that equation at the root. They don't need seats. They operate via APIs, execute tasks autonomously, and a single agent can replace what previously required dozens of individual licenses.”

https://www.linkedin.com/posts/datapills_in-february-2026-the-software-sector-lost-activity-7448027925230030848-MmiR

Leadership implication:

  • If agents operate via APIs (Application Programming Interfaces), then the “front-end only” experience starts to look like a constraint, not a convenience.
  • Your competitive edge moves towards data access, permissions, and workflow design, not who has the nicest dashboard.

The Lock-In Trap (And The Silo You Accidentally Built)

A lot of teams try to fix the mess by buying another platform to “unify” data. That can help. It can also create another layer of copying and another point of lock-in.

Chris O’Neill put it like this:

“So, you're just another customer data platform? No. And this misconception is costing companies millions. Here's what most people think a CDP does: scrape data from different places, copy it into a separate pool, and call it a 'single source of truth.' Sounds logical, right? Customers are important, so having all their data in one place must be better. But there’s a problem: you're creating yet another silo. More copying, more transfers, more delays, more lock-in with a single vendor. At GrowthLoop, we don't copy your data anywhere. We’re an intelligent data and AI platform that sits on top of your data cloud where it already lives. No duplication, no lag time, no vendor lock-in.”

https://www.linkedin.com/posts/croneill_so-youre-just-another-customer-data-platform-activity-7351275926841683968-_9KH

Leadership implication:

  • If your “single source of truth” requires another copy, you may be building a new truth that drifts over time.
  • If you want evidence-based decision making, you need fewer hand-offs, fewer extracts, and fewer hidden transformations.

A Balanced View: This Is A Shift, Not A Collapse

It’s easy to read all this and assume SaaS is dead. That is not the most useful conclusion.

Dirk Wakeham offered a more grounded framing:

“I’ve been asked that question multiple times in the past 48 hours. The short answer: no. The better answer: software is evolving — fast. Public SaaS valuations have been under pressure. AI has introduced real uncertainty. When the marginal cost of 'intelligence' drops toward zero, investors naturally ask: 𝑊ℎ𝑎𝑡 ℎ𝑎𝑝𝑝𝑒𝑛𝑠 𝑡𝑜 𝑎𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛-𝑙𝑎𝑦𝑒𝑟 𝑠𝑜𝑓𝑡𝑤𝑎𝑟𝑒? But history suggests we’re at an inflection point, not a funeral. In prior platform shifts — client/server, cloud, mobile — the narrative was similar. Incumbents were questioned. Multiples compressed. And then a new category of value creation emerged.”

https://www.linkedin.com/posts/dirkwakeham_𝗜𝘀-𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲-𝗗𝗲𝗮𝗱-ive-been-activity-7428519434870390785-VJMY

Leadership implication:

  • You do not need to “rip and replace” everything.
  • You do need to re-check your assumptions about how work gets done, and who (human or agent) needs access to what.

What You Can Do This Quarter (Leadership-Level, Low Drama)

You can respond without turning this into a multi-year transformation programme.

Start with a simple system audit:

  • Where is the raw data?
  • Name the systems of record.
  • Name the systems that copy data elsewhere.
  • Who can query it, and how?
  • Humans through dashboards.
  • Analysts through SQL.
  • Agents through APIs.
  • Where do definitions drift?
  • Same metric, different numbers across tools.
  • Manual exports in spreadsheets.
  • What is locked in by contract or architecture?
  • Cost to export.
  • Cost to integrate.
  • Restrictions on API access.

Then choose one workflow to rebuild as an agentic workflow:

  • Pick a process with repeated questions, repeated reports, and repeated manual steps.
  • Make the data path explicit end-to-end.
  • Reduce copying.
  • Increase direct access with clear permissions and audit trails.

If you do this well, you stop treating data literacy as a personal virtue.

You treat it as an outcome of a system that is designed to be learnable, inspectable, and improvable.

Links

No validated links were provided in this run.

Quotes

  • “In February 2026, the software sector lost roughly $2 trillion in market cap in less than 30 days. No recession. No rate shock. No accounting scandal. The market simply decided that selling software by the seat — the model that defined two decades of tech — is structurally broken. Here's the core of what's happening: SaaS was built on a linear equation: more employees = more licenses = more revenue for vendors. It was elegant, predictable, and produced gross margins of 70–85% that no other industry could touch. AI agents break that equation at the root. They don't need seats. They operate via APIs, execute tasks autonomously, and a single agent can replace what previously required dozens of individual licenses.”

https://www.linkedin.com/posts/datapills_in-february-2026-the-software-sector-lost-activity-7448027925230030848-MmiR

  • “So, you're just another customer data platform? No. And this misconception is costing companies millions. Here's what most people think a CDP does: scrape data from different places, copy it into a separate pool, and call it a 'single source of truth.' Sounds logical, right? Customers are important, so having all their data in one place must be better. But there’s a problem: you're creating yet another silo. More copying, more transfers, more delays, more lock-in with a single vendor. At GrowthLoop, we don't copy your data anywhere. We’re an intelligent data and AI platform that sits on top of your data cloud where it already lives. No duplication, no lag time, no vendor lock-in.”

https://www.linkedin.com/posts/croneill_so-youre-just-another-customer-data-platform-activity-7351275926841683968-_9KH